2018-02-22 00:31:14 -08:00
|
|
|
|
// Generated by CoffeeScript 2.2.2
|
2010-07-25 00:15:12 -07:00
|
|
|
|
(function() {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// The CoffeeScript language has a good deal of optional syntax, implicit syntax,
|
|
|
|
|
// and shorthand syntax. This can greatly complicate a grammar and bloat
|
|
|
|
|
// the resulting parse table. Instead of making the parser handle it all, we take
|
|
|
|
|
// a series of passes over the token stream, using this **Rewriter** to convert
|
|
|
|
|
// shorthand into the unambiguous long form, add implicit indentation and
|
|
|
|
|
// parentheses, and generally clean things up.
|
|
|
|
|
var BALANCED_PAIRS, CALL_CLOSERS, CONTROL_IN_IMPLICIT, DISCARDED, EXPRESSION_CLOSE, EXPRESSION_END, EXPRESSION_START, IMPLICIT_CALL, IMPLICIT_END, IMPLICIT_FUNC, IMPLICIT_UNSPACED_CALL, INVERSES, LINEBREAKS, Rewriter, SINGLE_CLOSERS, SINGLE_LINERS, generate, k, left, len, moveComments, right, throwSyntaxError,
|
2017-04-25 07:10:42 -07:00
|
|
|
|
indexOf = [].indexOf;
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2017-06-07 07:33:46 +01:00
|
|
|
|
({throwSyntaxError} = require('./helpers'));
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Move attached comments from one token to another.
|
|
|
|
|
moveComments = function(fromToken, toToken) {
|
|
|
|
|
var comment, k, len, ref, unshiftedComments;
|
|
|
|
|
if (!fromToken.comments) {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
if (toToken.comments && toToken.comments.length !== 0) {
|
|
|
|
|
unshiftedComments = [];
|
|
|
|
|
ref = fromToken.comments;
|
|
|
|
|
for (k = 0, len = ref.length; k < len; k++) {
|
|
|
|
|
comment = ref[k];
|
|
|
|
|
if (comment.unshift) {
|
|
|
|
|
unshiftedComments.push(comment);
|
|
|
|
|
} else {
|
|
|
|
|
toToken.comments.push(comment);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
toToken.comments = unshiftedComments.concat(toToken.comments);
|
|
|
|
|
} else {
|
|
|
|
|
toToken.comments = fromToken.comments;
|
|
|
|
|
}
|
|
|
|
|
return delete fromToken.comments;
|
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
// Create a generated token: one that exists due to a use of implicit syntax.
|
|
|
|
|
// Optionally have this new token take the attached comments from another token.
|
|
|
|
|
generate = function(tag, value, origin, commentsToken) {
|
|
|
|
|
var token;
|
|
|
|
|
token = [tag, value];
|
|
|
|
|
token.generated = true;
|
2014-01-22 02:44:50 +00:00
|
|
|
|
if (origin) {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
token.origin = origin;
|
|
|
|
|
}
|
|
|
|
|
if (commentsToken) {
|
|
|
|
|
moveComments(commentsToken, token);
|
2014-01-22 02:44:50 +00:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
return token;
|
2013-02-24 19:09:01 +01:00
|
|
|
|
};
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// The **Rewriter** class is used by the [Lexer](lexer.html), directly against
|
|
|
|
|
// its internal array of tokens.
|
2017-05-01 19:26:24 -07:00
|
|
|
|
exports.Rewriter = Rewriter = (function() {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
class Rewriter {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Rewrite the token stream in multiple passes, one logical filter at
|
|
|
|
|
// a time. This could certainly be changed into a single pass through the
|
|
|
|
|
// stream, with a big ol’ efficient switch, but it’s much nicer to work with
|
|
|
|
|
// like this. The order of these passes matters—indentation must be
|
|
|
|
|
// corrected before implicit parentheses can be wrapped around blocks of code.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
rewrite(tokens1) {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
var ref, ref1, t;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
this.tokens = tokens1;
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Set environment variable `DEBUG_TOKEN_STREAM` to `true` to output token
|
|
|
|
|
// debugging info. Also set `DEBUG_REWRITTEN_TOKEN_STREAM` to `true` to
|
|
|
|
|
// output the token stream after it has been rewritten by this file.
|
|
|
|
|
if (typeof process !== "undefined" && process !== null ? (ref = process.env) != null ? ref.DEBUG_TOKEN_STREAM : void 0 : void 0) {
|
|
|
|
|
if (process.env.DEBUG_REWRITTEN_TOKEN_STREAM) {
|
|
|
|
|
console.log('Initial token stream:');
|
|
|
|
|
}
|
|
|
|
|
console.log(((function() {
|
|
|
|
|
var k, len, ref1, results;
|
|
|
|
|
ref1 = this.tokens;
|
|
|
|
|
results = [];
|
|
|
|
|
for (k = 0, len = ref1.length; k < len; k++) {
|
|
|
|
|
t = ref1[k];
|
|
|
|
|
results.push(t[0] + '/' + t[1] + (t.comments ? '*' : ''));
|
|
|
|
|
}
|
|
|
|
|
return results;
|
|
|
|
|
}).call(this)).join(' '));
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
this.removeLeadingNewlines();
|
|
|
|
|
this.closeOpenCalls();
|
|
|
|
|
this.closeOpenIndexes();
|
|
|
|
|
this.normalizeLines();
|
|
|
|
|
this.tagPostfixConditionals();
|
|
|
|
|
this.addImplicitBracesAndParens();
|
2017-08-29 17:17:56 -04:00
|
|
|
|
this.addParensToChainedDoIife();
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
this.rescueStowawayComments();
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
this.addLocationDataToGeneratedTokens();
|
2017-06-07 07:33:46 +01:00
|
|
|
|
this.enforceValidCSXAttributes();
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
this.fixOutdentLocationData();
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
if (typeof process !== "undefined" && process !== null ? (ref1 = process.env) != null ? ref1.DEBUG_REWRITTEN_TOKEN_STREAM : void 0 : void 0) {
|
|
|
|
|
if (process.env.DEBUG_TOKEN_STREAM) {
|
|
|
|
|
console.log('Rewritten token stream:');
|
|
|
|
|
}
|
|
|
|
|
console.log(((function() {
|
|
|
|
|
var k, len, ref2, results;
|
|
|
|
|
ref2 = this.tokens;
|
|
|
|
|
results = [];
|
|
|
|
|
for (k = 0, len = ref2.length; k < len; k++) {
|
|
|
|
|
t = ref2[k];
|
|
|
|
|
results.push(t[0] + '/' + t[1] + (t.comments ? '*' : ''));
|
|
|
|
|
}
|
|
|
|
|
return results;
|
|
|
|
|
}).call(this)).join(' '));
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return this.tokens;
|
2010-10-02 18:57:16 +09:00
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Rewrite the token stream, looking one token ahead and behind.
|
|
|
|
|
// Allow the return value of the block to tell us how many tokens to move
|
|
|
|
|
// forwards (or backwards) in the stream, to make sure we don’t miss anything
|
|
|
|
|
// as tokens are inserted and removed, and the stream changes length under
|
|
|
|
|
// our feet.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
scanTokens(block) {
|
|
|
|
|
var i, token, tokens;
|
2017-04-06 10:06:45 -07:00
|
|
|
|
({tokens} = this);
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
i = 0;
|
|
|
|
|
while (token = tokens[i]) {
|
|
|
|
|
i += block.call(this, token, i, tokens);
|
2012-04-10 14:57:45 -04:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return true;
|
2010-11-11 21:48:08 -05:00
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2017-05-29 18:29:45 -06:00
|
|
|
|
detectEnd(i, condition, action, opts = {}) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
var levels, ref, ref1, token, tokens;
|
2017-04-06 10:06:45 -07:00
|
|
|
|
({tokens} = this);
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
levels = 0;
|
|
|
|
|
while (token = tokens[i]) {
|
|
|
|
|
if (levels === 0 && condition.call(this, token, i)) {
|
|
|
|
|
return action.call(this, token, i);
|
|
|
|
|
}
|
|
|
|
|
if (ref = token[0], indexOf.call(EXPRESSION_START, ref) >= 0) {
|
|
|
|
|
levels += 1;
|
2018-01-31 06:33:17 -08:00
|
|
|
|
} else if (ref1 = token[0], indexOf.call(EXPRESSION_END, ref1) >= 0) {
|
|
|
|
|
levels -= 1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
}
|
2017-05-29 18:29:45 -06:00
|
|
|
|
if (levels < 0) {
|
|
|
|
|
if (opts.returnOnNegativeLevel) {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
return action.call(this, token, i);
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
i += 1;
|
2012-04-10 14:57:45 -04:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return i - 1;
|
2012-04-10 14:57:45 -04:00
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Leading newlines would introduce an ambiguity in the grammar, so we
|
|
|
|
|
// dispatch them here.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
removeLeadingNewlines() {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
var i, k, l, leadingNewlineToken, len, len1, ref, ref1, tag;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
ref = this.tokens;
|
|
|
|
|
for (i = k = 0, len = ref.length; k < len; i = ++k) {
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[tag] = ref[i];
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (tag !== 'TERMINATOR') {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Find the index of the first non-`TERMINATOR` token.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
break;
|
|
|
|
|
}
|
2010-10-08 00:56:01 +09:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
if (i === 0) {
|
|
|
|
|
return;
|
2010-10-08 00:56:01 +09:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
ref1 = this.tokens.slice(0, i);
|
|
|
|
|
// If there are any comments attached to the tokens we’re about to discard,
|
|
|
|
|
// shift them forward to what will become the new first token.
|
|
|
|
|
for (l = 0, len1 = ref1.length; l < len1; l++) {
|
|
|
|
|
leadingNewlineToken = ref1[l];
|
|
|
|
|
moveComments(leadingNewlineToken, this.tokens[i]);
|
|
|
|
|
}
|
|
|
|
|
// Discard all the leading newline tokens.
|
|
|
|
|
return this.tokens.splice(0, i);
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// The lexer has tagged the opening parenthesis of a method call. Match it with
|
|
|
|
|
// its paired close.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
closeOpenCalls() {
|
|
|
|
|
var action, condition;
|
|
|
|
|
condition = function(token, i) {
|
|
|
|
|
var ref;
|
2017-05-29 18:29:45 -06:00
|
|
|
|
return (ref = token[0]) === ')' || ref === 'CALL_END';
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
action = function(token, i) {
|
2017-05-29 18:29:45 -06:00
|
|
|
|
return token[0] = 'CALL_END';
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
return this.scanTokens(function(token, i) {
|
|
|
|
|
if (token[0] === 'CALL_START') {
|
|
|
|
|
this.detectEnd(i + 1, condition, action);
|
|
|
|
|
}
|
|
|
|
|
return 1;
|
|
|
|
|
});
|
Fix #3597: Allow interpolations in object keys
The following is now allowed:
o =
a: 1
b: 2
"#{'c'}": 3
"#{'d'}": 4
e: 5
"#{'f'}": 6
g: 7
It compiles to:
o = (
obj = {
a: 1,
b: 2
},
obj["" + 'c'] = 3,
obj["" + 'd'] = 4,
obj.e = 5,
obj["" + 'f'] = 6,
obj.g = 7,
obj
);
- Closes #3039. Empty interpolations in object keys are now _supposed_ to be
allowed.
