1
0
Fork 0
mirror of https://github.com/jashkenas/coffeescript.git synced 2022-11-09 12:23:24 -05:00
jashkenas--coffeescript/lib/grammar.js
gfxmonk cc3c314988 Cleaned up return logic
- ReturnNodes are explicitly added during compilation
 - ReturnNode is used instead of scattering "return" throughout
   code compilation snippets
 - nodes gain a make_return method in order to do the most useful
   thing when a return is requested
2010-03-21 22:21:55 +11:00

661 lines
27 KiB
JavaScript

(function(){
var Parser, _a, _b, _c, _d, _e, _f, _g, _h, alt, alternatives, grammar, name, o, operators, token, tokens, unwrap;
var __hasProp = Object.prototype.hasOwnProperty;
// The CoffeeScript parser is generated by [Jison](http://github.com/zaach/jison)
// from this grammar file. Jison is a bottom-up parser generator, similar in
// style to [Bison](http://www.gnu.org/software/bison), implemented in JavaScript.
// It can recognize [LALR(1), LR(0), SLR(1), and LR(1)](http://en.wikipedia.org/wiki/LR_grammar)
// type grammars. To create the Jison parser, we list the pattern to match
// on the left-hand side, and the action to take (usually the creation of syntax
// tree nodes) on the right. As the parser runs, it
// shifts tokens from our token stream, from left to right, and
// [attempts to match](http://en.wikipedia.org/wiki/Bottom-up_parsing)
// the token sequence against the rules below. When a match can be made, it
// reduces into the [nonterminal](http://en.wikipedia.org/wiki/Terminal_and_nonterminal_symbols)
// (the enclosing name at the top), and we proceed from there.
// If you run the `cake build:parser` command, Jison constructs a parse table
// from our rules and saves it into `lib/parser.js`.
// The only dependency is on the **Jison.Parser**.
Parser = require('jison').Parser;
// Jison DSL
// ---------
// Since we're going to be wrapped in a function by Jison in any case, if our
// action immediately returns a value, we can optimize by removing the function
// wrapper and just returning the value directly.
unwrap = /function\s*\(\)\s*\{\s*return\s*([\s\S]*);\s*\}/;
// Our handy DSL for Jison grammar generation, thanks to
// [Tim Caswell](http://github.com/creationix). For every rule in the grammar,
// we pass the pattern-defining string, the action to run, and extra options,
// optionally. If no action is specified, we simply pass the value of the
// previous nonterminal.
o = function o(pattern_string, action, options) {
var match;
if (!(action)) {
return [pattern_string, '$$ = $1;', options];
}
action = (match = (action + '').match(unwrap)) ? match[1] : "(" + action + "())";
return [pattern_string, "$$ = " + action + ";", options];
};
// Grammatical Rules
// -----------------
// In all of the rules that follow, you'll see the name of the nonterminal as
// the key to a list of alternative matches. With each match's action, the
// dollar-sign variables are provided by Jison as references to the value of
// their numeric position, so in this rule:
// "Expression UNLESS Expression"
// `$1` would be the value of the first `Expression`, `$2` would be the token
// for the `UNLESS` terminal, and `$3` would be the value of the second
// `Expression`.
grammar = {
// The **Root** is the top-level node in the syntax tree. Since we parse bottom-up,
// all parsing must end here.
Root: [o("", function() {
return new Expressions();
}), o("TERMINATOR", function() {
return new Expressions();
}), o("Expressions"), o("Block TERMINATOR")
],
// Any list of expressions or method body, seperated by line breaks or
// semicolons.
Expressions: [o("Expression", function() {
return Expressions.wrap([$1]);
}), o("Expressions TERMINATOR Expression", function() {
return $1.push($3);
}), o("Expressions TERMINATOR")
],
// All the different types of expressions in our language. The basic unit of
// CoffeeScript is the **Expression** -- you'll notice that there is no
// "statement" nonterminal. Expressions serve as the building blocks
// of many other rules, making them somewhat circular.