- Closes #1131. No need to improve error messages for attempted key
interpolation anymore.
- Implementing this required fixing the following bug: `("" + a): 1` used to
error out on the colon, saying "unexpected colon". But really, it is the
attempted object key that is unexpected. Now the error is on the opening
parenthesis instead.
- However, the above fix broke some error message tests for regexes. The easiest
way to fix this was to make a seemingly unrelated change: The error messages
for unexpected identifiers, numbers, strings and regexes now say for example
'unexpected string' instead of 'unexpected """some #{really long} string"""'.
In other words, the tag _name_ is used instead of the tag _value_.
This was way easier to implement, and is more helpful to the user. Using the
tag value is good for operators, reserved words and the like, but not for
tokens which can contain any text. For example, 'unexpected identifier' is
better than 'unexpected expected' (if a variable called 'expected' was used
erraneously).
- While writing tests for the above point I found a few minor bugs with string
locations which have been fixed.
2015-02-07 20:16:59 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// The lexer has tagged the opening bracket of an indexing operation call.
|
|
|
|
|
// Match it with its paired close.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
closeOpenIndexes() {
|
|
|
|
|
var action, condition;
|
|
|
|
|
condition = function(token, i) {
|
Fix #3597: Allow interpolations in object keys
The following is now allowed:
o =
a: 1
b: 2
"#{'c'}": 3
"#{'d'}": 4
e: 5
"#{'f'}": 6
g: 7
It compiles to:
o = (
obj = {
a: 1,
b: 2
},
obj["" + 'c'] = 3,
obj["" + 'd'] = 4,
obj.e = 5,
obj["" + 'f'] = 6,
obj.g = 7,
obj
);
- Closes #3039. Empty interpolations in object keys are now _supposed_ to be
allowed.
- Closes #1131. No need to improve error messages for attempted key
interpolation anymore.
- Implementing this required fixing the following bug: `("" + a): 1` used to
error out on the colon, saying "unexpected colon". But really, it is the
attempted object key that is unexpected. Now the error is on the opening
parenthesis instead.
- However, the above fix broke some error message tests for regexes. The easiest
way to fix this was to make a seemingly unrelated change: The error messages
for unexpected identifiers, numbers, strings and regexes now say for example
'unexpected string' instead of 'unexpected """some #{really long} string"""'.
In other words, the tag _name_ is used instead of the tag _value_.
This was way easier to implement, and is more helpful to the user. Using the
tag value is good for operators, reserved words and the like, but not for
tokens which can contain any text. For example, 'unexpected identifier' is
better than 'unexpected expected' (if a variable called 'expected' was used
erraneously).
- While writing tests for the above point I found a few minor bugs with string
locations which have been fixed.
2015-02-07 20:16:59 +01:00
|
|
|
|
var ref;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return (ref = token[0]) === ']' || ref === 'INDEX_END';
|
|
|
|
|
};
|
|
|
|
|
action = function(token, i) {
|
|
|
|
|
return token[0] = 'INDEX_END';
|
|
|
|
|
};
|
|
|
|
|
return this.scanTokens(function(token, i) {
|
|
|
|
|
if (token[0] === 'INDEX_START') {
|
|
|
|
|
this.detectEnd(i + 1, condition, action);
|
|
|
|
|
}
|
|
|
|
|
return 1;
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Match tags in token stream starting at `i` with `pattern`.
|
|
|
|
|
// `pattern` may consist of strings (equality), an array of strings (one of)
|
|
|
|
|
// or null (wildcard). Returns the index of the match or -1 if no match.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
indexOfTag(i, ...pattern) {
|
|
|
|
|
var fuzz, j, k, ref, ref1;
|
|
|
|
|
fuzz = 0;
|
2018-02-22 00:31:14 -08:00
|
|
|
|
for (j = k = 0, ref = pattern.length; (0 <= ref ? k < ref : k > ref); j = 0 <= ref ? ++k : --k) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (pattern[j] == null) {
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
if (typeof pattern[j] === 'string') {
|
|
|
|
|
pattern[j] = [pattern[j]];
|
|
|
|
|
}
|
|
|
|
|
if (ref1 = this.tag(i + j + fuzz), indexOf.call(pattern[j], ref1) < 0) {
|
|
|
|
|
return -1;
|
|
|
|
|
}
|
Fix #3597: Allow interpolations in object keys
The following is now allowed:
o =
a: 1
b: 2
"#{'c'}": 3
"#{'d'}": 4
e: 5
"#{'f'}": 6
g: 7
It compiles to:
o = (
obj = {
a: 1,
b: 2
},
obj["" + 'c'] = 3,
obj["" + 'd'] = 4,
obj.e = 5,
obj["" + 'f'] = 6,
obj.g = 7,
obj
);
- Closes #3039. Empty interpolations in object keys are now _supposed_ to be
allowed.
- Closes #1131. No need to improve error messages for attempted key
interpolation anymore.
- Implementing this required fixing the following bug: `("" + a): 1` used to
error out on the colon, saying "unexpected colon". But really, it is the
attempted object key that is unexpected. Now the error is on the opening
parenthesis instead.
- However, the above fix broke some error message tests for regexes. The easiest
way to fix this was to make a seemingly unrelated change: The error messages
for unexpected identifiers, numbers, strings and regexes now say for example
'unexpected string' instead of 'unexpected """some #{really long} string"""'.
In other words, the tag _name_ is used instead of the tag _value_.
This was way easier to implement, and is more helpful to the user. Using the
tag value is good for operators, reserved words and the like, but not for
tokens which can contain any text. For example, 'unexpected identifier' is
better than 'unexpected expected' (if a variable called 'expected' was used
erraneously).
- While writing tests for the above point I found a few minor bugs with string
locations which have been fixed.
2015-02-07 20:16:59 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return i + j + fuzz - 1;
|
Fix #3597: Allow interpolations in object keys
The following is now allowed:
o =
a: 1
b: 2
"#{'c'}": 3
"#{'d'}": 4
e: 5
"#{'f'}": 6
g: 7
It compiles to:
o = (
obj = {
a: 1,
b: 2
},
obj["" + 'c'] = 3,
obj["" + 'd'] = 4,
obj.e = 5,
obj["" + 'f'] = 6,
obj.g = 7,
obj
);
- Closes #3039. Empty interpolations in object keys are now _supposed_ to be
allowed.
- Closes #1131. No need to improve error messages for attempted key
interpolation anymore.
- Implementing this required fixing the following bug: `("" + a): 1` used to
error out on the colon, saying "unexpected colon". But really, it is the
attempted object key that is unexpected. Now the error is on the opening
parenthesis instead.
- However, the above fix broke some error message tests for regexes. The easiest
way to fix this was to make a seemingly unrelated change: The error messages
for unexpected identifiers, numbers, strings and regexes now say for example
'unexpected string' instead of 'unexpected """some #{really long} string"""'.
In other words, the tag _name_ is used instead of the tag _value_.
This was way easier to implement, and is more helpful to the user. Using the
tag value is good for operators, reserved words and the like, but not for
tokens which can contain any text. For example, 'unexpected identifier' is
better than 'unexpected expected' (if a variable called 'expected' was used
erraneously).
- While writing tests for the above point I found a few minor bugs with string
locations which have been fixed.
2015-02-07 20:16:59 +01:00
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Returns `yes` if standing in front of something looking like
|
|
|
|
|
// `@<x>:`, `<x>:` or `<EXPRESSION_START><x>...<EXPRESSION_END>:`.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
looksObjectish(j) {
|
|
|
|
|
var end, index;
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
if (this.indexOfTag(j, '@', null, ':') !== -1 || this.indexOfTag(j, null, ':') !== -1) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return true;
|
2011-08-08 12:55:22 -04:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
index = this.indexOfTag(j, EXPRESSION_START);
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
if (index !== -1) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
end = null;
|
|
|
|
|
this.detectEnd(index + 1, (function(token) {
|
|
|
|
|
var ref;
|
|
|
|
|
return ref = token[0], indexOf.call(EXPRESSION_END, ref) >= 0;
|
|
|
|
|
}), (function(token, i) {
|
|
|
|
|
return end = i;
|
|
|
|
|
}));
|
|
|
|
|
if (this.tag(end + 1) === ':') {
|
|
|
|
|
return true;
|
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return false;
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Returns `yes` if current line of tokens contain an element of tags on same
|
|
|
|
|
// expression level. Stop searching at `LINEBREAKS` or explicit start of
|
|
|
|
|
// containing balanced expression.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
findTagsBackwards(i, tags) {
|
|
|
|
|
var backStack, ref, ref1, ref2, ref3, ref4, ref5;
|
|
|
|
|
backStack = [];
|
|
|
|
|
while (i >= 0 && (backStack.length || (ref2 = this.tag(i), indexOf.call(tags, ref2) < 0) && ((ref3 = this.tag(i), indexOf.call(EXPRESSION_START, ref3) < 0) || this.tokens[i].generated) && (ref4 = this.tag(i), indexOf.call(LINEBREAKS, ref4) < 0))) {
|
|
|
|
|
if (ref = this.tag(i), indexOf.call(EXPRESSION_END, ref) >= 0) {
|
|
|
|
|
backStack.push(this.tag(i));
|
|
|
|
|
}
|
|
|
|
|
if ((ref1 = this.tag(i), indexOf.call(EXPRESSION_START, ref1) >= 0) && backStack.length) {
|
|
|
|
|
backStack.pop();
|
|
|
|
|
}
|
|
|
|
|
i -= 1;
|
|
|
|
|
}
|
|
|
|
|
return ref5 = this.tag(i), indexOf.call(tags, ref5) >= 0;
|
|
|
|
|
}
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Look for signs of implicit calls and objects in the token stream and
|
|
|
|
|
// add them.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
addImplicitBracesAndParens() {
|
|
|
|
|
var stack, start;
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Track current balancing depth (both implicit and explicit) on stack.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
stack = [];
|
|
|
|
|
start = null;
|
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
2018-01-31 00:04:29 +01:00
|
|
|
|
var endImplicitCall, endImplicitObject, forward, implicitObjectContinues, inControlFlow, inImplicit, inImplicitCall, inImplicitControl, inImplicitObject, isImplicit, isImplicitCall, isImplicitObject, k, newLine, nextTag, nextToken, offset, prevTag, prevToken, ref, ref1, ref2, s, sameLine, stackIdx, stackItem, stackTag, stackTop, startIdx, startImplicitCall, startImplicitObject, startsLine, tag;
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[tag] = token;
|
|
|
|
|
[prevTag] = prevToken = i > 0 ? tokens[i - 1] : [];
|
2017-05-29 18:29:45 -06:00
|
|
|
|
[nextTag] = nextToken = i < tokens.length - 1 ? tokens[i + 1] : [];
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
stackTop = function() {
|
|
|
|
|
return stack[stack.length - 1];
|
|
|
|
|
};
|
|
|
|
|
startIdx = i;
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Helper function, used for keeping track of the number of tokens consumed
|
|
|
|
|
// and spliced, when returning for getting a new token.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
forward = function(n) {
|
|
|
|
|
return i - startIdx + n;
|
|
|
|
|
};
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Helper functions
|
2017-05-13 21:23:05 -07:00
|
|
|
|
isImplicit = function(stackItem) {
|
|
|
|
|
var ref;
|
|
|
|
|
return stackItem != null ? (ref = stackItem[2]) != null ? ref.ours : void 0 : void 0;
|
|
|
|
|
};
|
|
|
|
|
isImplicitObject = function(stackItem) {
|
|
|
|
|
return isImplicit(stackItem) && (stackItem != null ? stackItem[0] : void 0) === '{';
|
|
|
|
|
};
|
|
|
|
|
isImplicitCall = function(stackItem) {
|
|
|
|
|
return isImplicit(stackItem) && (stackItem != null ? stackItem[0] : void 0) === '(';
|
|
|
|
|
};
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
inImplicit = function() {
|
2017-05-13 21:23:05 -07:00
|
|
|
|
return isImplicit(stackTop());
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
inImplicitCall = function() {
|
2017-05-13 21:23:05 -07:00
|
|
|
|
return isImplicitCall(stackTop());
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
inImplicitObject = function() {
|
2017-05-13 21:23:05 -07:00
|
|
|
|
return isImplicitObject(stackTop());
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Unclosed control statement inside implicit parens (like
|
|
|
|
|
// class declaration or if-conditionals).
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
inImplicitControl = function() {
|
|
|
|
|
var ref;
|
2017-05-06 06:11:21 -07:00
|
|
|
|
return inImplicit() && ((ref = stackTop()) != null ? ref[0] : void 0) === 'CONTROL';
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
2017-05-29 18:29:45 -06:00
|
|
|
|
startImplicitCall = function(idx) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
stack.push([
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
'(',
|
|
|
|
|
idx,
|
|
|
|
|
{
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
ours: true
|
|
|
|
|
}
|
|
|
|
|
]);
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
return tokens.splice(idx, 0, generate('CALL_START', '(', ['', 'implicit function call', token[2]], prevToken));
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
endImplicitCall = function() {
|
|
|
|
|
stack.pop();
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
tokens.splice(i, 0, generate('CALL_END', ')', ['', 'end of input', token[2]], prevToken));
|
2013-02-24 19:09:01 +01:00
|
|
|
|
return i += 1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
2017-05-29 18:29:45 -06:00
|
|
|
|
startImplicitObject = function(idx, startsLine = true) {
|
|
|
|
|
var val;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
stack.push([
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
'{',
|
|
|
|
|
idx,
|
|
|
|
|
{
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
sameLine: true,
|
|
|
|
|
startsLine: startsLine,
|
|
|
|
|
ours: true
|
|
|
|
|
}
|
|
|
|
|
]);
|
|
|
|
|
val = new String('{');
|
|
|
|
|
val.generated = true;
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
return tokens.splice(idx, 0, generate('{', val, token, prevToken));
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
endImplicitObject = function(j) {
|
|
|
|
|
j = j != null ? j : i;
|
|
|
|
|
stack.pop();
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
tokens.splice(j, 0, generate('}', '}', token, prevToken));
|
2013-02-24 19:09:01 +01:00
|
|
|
|
return i += 1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
2017-05-29 18:29:45 -06:00
|
|
|
|
implicitObjectContinues = (j) => {
|
|
|
|
|
var nextTerminatorIdx;
|
|
|
|
|
nextTerminatorIdx = null;
|
|
|
|
|
this.detectEnd(j, function(token) {
|
|
|
|
|
return token[0] === 'TERMINATOR';
|
|
|
|
|
}, function(token, i) {
|
|
|
|
|
return nextTerminatorIdx = i;
|
|
|
|
|
}, {
|
|
|
|
|
returnOnNegativeLevel: true
|
|
|
|
|
});
|
|
|
|
|
if (nextTerminatorIdx == null) {
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
return this.looksObjectish(nextTerminatorIdx + 1);
|
|
|
|
|
};
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Don’t end an implicit call/object on next indent if any of these are in an argument/value.
|
2017-06-15 11:07:36 -05:00
|
|
|
|
if ((inImplicitCall() || inImplicitObject()) && indexOf.call(CONTROL_IN_IMPLICIT, tag) >= 0 || inImplicitObject() && prevTag === ':' && tag === 'FOR') {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
stack.push([
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
'CONTROL',
|
|
|
|
|
i,
|
|
|
|
|
{
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
ours: true
|
|
|
|
|
}
|
|
|
|
|
]);
|
|
|
|
|
return forward(1);
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (tag === 'INDENT' && inImplicit()) {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// An `INDENT` closes an implicit call unless
|
|
|
|
|
|
|
|
|
|
// 1. We have seen a `CONTROL` argument on the line.
|
|
|
|
|
// 2. The last token before the indent is part of the list below.
|
2017-05-29 18:29:45 -06:00
|
|
|
|
if (prevTag !== '=>' && prevTag !== '->' && prevTag !== '[' && prevTag !== '(' && prevTag !== ',' && prevTag !== '{' && prevTag !== 'ELSE' && prevTag !== '=') {
|
2017-06-15 11:07:36 -05:00
|
|
|
|
while (inImplicitCall() || inImplicitObject() && prevTag !== ':') {
|
|
|
|
|
if (inImplicitCall()) {
|
|
|
|
|
endImplicitCall();
|
|
|
|
|
} else {
|
|
|
|
|
endImplicitObject();
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (inImplicitControl()) {
|
|
|
|
|
stack.pop();
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
stack.push([tag, i]);
|
|
|
|
|
return forward(1);
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Straightforward start of explicit expression.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (indexOf.call(EXPRESSION_START, tag) >= 0) {
|
|
|
|
|
stack.push([tag, i]);
|
|
|
|
|
return forward(1);
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Close all implicit expressions inside of explicitly closed expressions.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (indexOf.call(EXPRESSION_END, tag) >= 0) {
|
|
|
|
|
while (inImplicit()) {
|
|
|
|
|
if (inImplicitCall()) {
|
|
|
|
|
endImplicitCall();
|
2018-01-31 06:33:17 -08:00
|
|
|
|
} else if (inImplicitObject()) {
|
|
|
|
|
endImplicitObject();
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
} else {
|
2018-01-31 06:33:17 -08:00
|
|
|
|
stack.pop();
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
start = stack.pop();
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
2018-01-31 00:04:29 +01:00
|
|
|
|
inControlFlow = () => {
|
|
|
|
|
var controlFlow, isFunc, seenFor, tagCurrentLine;
|
|
|
|
|
seenFor = this.findTagsBackwards(i, ['FOR']) && this.findTagsBackwards(i, ['FORIN', 'FOROF', 'FORFROM']);
|
|
|
|
|
controlFlow = seenFor || this.findTagsBackwards(i, ['WHILE', 'UNTIL', 'LOOP', 'LEADING_WHEN']);
|
|
|
|
|
if (!controlFlow) {
|
|
|
|
|
return false;
|
|
|
|
|
}
|
|
|
|
|
isFunc = false;
|
|
|
|
|
tagCurrentLine = token[2].first_line;
|
|
|
|
|
this.detectEnd(i, function(token, i) {
|
|
|
|
|
var ref;
|
|
|
|
|
return ref = token[0], indexOf.call(LINEBREAKS, ref) >= 0;
|
|
|
|
|
}, function(token, i) {
|
|
|
|
|
var first_line;
|
|
|
|
|
[prevTag, , {first_line}] = tokens[i - 1] || [];
|
|
|
|
|
return isFunc = tagCurrentLine === first_line && (prevTag === '->' || prevTag === '=>');
|
|
|
|
|
}, {
|
|
|
|
|
returnOnNegativeLevel: true
|
|
|
|
|
});
|
|
|
|
|
return isFunc;
|
|
|
|
|
};
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Recognize standard implicit calls like
|
|
|
|
|
// f a, f() b, f? c, h[0] d etc.
|
|
|
|
|
// Added support for spread dots on the left side: f ...a
|
2018-01-31 00:04:29 +01:00
|
|
|
|
if ((indexOf.call(IMPLICIT_FUNC, tag) >= 0 && token.spaced || tag === '?' && i > 0 && !tokens[i - 1].spaced) && (indexOf.call(IMPLICIT_CALL, nextTag) >= 0 || (nextTag === '...' && (ref = this.tag(i + 2), indexOf.call(IMPLICIT_CALL, ref) >= 0) && !this.findTagsBackwards(i, ['INDEX_START', '['])) || indexOf.call(IMPLICIT_UNSPACED_CALL, nextTag) >= 0 && !nextToken.spaced && !nextToken.newLine) && !inControlFlow()) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (tag === '?') {
|
|
|
|
|
tag = token[0] = 'FUNC_EXIST';
|
|
|
|
|
}
|
|
|
|
|
startImplicitCall(i + 1);
|
|
|
|
|
return forward(2);
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Implicit call taking an implicit indented object as first argument.
|
|
|
|
|
|
|
|
|
|
// f
|
|
|
|
|
// a: b
|
|
|
|
|
// c: d
|
|
|
|
|
|
|
|
|
|
// Don’t accept implicit calls of this type, when on the same line
|
|
|
|
|
// as the control structures below as that may misinterpret constructs like:
|
|
|
|
|
|
|
|
|
|
// if f
|
|
|
|
|
// a: 1
|
|
|
|
|
// as
|
|
|
|
|
|
|
|
|
|
// if f(a: 1)
|
|
|
|
|
|
|
|
|
|
// which is probably always unintended.
|
|
|
|
|
// Furthermore don’t allow this in literal arrays, as
|
|
|
|
|
// that creates grammatical ambiguities.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (indexOf.call(IMPLICIT_FUNC, tag) >= 0 && this.indexOfTag(i + 1, 'INDENT') > -1 && this.looksObjectish(i + 2) && !this.findTagsBackwards(i, ['CLASS', 'EXTENDS', 'IF', 'CATCH', 'SWITCH', 'LEADING_WHEN', 'FOR', 'WHILE', 'UNTIL'])) {
|
|
|
|
|
startImplicitCall(i + 1);
|
|
|
|
|
stack.push(['INDENT', i + 2]);
|
|
|
|
|
return forward(3);
|
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Implicit objects start here.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (tag === ':') {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Go back to the (implicit) start of the object.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
s = (function() {
|
2017-08-17 22:13:52 +02:00
|
|
|
|
var ref1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
switch (false) {
|
2017-08-17 22:13:52 +02:00
|
|
|
|
case ref1 = this.tag(i - 1), indexOf.call(EXPRESSION_END, ref1) < 0:
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return start[1];
|
|
|
|
|
case this.tag(i - 2) !== '@':
|
|
|
|
|
return i - 2;
|
|
|
|
|
default:
|
|
|
|
|
return i - 1;
|
|
|
|
|
}
|
|
|
|
|
}).call(this);
|
2017-08-24 00:02:57 -07:00
|
|
|
|
startsLine = s <= 0 || (ref1 = this.tag(s - 1), indexOf.call(LINEBREAKS, ref1) >= 0) || tokens[s - 1].newLine;
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Are we just continuing an already declared object?