Expression: [o("Value"), o("Call"), o("Curry"), o("Code"), o("Operation"), o("Assign"), o("If"), o("Try"), o("Throw"), o("Return"), o("While"), o("For"), o("Switch"), o("Extends"), o("Class"), o("Splat"), o("Existence"), o("Comment"), o("Extension")],
// A an indented block of expressions. Note that the [Rewriter](rewriter.html)
// will convert some postfix forms into blocks for us, by adjusting the
// token stream.
Block: [o("INDENT Expressions OUTDENT", function() {
return $2;
}), o("INDENT OUTDENT", function() {
return new Expressions();
}), o("TERMINATOR Comment", function() {
return Expressions.wrap([$2]);
})
],
// A literal identifier, a variable name or property.
Identifier: [o("IDENTIFIER", function() {
return new LiteralNode(yytext);
})
],
// Alphanumerics are separated from the other **Literal** matchers because
// they can also serve as keys in object literals.
AlphaNumeric: [o("NUMBER", function() {
return new LiteralNode(yytext);
}), o("STRING", function() {
return new LiteralNode(yytext);
})
],
// All of our immediate values. These can (in general), be passed straight
// through and printed to JavaScript.
Literal: [o("AlphaNumeric"), o("JS", function() {
return new LiteralNode(yytext);
}), o("REGEX", function() {
return new LiteralNode(yytext);
}), o("BREAK", function() {
return new LiteralNode(yytext);
}), o("CONTINUE", function() {
return new LiteralNode(yytext);
}), o("TRUE", function() {
return new LiteralNode(true);
}), o("FALSE", function() {
return new LiteralNode(false);
}), o("YES", function() {
return new LiteralNode(true);
}), o("NO", function() {
return new LiteralNode(false);
}), o("ON", function() {
return new LiteralNode(true);
}), o("OFF", function() {
return new LiteralNode(false);
})
],
// Assignment of a variable, property, or index to a value.
Assign: [o("Value ASSIGN Expression", function() {
return new AssignNode($1, $3);
})
],
// Assignment when it happens within an object literal. The difference from
// the ordinary **Assign** is that these allow numbers and strings as keys.
AssignObj: [o("Identifier ASSIGN Expression", function() {
return new AssignNode(new ValueNode($1), $3, 'object');
}), o("AlphaNumeric ASSIGN Expression", function() {
return new AssignNode(new ValueNode($1), $3, 'object');
}), o("Comment")
],
// A return statement from a function body.
Return: [o("RETURN Expression", function() {
return new ReturnNode($2);
}), o("RETURN", function() {
return new ReturnNode(new ValueNode(new LiteralNode('null')));
})
],
// A comment. Because CoffeeScript passes comments through to JavaScript, we
// have to parse comments like any other construct, and identify all of the
// positions in which they can occur in the grammar.
Comment: [o("COMMENT", function() {
return new CommentNode(yytext);
})
],
// [The existential operator](http://jashkenas.github.com/coffee-script/#existence).
Existence: [o("Expression ?", function() {
return new ExistenceNode($1);
})
],
// The **Code** node is the function literal. It's defined by an indented block
// of **Expressions** preceded by a function arrow, with an optional parameter
// list.
Code: [o("PARAM_START ParamList PARAM_END FuncGlyph Block", function() {
return new CodeNode($2, $5, $4);
}), o("FuncGlyph Block", function() {
return new CodeNode([], $2, $1);
})
],
// CoffeeScript has two different symbols for functions. `->` is for ordinary
// functions, and `=>` is for functions bound to the current value of *this*.
FuncGlyph: [o("->", function() {
return 'func';
}), o("=>", function() {
return 'boundfunc';
})
],
// The list of parameters that a function accepts can be of any length.
ParamList: [o("", function() {
return [];
}), o("Param", function() {
return [$1];
}), o("ParamList , Param", function() {
return $1.concat([$3]);
})
],
// A single parameter in a function definition can be ordinary, or a splat
// that hoovers up the remaining arguments.
Param: [o("PARAM", function() {
return new LiteralNode(yytext);
}), o("Param . . .", function() {
return new SplatNode($1);
})
],
// A splat that occurs outside of a parameter list.
Splat: [o("Expression . . .", function() {
return new SplatNode($1);
})
],
// The types of things that can be treated as values -- assigned to, invoked
// as functions, indexed into, named as a class, etc.