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (stackTop()) {
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[stackTag, stackIdx] = stackTop();
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if ((stackTag === '{' || stackTag === 'INDENT' && this.tag(stackIdx - 1) === '{') && (startsLine || this.tag(s - 1) === ',' || this.tag(s - 1) === '{')) {
|
|
|
|
|
return forward(1);
|
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
startImplicitObject(s, !!startsLine);
|
|
|
|
|
return forward(2);
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// End implicit calls when chaining method calls
|
|
|
|
|
// like e.g.:
|
|
|
|
|
|
|
|
|
|
// f ->
|
|
|
|
|
// a
|
|
|
|
|
// .g b, ->
|
|
|
|
|
// c
|
|
|
|
|
// .h a
|
|
|
|
|
|
|
|
|
|
// and also
|
|
|
|
|
|
|
|
|
|
// f a
|
|
|
|
|
// .g b
|
|
|
|
|
// .h a
|
|
|
|
|
|
|
|
|
|
// Mark all enclosing objects as not sameLine
|
2017-05-13 21:23:05 -07:00
|
|
|
|
if (indexOf.call(LINEBREAKS, tag) >= 0) {
|
|
|
|
|
for (k = stack.length - 1; k >= 0; k += -1) {
|
|
|
|
|
stackItem = stack[k];
|
2017-07-04 13:31:51 -05:00
|
|
|
|
if (!isImplicit(stackItem)) {
|
|
|
|
|
break;
|
|
|
|
|
}
|
2017-05-13 21:23:05 -07:00
|
|
|
|
if (isImplicitObject(stackItem)) {
|
|
|
|
|
stackItem[2].sameLine = false;
|
|
|
|
|
}
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
}
|
|
|
|
|
newLine = prevTag === 'OUTDENT' || prevToken.newLine;
|
2018-01-29 02:51:57 +01:00
|
|
|
|
if (indexOf.call(IMPLICIT_END, tag) >= 0 || (indexOf.call(CALL_CLOSERS, tag) >= 0 && newLine) || ((tag === '..' || tag === '...') && this.findTagsBackwards(i, ["INDEX_START"]))) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
while (inImplicit()) {
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[stackTag, stackIdx, {sameLine, startsLine}] = stackTop();
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Close implicit calls when reached end of argument list
|
2018-01-31 06:09:50 +01:00
|
|
|
|
if (inImplicitCall() && prevTag !== ',' || (prevTag === ',' && tag === 'TERMINATOR' && (nextTag == null))) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
endImplicitCall();
|
2018-01-31 06:33:17 -08:00
|
|
|
|
// Close implicit objects such as:
|
|
|
|
|
// return a: 1, b: 2 unless true
|
|
|
|
|
} else if (inImplicitObject() && sameLine && tag !== 'TERMINATOR' && prevTag !== ':' && !((tag === 'POST_IF' || tag === 'FOR' || tag === 'WHILE' || tag === 'UNTIL') && startsLine && implicitObjectContinues(i + 1))) {
|
|
|
|
|
endImplicitObject();
|
|
|
|
|
// Close implicit objects when at end of line, line didn't end with a comma
|
|
|
|
|
// and the implicit object didn't start the line or the next line doesn’t look like
|
|
|
|
|
// the continuation of an object.
|
|
|
|
|
} else if (inImplicitObject() && tag === 'TERMINATOR' && prevTag !== ',' && !(startsLine && this.looksObjectish(i + 1))) {
|
|
|
|
|
endImplicitObject();
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
} else {
|
2018-01-31 06:33:17 -08:00
|
|
|
|
break;
|
2015-01-23 19:39:26 -05:00
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
|
|
|
|
}
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Close implicit object if comma is the last character
|
|
|
|
|
// and what comes after doesn’t look like it belongs.
|
|
|
|
|
// This is used for trailing commas and calls, like:
|
|
|
|
|
|
|
|
|
|
// x =
|
|
|
|
|
// a: b,
|
|
|
|
|
// c: d,
|
|
|
|
|
// e = 2
|
|
|
|
|
|
|
|
|
|
// and
|
|
|
|
|
|
|
|
|
|
// f a, b: c, d: e, f, g: h: i, j
|
|
|
|
|
|
2017-11-20 18:15:19 +01:00
|
|
|
|
if (tag === ',' && !this.looksObjectish(i + 1) && inImplicitObject() && !((ref2 = this.tag(i + 2)) === 'FOROF' || ref2 === 'FORIN') && (nextTag !== 'TERMINATOR' || !this.looksObjectish(i + 2))) {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// When nextTag is OUTDENT the comma is insignificant and
|
|
|
|
|
// should just be ignored so embed it in the implicit object.
|
|
|
|
|
|
|
|
|
|
// When it isn’t the comma go on to play a role in a call or
|
|
|
|
|
// array further up the stack, so give it a chance.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
offset = nextTag === 'OUTDENT' ? 1 : 0;
|
|
|
|
|
while (inImplicitObject()) {
|
|
|
|
|
endImplicitObject(i + offset);
|
|
|
|
|
}
|
2013-02-24 19:09:01 +01:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return forward(1);
|
|
|
|
|
});
|
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Make sure only strings and wrapped expressions are used in CSX attributes.
|
2017-06-07 07:33:46 +01:00
|
|
|
|
enforceValidCSXAttributes() {
|
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
|
|
|
|
var next, ref;
|
|
|
|
|
if (token.csxColon) {
|
|
|
|
|
next = tokens[i + 1];
|
|
|
|
|
if ((ref = next[0]) !== 'STRING_START' && ref !== 'STRING' && ref !== '(') {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
throwSyntaxError('expected wrapped or quoted JSX attribute', next[2]);
|
2017-06-07 07:33:46 +01:00
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return 1;
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Not all tokens survive processing by the parser. To avoid comments getting
|
|
|
|
|
// lost into the ether, find comments attached to doomed tokens and move them
|
|
|
|
|
// to a token that will make it to the other side.
|
|
|
|
|
rescueStowawayComments() {
|
|
|
|
|
var insertPlaceholder, shiftCommentsBackward, shiftCommentsForward;
|
|
|
|
|
insertPlaceholder = function(token, j, tokens, method) {
|
|
|
|
|
if (tokens[j][0] !== 'TERMINATOR') {
|
|
|
|
|
tokens[method](generate('TERMINATOR', '\n', tokens[j]));
|
|
|
|
|
}
|
|
|
|
|
return tokens[method](generate('JS', '', tokens[j], token));
|
|
|
|
|
};
|
|
|
|
|
shiftCommentsForward = function(token, i, tokens) {
|
|
|
|
|
var comment, j, k, len, ref, ref1, ref2;
|
|
|
|
|
// Find the next surviving token and attach this token’s comments to it,
|
|
|
|
|
// with a flag that we know to output such comments *before* that
|
|
|
|
|
// token’s own compilation. (Otherwise comments are output following
|
|
|
|
|
// the token they’re attached to.)
|
|
|
|
|
j = i;
|
|
|
|
|
while (j !== tokens.length && (ref = tokens[j][0], indexOf.call(DISCARDED, ref) >= 0)) {
|
|
|
|
|
j++;
|
|
|
|
|
}
|
|
|
|
|
if (!(j === tokens.length || (ref1 = tokens[j][0], indexOf.call(DISCARDED, ref1) >= 0))) {
|
|
|
|
|
ref2 = token.comments;
|
|
|
|
|
for (k = 0, len = ref2.length; k < len; k++) {
|
|
|
|
|
comment = ref2[k];
|
|
|
|
|
comment.unshift = true;
|
|
|
|
|
}
|
|
|
|
|
moveComments(token, tokens[j]);
|
|
|
|
|
return 1; // All following tokens are doomed!
|
|
|
|
|
} else {
|
|
|
|
|
j = tokens.length - 1;
|
|
|
|
|
insertPlaceholder(token, j, tokens, 'push');
|
|
|
|
|
// The generated tokens were added to the end, not inline, so we don’t skip.
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
};
|
|
|
|
|
shiftCommentsBackward = function(token, i, tokens) {
|
|
|
|
|
var j, ref, ref1;
|
|
|
|
|
// Find the last surviving token and attach this token’s comments to it.
|
|
|
|
|
j = i;
|
|
|
|
|
while (j !== -1 && (ref = tokens[j][0], indexOf.call(DISCARDED, ref) >= 0)) {
|
|
|
|
|
j--;
|
|
|
|
|
}
|
|
|
|
|
if (!(j === -1 || (ref1 = tokens[j][0], indexOf.call(DISCARDED, ref1) >= 0))) {
|
|
|
|
|
moveComments(token, tokens[j]);
|
|
|
|
|
return 1; // All previous tokens are doomed!
|
|
|
|
|
} else {
|
|
|
|
|
insertPlaceholder(token, 0, tokens, 'unshift');
|
|
|
|
|
// We added two tokens, so shift forward to account for the insertion.
|
|
|
|
|
return 3;
|
|
|
|
|
}
|
|
|
|
|
};
|
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
|
|
|
|
var dummyToken, j, ref, ref1, ret;
|
|
|
|
|
if (!token.comments) {
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
ret = 1;
|
|
|
|
|
if (ref = token[0], indexOf.call(DISCARDED, ref) >= 0) {
|
|
|
|
|
// This token won’t survive passage through the parser, so we need to
|
|
|
|
|
// rescue its attached tokens and redistribute them to nearby tokens.
|
|
|
|
|
// Comments that don’t start a new line can shift backwards to the last
|
|
|
|
|
// safe token, while other tokens should shift forward.
|
|
|
|
|
dummyToken = {
|
|
|
|
|
comments: []
|
|
|
|
|
};
|
|
|
|
|
j = token.comments.length - 1;
|
|
|
|
|
while (j !== -1) {
|
|
|
|
|
if (token.comments[j].newLine === false && token.comments[j].here === false) {
|
|
|
|
|
dummyToken.comments.unshift(token.comments[j]);
|
|
|
|
|
token.comments.splice(j, 1);
|
|
|
|
|
}
|
|
|
|
|
j--;
|
|
|
|
|
}
|
|
|
|
|
if (dummyToken.comments.length !== 0) {
|
|
|
|
|
ret = shiftCommentsBackward(dummyToken, i - 1, tokens);
|
|
|
|
|
}
|
|
|
|
|
if (token.comments.length !== 0) {
|
|
|
|
|
shiftCommentsForward(token, i, tokens);
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
// If any of this token’s comments start a line—there’s only
|
|
|
|
|
// whitespace between the preceding newline and the start of the
|
|
|
|
|
// comment—and this isn’t one of the special `JS` tokens, then
|
|
|
|
|
// shift this comment forward to precede the next valid token.
|
|
|
|
|
// `Block.compileComments` also has logic to make sure that
|
|
|
|
|
// “starting new line” comments follow or precede the nearest
|
|
|
|
|
// newline relative to the token that the comment is attached to,
|
|
|
|
|
// but that newline might be inside a `}` or `)` or other generated
|
|
|
|
|
// token that we really want this comment to output after. Therefore
|
|
|
|
|
// we need to shift the comments here, avoiding such generated and
|
|
|
|
|
// discarded tokens.
|
|
|
|
|
dummyToken = {
|
|
|
|
|
comments: []
|
|
|
|
|
};
|
|
|
|
|
j = token.comments.length - 1;
|
|
|
|
|
while (j !== -1) {
|
|
|
|
|
if (token.comments[j].newLine && !token.comments[j].unshift && !(token[0] === 'JS' && token.generated)) {
|
|
|
|
|
dummyToken.comments.unshift(token.comments[j]);
|
|
|
|
|
token.comments.splice(j, 1);
|
|
|
|
|
}
|
|
|
|
|
j--;
|
|
|
|
|
}
|
|
|
|
|
if (dummyToken.comments.length !== 0) {
|
|
|
|
|
ret = shiftCommentsForward(dummyToken, i + 1, tokens);
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (((ref1 = token.comments) != null ? ref1.length : void 0) === 0) {
|
|
|
|
|
delete token.comments;
|
|
|
|
|
}
|
|
|
|
|
return ret;
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
// Add location data to all tokens generated by the rewriter.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
addLocationDataToGeneratedTokens() {
|
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
|
|
|
|
var column, line, nextLocation, prevLocation, ref, ref1;
|
|
|
|
|
if (token[2]) {
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
if (!(token.generated || token.explicit)) {
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
if (token[0] === '{' && (nextLocation = (ref = tokens[i + 1]) != null ? ref[2] : void 0)) {
|
2017-04-06 10:06:45 -07:00
|
|
|
|
({
|
|
|
|
|
first_line: line,
|
|
|
|
|
first_column: column
|
|
|
|
|
} = nextLocation);
|
2018-01-31 06:33:17 -08:00
|
|
|
|
} else if (prevLocation = (ref1 = tokens[i - 1]) != null ? ref1[2] : void 0) {
|
|
|
|
|
({
|
|
|
|
|
last_line: line,
|
|
|
|
|
last_column: column
|
|
|
|
|
} = prevLocation);
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
} else {
|
2018-01-31 06:33:17 -08:00
|
|
|
|
line = column = 0;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
}
|
|
|
|
|
token[2] = {
|
|
|
|
|
first_line: line,
|
|
|
|
|
first_column: column,
|
|
|
|
|
last_line: line,
|
|
|
|
|
last_column: column
|
|
|
|
|
};
|
2013-02-24 19:09:01 +01:00
|
|
|
|
return 1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
});
|
|
|
|
|
}
|
2012-11-19 11:34:09 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// `OUTDENT` tokens should always be positioned at the last character of the
|
|
|
|
|
// previous token, so that AST nodes ending in an `OUTDENT` token end up with a
|
|
|
|
|
// location corresponding to the last “real” token under the node.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
fixOutdentLocationData() {
|
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
|
|
|
|
var prevLocationData;
|
|
|
|
|
if (!(token[0] === 'OUTDENT' || (token.generated && token[0] === 'CALL_END') || (token.generated && token[0] === '}'))) {
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
prevLocationData = tokens[i - 1][2];
|
|
|
|
|
token[2] = {
|
|
|
|
|
first_line: prevLocationData.last_line,
|
|
|
|
|
first_column: prevLocationData.last_column,
|
|
|
|
|
last_line: prevLocationData.last_line,
|
|
|
|
|
last_column: prevLocationData.last_column
|
|
|
|
|
};
|
Change OUTDENT tokens to be positioned at the end of the previous token
This commit adds another post-processing step after normal lexing that sets the
locationData on all OUTDENT tokens to be at the last character of the previous
token. This does feel like a little bit of a hack. Ideally the location data
would be set correctly in the first place and not in a post-processing step, but
I tried that and some temporary intermediate tokens were causing problems, so I
decided to set the location data once those intermediate tokens were removed.
Also, having this as a separate processing step makes it more robust and
isolated.
This fixes the problem in https://github.com/decaffeinate/decaffeinate/issues/371 .
In that issue, the CoffeeScript tokens had three OUTDENT tokens in a row, and
the last two overlapped with the `]`. Since at least one of those OUTDENT tokens
was considered part of the function body, the function expression had an ending
position just after the end of the `]`.
OUTDENT tokens are sort of a weird case in the lexer anyway, since they often
don't correspond to an actual location in the source code. It seems like the
code in `lexer.coffee` makes an attempt at finding a good place for them, but in
some cases, it has a bad result. This seems hard to avoid in the general case.
For example, in this code:
```coffee
[->
a]
```
There must be an OUTDENT between the `a` and the `]`, but CoffeeScript tokens
have an inclusive start and end, so they must always be at least one character
wide (I think). In this case, the lexer was choosing the `]` as the location,
and the parser ended up generating correct location data, I believe because
it ignores the outermost INDENT and OUTDENT tokens. However, with multiple
OUTDENT tokens in a row, the parser ends up producing location data that is
wrong.
It seems to me like there isn't a solid answer to "what location do OUTDENT
tokens have", since it hasn't mattered much, but for this commit, I'm defining
it: they always have the location of the last character of the previous token.
This should hopefully be fairly safe because tokens are still in the same order
relative to each other. Also, it's worth noting that this makes the start
location for OUTDENT tokens awkward. However, OUTDENT tokens are always used to
mark the end of something, so their `last_line` and `last_column` values are
always what matter when determining AST node bounds, so it is most important for
those to be correct.
2016-08-11 06:28:17 -07:00
|
|
|
|
return 1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
});
|
|
|
|
|
}
|
Change OUTDENT tokens to be positioned at the end of the previous token
This commit adds another post-processing step after normal lexing that sets the
locationData on all OUTDENT tokens to be at the last character of the previous
token. This does feel like a little bit of a hack. Ideally the location data
would be set correctly in the first place and not in a post-processing step, but
I tried that and some temporary intermediate tokens were causing problems, so I
decided to set the location data once those intermediate tokens were removed.
Also, having this as a separate processing step makes it more robust and
isolated.
This fixes the problem in https://github.com/decaffeinate/decaffeinate/issues/371 .
In that issue, the CoffeeScript tokens had three OUTDENT tokens in a row, and
the last two overlapped with the `]`. Since at least one of those OUTDENT tokens
was considered part of the function body, the function expression had an ending
position just after the end of the `]`.
OUTDENT tokens are sort of a weird case in the lexer anyway, since they often
don't correspond to an actual location in the source code. It seems like the
code in `lexer.coffee` makes an attempt at finding a good place for them, but in
some cases, it has a bad result. This seems hard to avoid in the general case.
For example, in this code:
```coffee
[->
a]
```
There must be an OUTDENT between the `a` and the `]`, but CoffeeScript tokens
have an inclusive start and end, so they must always be at least one character
wide (I think). In this case, the lexer was choosing the `]` as the location,
and the parser ended up generating correct location data, I believe because
it ignores the outermost INDENT and OUTDENT tokens. However, with multiple
OUTDENT tokens in a row, the parser ends up producing location data that is
wrong.
It seems to me like there isn't a solid answer to "what location do OUTDENT
tokens have", since it hasn't mattered much, but for this commit, I'm defining
it: they always have the location of the last character of the previous token.
This should hopefully be fairly safe because tokens are still in the same order
relative to each other. Also, it's worth noting that this makes the start
location for OUTDENT tokens awkward. However, OUTDENT tokens are always used to
mark the end of something, so their `last_line` and `last_column` values are
always what matter when determining AST node bounds, so it is most important for
those to be correct.