Value: [o("Identifier", function() {
return new ValueNode($1);
}), o("Literal", function() {
return new ValueNode($1);
}), o("Array", function() {
return new ValueNode($1);
}), o("Object", function() {
return new ValueNode($1);
}), o("Parenthetical", function() {
return new ValueNode($1);
}), o("Range", function() {
return new ValueNode($1);
}), o("This"), o("Value Accessor", function() {
return $1.push($2);
}), o("Invocation Accessor", function() {
return new ValueNode($1, [$2]);
})
],
// The general group of accessors into an object, by property, by prototype
// or by array index or slice.
Accessor: [o("PROPERTY_ACCESS Identifier", function() {
return new AccessorNode($2);
}), o("PROTOTYPE_ACCESS Identifier", function() {
return new AccessorNode($2, 'prototype');
}), o("::", function() {
return new AccessorNode(new LiteralNode('prototype'));
}), o("SOAK_ACCESS Identifier", function() {
return new AccessorNode($2, 'soak');
}), o("Index"), o("Slice", function() {
return new SliceNode($1);
})
],
// Indexing into an object or array using bracket notation.
Index: [o("INDEX_START Expression INDEX_END", function() {
return new IndexNode($2);
}), o("SOAKED_INDEX_START Expression SOAKED_INDEX_END", function() {
return new IndexNode($2, 'soak');
})
],
// In CoffeeScript, an object literal is simply a list of assignments.
Object: [o("{ AssignList }", function() {
return new ObjectNode($2);
}), o("{ IndentedAssignList }", function() {
return new ObjectNode($2);
})
],
// Class definitions have optional bodies of prototype property assignments,
// and optional references to the superclass.
Class: [o("CLASS Value", function() {
return new ClassNode($2);
}), o("CLASS Value EXTENDS Value", function() {
return new ClassNode($2, $4);
}), o("CLASS Value IndentedAssignList", function() {
return new ClassNode($2, null, $3);
}), o("CLASS Value EXTENDS Value IndentedAssignList", function() {
return new ClassNode($2, $4, $5);
})
],
// Assignment of properties within an object literal can be separated by
// comma, as in JavaScript, or simply by newline.
AssignList: [o("", function() {
return [];
}), o("AssignObj", function() {
return [$1];
}), o("AssignList , AssignObj", function() {
return $1.concat([$3]);
}), o("AssignList TERMINATOR AssignObj", function() {
return $1.concat([$3]);
}), o("AssignList , TERMINATOR AssignObj", function() {
return $1.concat([$4]);
})
],
// An **AssignList** within a block indentation.
IndentedAssignList: [o("INDENT AssignList OUTDENT", function() {
return $2;
})
],
// The three flavors of function call: normal, object instantiation with `new`,
// and calling `super()`
Call: [o("Invocation"), o("NEW Invocation", function() {
return $2.new_instance();
}), o("Super")
],
Curry: [o("Value <- Arguments", function() {
return new CurryNode($1, $3);
})
],
// Extending an object by setting its prototype chain to reference a parent
// object.
Extends: [o("Value EXTENDS Value", function() {
return new ExtendsNode($1, $3);
})
],
// Ordinary function invocation, or a chained series of calls.
Invocation: [o("Value Arguments", function() {
return new CallNode($1, $2);
}), o("Invocation Arguments", function() {
return new CallNode($1, $2);
})
],
// The list of arguments to a function call.
Arguments: [o("CALL_START ArgList CALL_END", function() {
return $2;
})
],
// Calling super.
Super: [o("SUPER CALL_START ArgList CALL_END", function() {
return new CallNode('super', $3);
})
],
// A reference to the *this* current object, either naked or to a property.
This: [o("@", function() {
return new ValueNode(new LiteralNode('this'));
}), o("@ Identifier", function() {
return new ValueNode(new LiteralNode('this'), [new AccessorNode($2)]);
})
],
// The CoffeeScript range literal.
Range: [o("[ Expression . . Expression ]", function() {
return new RangeNode($2, $5);
}), o("[ Expression . . . Expression ]", function() {
return new RangeNode($2, $6, true);
})
],
// The slice literal.