2016-08-11 06:28:17 -07:00
|
|
|
|
|
2017-08-29 17:17:56 -04:00
|
|
|
|
// Add parens around a `do` IIFE followed by a chained `.` so that the
|
|
|
|
|
// chaining applies to the executed function rather than the function
|
|
|
|
|
// object (see #3736)
|
|
|
|
|
addParensToChainedDoIife() {
|
|
|
|
|
var action, condition, doIndex;
|
|
|
|
|
condition = function(token, i) {
|
|
|
|
|
return this.tag(i - 1) === 'OUTDENT';
|
|
|
|
|
};
|
|
|
|
|
action = function(token, i) {
|
|
|
|
|
var ref;
|
|
|
|
|
if (ref = token[0], indexOf.call(CALL_CLOSERS, ref) < 0) {
|
|
|
|
|
return;
|
|
|
|
|
}
|
|
|
|
|
this.tokens.splice(doIndex, 0, generate('(', '(', this.tokens[doIndex]));
|
|
|
|
|
return this.tokens.splice(i + 1, 0, generate(')', ')', this.tokens[i]));
|
|
|
|
|
};
|
|
|
|
|
doIndex = null;
|
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
2017-08-31 01:42:50 -04:00
|
|
|
|
var glyphIndex, ref;
|
|
|
|
|
if (token[1] !== 'do') {
|
2017-08-29 17:17:56 -04:00
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
doIndex = i;
|
2017-08-31 01:42:50 -04:00
|
|
|
|
glyphIndex = i + 1;
|
|
|
|
|
if (this.tag(i + 1) === 'PARAM_START') {
|
|
|
|
|
glyphIndex = null;
|
|
|
|
|
this.detectEnd(i + 1, function(token, i) {
|
|
|
|
|
return this.tag(i - 1) === 'PARAM_END';
|
|
|
|
|
}, function(token, i) {
|
|
|
|
|
return glyphIndex = i;
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
if (!((glyphIndex != null) && ((ref = this.tag(glyphIndex)) === '->' || ref === '=>') && this.tag(glyphIndex + 1) === 'INDENT')) {
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
this.detectEnd(glyphIndex + 1, condition, action);
|
2017-08-29 17:17:56 -04:00
|
|
|
|
return 2;
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Because our grammar is LALR(1), it can’t handle some single-line
|
|
|
|
|
// expressions that lack ending delimiters. The **Rewriter** adds the implicit
|
|
|
|
|
// blocks, so it doesn’t need to. To keep the grammar clean and tidy, trailing
|
|
|
|
|
// newlines within expressions are removed and the indentation tokens of empty
|
|
|
|
|
// blocks are added.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
normalizeLines() {
|
2018-01-29 02:51:57 +01:00
|
|
|
|
var action, closeElseTag, condition, ifThens, indent, leading_if_then, leading_switch_when, outdent, starter;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
starter = indent = outdent = null;
|
2018-01-29 02:51:57 +01:00
|
|
|
|
leading_switch_when = null;
|
|
|
|
|
leading_if_then = null;
|
|
|
|
|
// Count `THEN` tags
|
|
|
|
|
ifThens = [];
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
condition = function(token, i) {
|
|
|
|
|
var ref, ref1, ref2, ref3;
|
2018-01-29 02:51:57 +01:00
|
|
|
|
return token[1] !== ';' && (ref = token[0], indexOf.call(SINGLE_CLOSERS, ref) >= 0) && !(token[0] === 'TERMINATOR' && (ref1 = this.tag(i + 1), indexOf.call(EXPRESSION_CLOSE, ref1) >= 0)) && !(token[0] === 'ELSE' && (starter !== 'THEN' || (leading_if_then || leading_switch_when))) && !(((ref2 = token[0]) === 'CATCH' || ref2 === 'FINALLY') && (starter === '->' || starter === '=>')) || (ref3 = token[0], indexOf.call(CALL_CLOSERS, ref3) >= 0) && (this.tokens[i - 1].newLine || this.tokens[i - 1][0] === 'OUTDENT');
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
action = function(token, i) {
|
2018-01-29 02:51:57 +01:00
|
|
|
|
if (token[0] === 'ELSE' && starter === 'THEN') {
|
|
|
|
|
ifThens.pop();
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return this.tokens.splice((this.tag(i - 1) === ',' ? i - 1 : i), 0, outdent);
|
|
|
|
|
};
|
2018-01-29 02:51:57 +01:00
|
|
|
|
closeElseTag = (tokens, i) => {
|
|
|
|
|
var lastThen, outdentElse, tlen;
|
|
|
|
|
tlen = ifThens.length;
|
|
|
|
|
if (!(tlen > 0)) {
|
|
|
|
|
return i;
|
|
|
|
|
}
|
|
|
|
|
lastThen = ifThens.pop();
|
|
|
|
|
[, outdentElse] = this.indentation(tokens[lastThen]);
|
|
|
|
|
// Insert `OUTDENT` to close inner `IF`.
|
|
|
|
|
outdentElse[1] = tlen * 2;
|
|
|
|
|
tokens.splice(i, 0, outdentElse);
|
|
|
|
|
// Insert `OUTDENT` to close outer `IF`.
|
|
|
|
|
outdentElse[1] = 2;
|
|
|
|
|
tokens.splice(i + 1, 0, outdentElse);
|
|
|
|
|
// Remove outdents from the end.
|
|
|
|
|
this.detectEnd(i + 2, function(token, i) {
|
|
|
|
|
var ref;
|
|
|
|
|
return (ref = token[0]) === 'OUTDENT' || ref === 'TERMINATOR';
|
|
|
|
|
}, function(token, i) {
|
|
|
|
|
if (this.tag(i) === 'OUTDENT' && this.tag(i + 1) === 'OUTDENT') {
|
|
|
|
|
return tokens.splice(i, 2);
|
|
|
|
|
}
|
|
|
|
|
});
|
|
|
|
|
return i + 2;
|
|
|
|
|
};
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return this.scanTokens(function(token, i, tokens) {
|
2018-01-29 02:51:57 +01:00
|
|
|
|
var conditionTag, j, k, ref, ref1, tag;
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[tag] = token;
|
2018-01-29 02:51:57 +01:00
|
|
|
|
conditionTag = (tag === '->' || tag === '=>') && this.findTagsBackwards(i, ['IF', 'WHILE', 'FOR', 'UNTIL', 'SWITCH', 'WHEN', 'LEADING_WHEN', '[', 'INDEX_START']) && !(this.findTagsBackwards(i, ['THEN', '..', '...']));
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (tag === 'TERMINATOR') {
|
|
|
|
|
if (this.tag(i + 1) === 'ELSE' && this.tag(i - 1) !== 'OUTDENT') {
|
|
|
|
|
tokens.splice(i, 1, ...this.indentation());
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
|
|
|
|
if (ref = this.tag(i + 1), indexOf.call(EXPRESSION_CLOSE, ref) >= 0) {
|
|
|
|
|
tokens.splice(i, 1);
|
|
|
|
|
return 0;
|
|
|
|
|
}
|
2013-07-30 17:24:46 +02:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (tag === 'CATCH') {
|
|
|
|
|
for (j = k = 1; k <= 2; j = ++k) {
|
|
|
|
|
if (!((ref1 = this.tag(i + j)) === 'OUTDENT' || ref1 === 'TERMINATOR' || ref1 === 'FINALLY')) {
|
|
|
|
|
continue;
|
|
|
|
|
}
|
|
|
|
|
tokens.splice(i + j, 0, ...this.indentation());
|
|
|
|
|
return 2 + j;
|
|
|
|
|
}
|
2013-07-30 17:24:46 +02:00
|
|
|
|
}
|
2017-06-29 21:00:01 -05:00
|
|
|
|
if ((tag === '->' || tag === '=>') && (this.tag(i + 1) === ',' || this.tag(i + 1) === '.' && token.newLine)) {
|
2017-06-26 01:46:54 -05:00
|
|
|
|
[indent, outdent] = this.indentation(tokens[i]);
|
|
|
|
|
tokens.splice(i + 1, 0, indent, outdent);
|
|
|
|
|
return 1;
|
|
|
|
|
}
|
2018-01-31 15:31:52 +01:00
|
|
|
|
if (indexOf.call(SINGLE_LINERS, tag) >= 0 && this.tag(i + 1) !== 'INDENT' && !(tag === 'ELSE' && this.tag(i + 1) === 'IF') && !conditionTag) {
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
starter = tag;
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[indent, outdent] = this.indentation(tokens[i]);
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
if (starter === 'THEN') {
|
|
|
|
|
indent.fromThen = true;
|
2013-04-11 20:42:38 +02:00
|
|
|
|
}
|
2018-01-29 02:51:57 +01:00
|
|
|
|
if (tag === 'THEN') {
|
|
|
|
|
leading_switch_when = this.findTagsBackwards(i, ['LEADING_WHEN']) && this.tag(i + 1) === 'IF';
|
|
|
|
|
leading_if_then = this.findTagsBackwards(i, ['IF']) && this.tag(i + 1) === 'IF';
|
|
|
|
|
}
|
|
|
|
|
if (tag === 'THEN' && this.findTagsBackwards(i, ['IF'])) {
|
|
|
|
|
ifThens.push(i);
|
|
|
|
|
}
|
|
|
|
|
// `ELSE` tag is not closed.
|
|
|
|
|
if (tag === 'ELSE' && this.tag(i - 1) !== 'OUTDENT') {
|
|
|
|
|
i = closeElseTag(tokens, i);
|
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
tokens.splice(i + 1, 0, indent);
|
|
|
|
|
this.detectEnd(i + 2, condition, action);
|
|
|
|
|
if (tag === 'THEN') {
|
|
|
|
|
tokens.splice(i, 1);
|
|
|
|
|
}
|
|
|
|
|
return 1;
|
2013-04-11 20:42:38 +02:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return 1;
|
|
|
|
|
});
|
|
|
|
|
}
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tag postfix conditionals as such, so that we can parse them with a
|
|
|
|
|
// different precedence.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
tagPostfixConditionals() {
|
|
|
|
|
var action, condition, original;
|
|
|
|
|
original = null;
|
|
|
|
|
condition = function(token, i) {
|
|
|
|
|
var prevTag, tag;
|
2017-04-06 10:06:45 -07:00
|
|
|
|
[tag] = token;
|
|
|
|
|
[prevTag] = this.tokens[i - 1];
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return tag === 'TERMINATOR' || (tag === 'INDENT' && indexOf.call(SINGLE_LINERS, prevTag) < 0);
|
|
|
|
|
};
|
|
|
|
|
action = function(token, i) {
|
|
|
|
|
if (token[0] !== 'INDENT' || (token.generated && !token.fromThen)) {
|
|
|
|
|
return original[0] = 'POST_' + original[0];
|
2012-04-10 14:57:45 -04:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
};
|
|
|
|
|
return this.scanTokens(function(token, i) {
|
|
|
|
|
if (token[0] !== 'IF') {
|
|
|
|
|
return 1;
|
2012-04-10 14:57:45 -04:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
original = token;
|
|
|
|
|
this.detectEnd(i + 1, condition, action);
|
2010-11-11 21:48:08 -05:00
|
|
|
|
return 1;
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
});
|
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Generate the indentation tokens, based on another token on the same line.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
indentation(origin) {
|
|
|
|
|
var indent, outdent;
|
|
|
|
|
indent = ['INDENT', 2];
|
|
|
|
|
outdent = ['OUTDENT', 2];
|
|
|
|
|
if (origin) {
|
|
|
|
|
indent.generated = outdent.generated = true;
|
|
|
|
|
indent.origin = outdent.origin = origin;
|
|
|
|
|
} else {
|
|
|
|
|
indent.explicit = outdent.explicit = true;
|
2012-04-10 14:57:45 -04:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
return [indent, outdent];
|
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Look up a tag by token index.
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
tag(i) {
|
|
|
|
|
var ref;
|
|
|
|
|
return (ref = this.tokens[i]) != null ? ref[0] : void 0;
|
2012-11-22 14:22:38 -05:00
|
|
|
|
}
|
[CS2] Compile class constructors to ES2015 classes (#4354)
* Compile classes to ES2015 classes
Rather than compiling classes to named functions with prototype and
class assignments, they are now compiled to ES2015 class declarations.
Backwards compatibility has been maintained by compiling ES2015-
incompatible properties as prototype or class assignments. `super`
continues to be compiled as before.
Where possible, classes will be compiled "bare", without an enclosing
IIFE. This is possible when the class contains only ES2015 compatible
expressions (methods and static methods), and has no parent (this last
constraint is a result of the legacy `super` compilation, and could be
removed once ES2015 `super` is being used). Classes are still assigned
to variables to maintain compatibility for assigned class expressions.
There are a few changes to existing functionality that could break
backwards compatibility:
- Derived constructors that deliberately don't call `super` are no
longer possible. ES2015 derived classes can't use `this` unless the
parent constructor has been called, so it's now called implicitly when
not present.
- As a consequence of the above, derived constructors with @ parameters
or bound methods and explicit `super` calls are not allowed. The
implicit `super` must be used in these cases.
* Add tests to verify class interoperability with ES
* Refactor class nodes to separate executable body logic
Logic has been redistributed amongst the class nodes so that:
- `Class` contains the logic necessary to compile an ES class
declaration.
- `ExecutableClassBody` contains the logic necessary to compile CS'
class extensions that require an executable class body.
`Class` still necessarily contains logic to determine whether an
expression is valid in an ES class initializer or not. If any invalid
expressions are found then `Class` will wrap itself in an
`ExecutableClassBody` when compiling.
* Rename `Code#static` to `Code#isStatic`
This naming is more consistent with other `Code` flags.
* Output anonymous classes when possible
Anonymous classes can be output when:
- The class has no parent. The current super compilation needs a class
variable to reference. This condition will go away when ES2015 super
is in use.
- The class contains no bound static methods. Bound static methods have
their context set to the class name.
* Throw errors at compile time for async or generator constructors
* Improve handling of anonymous classes
Anonymous classes are now always anonymous. If a name is required (e.g.
for bound static methods or derived classes) then the class is compiled
in an `ExecutableClassBody` which will give the anonymous class a stable
reference.
* Add a `replaceInContext` method to `Node`
`replaceInContext` will traverse children looking for a node for which
`match` returns true. Once found, the matching node will be replaced by
the result of calling `replacement`.
* Separate `this` assignments from function parameters
This change has been made to simplify two future changes:
1. Outputting `@`-param assignments after a `super` call.
In this case it is necessary that non-`@` parameters are available
before `super` is called, so destructuring has to happen before
`this` assignment.
2. Compiling destructured assignment to ES6
In this case also destructuring has to happen before `this`,
as destructuring can happen in the arguments list, but `this`
assignment can not.
A bonus side-effect is that default values for `@` params are now output
as ES6 default parameters, e.g.
(@a = 1) ->
becomes
function a (a = 1) {
this.a = a;
}
* Change `super` handling in class constructors
Inside an ES derived constructor (a constructor for a class that extends
another class), it is impossible to access `this` until `super` has been
called. This conflicts with CoffeeScript's `@`-param and bound method
features, which compile to `this` references at the top of a function
body. For example:
class B extends A
constructor: (@param) -> super
method: =>
This would compile to something like:
class B extends A {
constructor (param) {
this.param = param;
this.method = bind(this.method, this);
super(...arguments);
}
}
This would break in an ES-compliant runtime as there are `this`
references before the call to `super`. Before this commit we were
dealing with this by injecting an implicit `super` call into derived
constructors that do not already have an explicit `super` call.
Furthermore, we would disallow explicit `super` calls in derived
constructors that used bound methods or `@`-params, meaning the above
example would need to be rewritten as:
class B extends A
constructor: (@param) ->
method: =>
This would result in a call to `super(...arguments)` being generated as
the first expression in `B#constructor`.
Whilst this approach seems to work pretty well, and is arguably more
convenient than having to manually call `super` when you don't
particularly care about the arguments, it does introduce some 'magic'
and separation from ES, and would likely be a pain point in a project
that made use of significant constructor overriding.
This commit introduces a mechanism through which `super` in constructors
is 'expanded' to include any generated `this` assignments, whilst
retaining the same semantics of a super call. The first example above
now compiles to something like:
class B extends A {
constructor (param) {
var ref
ref = super(...arguments), this.param = param, this.method = bind(this.method, this), ref;
}
}
* Improve `super` handling in constructors
Rather than functions expanding their `super` calls, the `SuperCall`
node can now be given a list of `thisAssignments` to apply when it is
compiled.
This allows us to use the normal compiler machinery to determine whether
the `super` result needs to be cached, whether it appears inline or not,
etc.
* Fix anonymous classes at the top level
Anonymous classes in ES are only valid within expressions. If an
anonymous class is at the top level it will now be wrapped in
parenthses to force it into an expression.
* Re-add Parens wrapper around executable class bodies
This was lost in the refactoring, but it necessary to ensure
`new class ...` works as expected when there's an executable body.
* Throw compiler errors for badly configured derived constructors
Rather than letting them become runtime errors, the following checks are
now performed when compiling a derived constructor:
- The constructor **must** include a call to `super`.
- The constructor **must not** reference `this` in the function body
before `super` has been called.
* Add some tests exercising new class behaviour
- async methods in classes
- `this` access after `super` in extended classes
- constructor super in arrow functions
- constructor functions can't be async
- constructor functions can't be generators
- derived constructors must call super
- derived constructors can't reference `this` before calling super
- generator methods in classes
- 'new' target
* Improve constructor `super` errors
Add a check for `super` in non-extended class constructors, and
explicitly mention derived constructors in the "can't reference this
before super" error.
* Fix compilation of multiple `super` paths in derived constructors
`super` can only be called once, but it can be called conditionally from
multiple locations. The chosen fix is to add the `this` assignments to
every super call.
* Additional class tests, added as a separate file to simplify testing and merging.
Some methods are commented out because they currently throw and I'm not sure how
to test for compilation errors like those.
There is also one test which I deliberately left without passing, `super` in an external prototype override.
This test should 'pass' but is really a variation on the failing `super only allowed in an instance method`
tests above it.
* Changes to the tests. Found bug in super in prototype method. fixed.
* Added failing test back in, dealing with bound functions in external prototype overrides.
* Located a bug in the compiler relating to assertions and escaped ES6 classes.
* Move tests from classes-additional.coffee into classes.coffee; comment out console.log
* Cleaned up tests and made changes based on feedback. Test at the end still has issues, but it's commented out for now.
* Make HoistTarget.expand recursive
It's possible that a hoisted node may itself contain hoisted nodes (e.g.
a class method inside a class method). For this to work the hoisted
fragments need to be expanded recursively.
* Uncomment final test in classes.coffee
The test case now compiles, however another issue is affecting the test
due to the error for `this` before `super` triggering based on source
order rather than execution order. These have been commented out for
now.
* Fixed last test TODOs in test/classes.coffee
Turns out an expression like `this.foo = super()` won't run in JS as it
attempts to lookup `this` before evaluating `super` (i.e. throws "this
is not defined").
* Added more tests for compatability checks, statics, prototypes and ES6 expectations. Cleaned test "nested classes with super".
* Changes to reflect feedback and to comment out issues that will be addressed seperately.
* Clean up test/classes.coffee
- Trim trailing whitespace.
- Rephrase a condition to be more idiomatic.
* Remove check for `super` in derived constructors
In order to be usable at runtime, an extended ES class must call `super`
OR return an alternative object. This check prevented the latter case,
and checking for an alternative return can't be completed statically
without control flow analysis.
* Disallow 'super' in constructor parameter defaults
There are many edge cases when combining 'super' in parameter defaults
with @-parameters and bound functions (and potentially property
initializers in the future).
Rather than attempting to resolve these edge cases, 'super' is now
explicitly disallowed in constructor parameter defaults.
* Disallow @-params in derived constructors without 'super'
@-parameters can't be assigned unless 'super' is called.
2017-01-13 05:55:30 +00:00
|
|
|
|
|
2011-12-18 18:53:34 -05:00
|
|
|
|
};
|
|
|
|
|
|
2013-02-24 19:09:01 +01:00
|
|
|
|
Rewriter.prototype.generate = generate;
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2010-11-11 21:48:08 -05:00
|
|
|
|
return Rewriter;
|
2011-12-14 10:39:20 -05:00
|
|
|
|
|
2017-12-18 18:51:08 +00:00
|
|
|
|
}).call(this);
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Constants
|
|
|
|
|
// ---------
|
|
|
|
|
|
|
|
|
|
// List of the token pairs that must be balanced.
|
Fix #3597: Allow interpolations in object keys
The following is now allowed:
o =
a: 1
b: 2
"#{'c'}": 3
"#{'d'}": 4
e: 5
"#{'f'}": 6
g: 7
It compiles to:
o = (
obj = {
a: 1,
b: 2
},
obj["" + 'c'] = 3,
obj["" + 'd'] = 4,
obj.e = 5,
obj["" + 'f'] = 6,
obj.g = 7,
obj
);
- Closes #3039. Empty interpolations in object keys are now _supposed_ to be
allowed.
- Closes #1131. No need to improve error messages for attempted key
interpolation anymore.
- Implementing this required fixing the following bug: `("" + a): 1` used to
error out on the colon, saying "unexpected colon". But really, it is the
attempted object key that is unexpected. Now the error is on the opening
parenthesis instead.
- However, the above fix broke some error message tests for regexes. The easiest
way to fix this was to make a seemingly unrelated change: The error messages
for unexpected identifiers, numbers, strings and regexes now say for example
'unexpected string' instead of 'unexpected """some #{really long} string"""'.
In other words, the tag _name_ is used instead of the tag _value_.
This was way easier to implement, and is more helpful to the user. Using the
tag value is good for operators, reserved words and the like, but not for
tokens which can contain any text. For example, 'unexpected identifier' is
better than 'unexpected expected' (if a variable called 'expected' was used
erraneously).
- While writing tests for the above point I found a few minor bugs with string
locations which have been fixed.
2015-02-07 20:16:59 +01:00
|
|
|
|
BALANCED_PAIRS = [['(', ')'], ['[', ']'], ['{', '}'], ['INDENT', 'OUTDENT'], ['CALL_START', 'CALL_END'], ['PARAM_START', 'PARAM_END'], ['INDEX_START', 'INDEX_END'], ['STRING_START', 'STRING_END'], ['REGEX_START', 'REGEX_END']];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// The inverse mappings of `BALANCED_PAIRS` we’re trying to fix up, so we can
|
|
|
|
|
// look things up from either end.
|
2011-09-17 08:26:04 +09:00
|
|
|
|
exports.INVERSES = INVERSES = {};
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// The tokens that signal the start/end of a balanced pair.
|
2010-10-02 18:57:16 +09:00
|
|
|
|
EXPRESSION_START = [];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2010-10-02 18:57:16 +09:00
|
|
|
|
EXPRESSION_END = [];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2015-01-30 20:33:03 +01:00
|
|
|
|
for (k = 0, len = BALANCED_PAIRS.length; k < len; k++) {
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
[left, right] = BALANCED_PAIRS[k];
|
|
|
|
|
EXPRESSION_START.push(INVERSES[right] = left);
|
|
|
|
|
EXPRESSION_END.push(INVERSES[left] = right);
|
2010-03-07 14:41:52 -05:00
|
|
|
|
}
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that indicate the close of a clause of an expression.
|
2013-07-31 22:12:44 +02:00
|
|
|
|
EXPRESSION_CLOSE = ['CATCH', 'THEN', 'ELSE', 'FINALLY'].concat(EXPRESSION_END);
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that, if followed by an `IMPLICIT_CALL`, indicate a function invocation.
|
2016-03-05 17:41:15 +01:00
|
|
|
|
IMPLICIT_FUNC = ['IDENTIFIER', 'PROPERTY', 'SUPER', ')', 'CALL_END', ']', 'INDEX_END', '@', 'THIS'];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// If preceded by an `IMPLICIT_FUNC`, indicates a function invocation.
|
2017-06-07 07:33:46 +01:00
|
|
|
|
IMPLICIT_CALL = ['IDENTIFIER', 'CSX_TAG', 'PROPERTY', 'NUMBER', 'INFINITY', 'NAN', 'STRING', 'STRING_START', 'REGEX', 'REGEX_START', 'JS', 'NEW', 'PARAM_START', 'CLASS', 'IF', 'TRY', 'SWITCH', 'THIS', 'UNDEFINED', 'NULL', 'BOOL', 'UNARY', 'YIELD', 'AWAIT', 'UNARY_MATH', 'SUPER', 'THROW', '@', '->', '=>', '[', '(', '{', '--', '++'];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2010-10-20 04:52:07 +13:00
|
|
|
|
IMPLICIT_UNSPACED_CALL = ['+', '-'];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that always mark the end of an implicit call for single-liners.
|
2011-05-15 10:41:41 -04:00
|
|
|
|
IMPLICIT_END = ['POST_IF', 'FOR', 'WHILE', 'UNTIL', 'WHEN', 'BY', 'LOOP', 'TERMINATOR'];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Single-line flavors of block expressions that have unclosed endings.
|
|
|
|
|
// The grammar can’t disambiguate them, so we insert the implicit indentation.
|
2014-09-06 12:55:27 +02:00
|
|
|
|
SINGLE_LINERS = ['ELSE', '->', '=>', 'TRY', 'FINALLY', 'THEN'];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
2010-03-07 14:41:52 -05:00
|
|
|
|
SINGLE_CLOSERS = ['TERMINATOR', 'CATCH', 'FINALLY', 'ELSE', 'OUTDENT', 'LEADING_WHEN'];
|
2011-09-18 17:16:39 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that end a line.
|
2010-08-14 19:56:00 -04:00
|
|
|
|
LINEBREAKS = ['TERMINATOR', 'INDENT', 'OUTDENT'];
|
2011-12-14 10:39:20 -05:00
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that close open calls when they follow a newline.
|
2013-11-27 04:52:52 +00:00
|
|
|
|
CALL_CLOSERS = ['.', '?.', '::', '?::'];
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that prevent a subsequent indent from ending implicit calls/objects
|
2017-06-15 11:07:36 -05:00
|
|
|
|
CONTROL_IN_IMPLICIT = ['IF', 'TRY', 'FINALLY', 'CATCH', 'CLASS', 'SWITCH'];
|
|
|
|
|
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
// Tokens that are swallowed up by the parser, never leading to code generation.
|
|
|
|
|
// You can spot these in `grammar.coffee` because the `o` function second
|
|
|
|
|
// argument doesn’t contain a `new` call for these tokens.
|
|
|
|
|
// `STRING_START` isn’t on this list because its `locationData` matches that of
|
|
|
|
|
// the node that becomes `StringWithInterpolations`, and therefore
|
|
|
|
|
// `addDataToNode` attaches `STRING_START`’s tokens to that node.
|
2017-10-18 17:22:02 -07:00
|
|
|
|
DISCARDED = ['(', ')', '[', ']', '{', '}', '.', '..', '...', ',', '=', '++', '--', '?', 'AS', 'AWAIT', 'CALL_START', 'CALL_END', 'DEFAULT', 'ELSE', 'EXTENDS', 'EXPORT', 'FORIN', 'FOROF', 'FORFROM', 'IMPORT', 'INDENT', 'INDEX_SOAK', 'LEADING_WHEN', 'OUTDENT', 'PARAM_END', 'REGEX_START', 'REGEX_END', 'RETURN', 'STRING_END', 'THROW', 'UNARY', 'YIELD'].concat(IMPLICIT_UNSPACED_CALL.concat(IMPLICIT_END.concat(CALL_CLOSERS.concat(CONTROL_IN_IMPLICIT))));
|
[CS2] Comments (#4572)
* Make `addLocationDataFn` more DRY
* Style fixes
* Provide access to full parser inside our custom function running in parser.js; rename the function to lay the groundwork for adding data aside from location data
* Fix style.
* Fix style.
* Label test comments
* Update grammar to remove comment tokens; update DSL to call new helper function that preserves comments through parsing
* New implementation of compiling block comments: the lexer pulls them out of the token stream, attaching them as a property to a token; the rewriter moves the attachment around so it lives on a token that is destined to make it through to compilation (and in a good placement); and the nodes render the block comment. All tests but one pass (commented out).
* If a comment follows a class declaration, move the comment inside the class body
* Style
* Improve indentation of multiline comments
* Fix indentation for block comments, at least in the cases covered by the one failing test
* Don’t reverse the order of unshifted comments
* Simplify rewriter’s handling of comments, generalizing the special case
* Expand the list of tokens we need to avoid for passing comments through the parser; get some literal tokens to have nodes created for them so that the comments pass through
* Improve comments; fix multiline flag
* Prepare HereComments for processing line comments
* Line comments, first draft: the tests pass, but the line comments aren’t indented and sometimes trail previous lines when they shouldn’t; updated compiler output in following commit
* Updated compiler, now with line comments
* `process` doesn’t exist in the browser, so we should check for its existence first
* Update parser output
* Test that proves #4290 is fixed
* Indent line comments, first pass
* Compiled output with indented line comments
* Comments that start a new line shouldn’t trail; don’t skip comments attached to generated tokens; stop looking for indentation once we hit a newline
* Revised output
* Cleanup
* Split “multiline” line comment tokens, shifting them forward or back as appropriate
* Fix comments in module specifiers
* Abstract attaching comments to a node
* Line comments in interpolated strings
* Line comments can’t be multiline anymore
* Improve handling of blank lines and indentation of following comments that start a new line (i.e. don’t trail)
* Make comments compilation more object-oriented
* Remove lots of dead code that we don’t need anymore because a comment is never a node, only a fragment
* Improve eqJS helper
* Fix #4290 definitively, with improved output for arrays with interspersed block comments
* Add support for line comments output interspersed within arrays
* Fix mistake, don’t lose the variable we’re working on
* Remove redundant replacements
* Check for indentation only from the start of the string
* Indentations in generated JS are always multiples of two spaces (never tabs) so just look for 2+ spaces
* Update package versions; run Babel twice, once for each preset, temporarily until a Babili bug is fixed that prevents it from running with the env preset
* Don’t rely on `fragment.type`, which can break when the compiler is minified
* Updated generated docs and browser compiler
* Output block comments after function arguments
* Comments appear above scope `var` declarations; better tracking of generated `JS` tokens created only to shepherd comments through to the output
* Create new FuncGlyph node, to hold comments we want to output near the function parameters
* Block comments between `)` and `->`/`=>` get output between `)` and `{`.
* Fix indentation of comments that are the first line inside a bare mode block
* Updated output
* Full Flow example
* Updated browser compiler
* Abstract and organize comment fragment generation code; store more properties on the comment fragment objects; make `throw` behave like `return`
* Abstract token insertion code
* Add missing locationData to STRING_START token, giving it the locationData of the overall StringWithInterpolations token so that comments attached to STRING_START end up on the StringWithInterpolations node
* Allow `SUPER` tokens to carry comments
* Rescue comments from `Existence` nodes and `If` nodes’ conditions
* Rescue comments after `\` line continuation tokens
* Updated compiled output
* Updated browser compiler
* Output block comments in the same `compileFragments` method as line comments, except for inline block comments
* Comments before splice
* Updated browser compiler
* Track compiledComments as a property of Base, to ensure that it’s not a global variable
* Docs: split up the Usage section
* Docs for type annotations via Flow; updated docs output
* Update regular comments documentation
* Updated browser compiler
* Comments before soak
* Comments before static methods, and probably before `@variable =` (this) assignments generally
* Comments before ‘if exists?’, refactor comment before ‘if this.var’ to be more precise, improve helper methods
* Comments before a method that contains ‘super()’ should output above the method property, not above the ‘super.method()’ call
* Fix missing comments before `if not` (i.e. before a UNARY token)
* Fix comments before ‘for’; add test for comment before assignment if (fixed in earlier commit)
* Comments within heregexes
* Updated browser compiler
* Update description to reflect what’s now happening in compileCommentFragments
* Preserve blank lines between line comments; output “whitespace-only” line comments as blank lines, rather than `//` following by whitespace
* Better future-proof comments tests
* Comments before object destructuring; abstract method for setting comments aside before compilation
* Handle more cases of comments before or after `for` loop declaration lines
* Fix indentation of comments preceding `for` loops
* Fix comment before splat function parameter
* Catch another RegexWithInterpolations comment edge case
* Updated browser compiler
* Change heregex example to one that’s more readable; update output
* Remove a few last references to the defunct HERECOMMENT token
* Abstract location hash creation into a function
* Improved clarity per code review notes
* Updated browser compiler
2017-08-02 19:34:34 -07:00
|
|
|
|
|
2010-09-21 16:53:58 +09:00
|
|
|
|
}).call(this);
|