Slice: [o("INDEX_START Expression . . Expression INDEX_END", function() {
return new RangeNode($2, $5);
}), o("INDEX_START Expression . . . Expression INDEX_END", function() {
return new RangeNode($2, $6, true);
})
],
// The array literal.
Array: [o("[ ArgList ]", function() {
return new ArrayNode($2);
})
],
// The **ArgList** is both the list of objects passed into a function call,
// as well as the contents of an array literal
// (i.e. comma-separated expressions). Newlines work as well.
ArgList: [o("", function() {
return [];
}), o("Expression", function() {
return [$1];
}), o("INDENT Expression", function() {
return [$2];
}), o("ArgList , Expression", function() {
return $1.concat([$3]);
}), o("ArgList TERMINATOR Expression", function() {
return $1.concat([$3]);
}), o("ArgList , TERMINATOR Expression", function() {
return $1.concat([$4]);
}), o("ArgList , INDENT Expression", function() {
return $1.concat([$4]);
}), o("ArgList OUTDENT")
],
// Just simple, comma-separated, required arguments (no fancy syntax). We need
// this to be separate from the **ArgList** for use in **Switch** blocks, where
// having the newlines wouldn't make sense.
SimpleArgs: [o("Expression"), o("SimpleArgs , Expression", function() {
if ($1 instanceof Array) {
return $1.concat([$3]);
} else {
return [$1].concat([$3]);
}
})
],
// The variants of *try/catch/finally* exception handling blocks.
Try: [o("TRY Block Catch", function() {
return new TryNode($2, $3[0], $3[1]);
}), o("TRY Block FINALLY Block", function() {
return new TryNode($2, null, null, $4);
}), o("TRY Block Catch FINALLY Block", function() {
return new TryNode($2, $3[0], $3[1], $5);
})
],
// A catch clause names its error and runs a block of code.
Catch: [o("CATCH Identifier Block", function() {
return [$2, $3];
})
],
// Throw an exception object.
Throw: [o("THROW Expression", function() {
return new ThrowNode($2);
})
],
// Parenthetical expressions. Note that the **Parenthetical** is a **Value**,
// not an **Expression**, so if you need to use an expression in a place
// where only values are accepted, wrapping it in parentheses will always do
// the trick.
Parenthetical: [o("( Expression )", function() {
return new ParentheticalNode($2);
})
],
// A language extension to CoffeeScript from the outside. We simply pass
// it through unaltered.
Extension: [o("EXTENSION", function() {
return yytext;
})
],
// The condition portion of a while loop.
WhileSource: [o("WHILE Expression", function() {
return new WhileNode($2);
}), o("WHILE Expression WHEN Expression", function() {
return new WhileNode($2, {
filter: $4
});
})
],
// The while loop can either be normal, with a block of expressions to execute,
// or postfix, with a single expression. There is no do..while.
While: [o("WhileSource Block", function() {
return $1.add_body($2);
}), o("Expression WhileSource", function() {
return $2.add_body(Expressions.wrap([$1]));
})
],
// Array, object, and range comprehensions, at the most generic level.
// Comprehensions can either be normal, with a block of expressions to execute,
// or postfix, with a single expression.
For: [o("Expression FOR ForVariables ForSource", function() {
return new ForNode($1, $4, $3[0], $3[1]);
}), o("FOR ForVariables ForSource Block", function() {
return new ForNode($4, $3, $2[0], $2[1]);
})
],
// An array or range comprehension has variables for the current element and
// (optional) reference to the current index. Or, *key, value*, in the case
// of object comprehensions.
ForVariables: [o("Identifier", function() {
return [$1];
}), o("Identifier , Identifier", function() {
return [$1, $3];
})
],
// The source of a comprehension is an array or object with an optional filter
// clause. If it's an array comprehension, you can also choose to step throug
// in fixed-size increments.
ForSource: [o("IN Expression", function() {
return {
source: $2
};
}), o("OF Expression", function() {
return {
source: $2,
object: true
};
}), o("ForSource WHEN Expression", function() {
$1.filter = $3;
return $1;
}), o("ForSource BY Expression", function() {
$1.step = $3;
return $1;
})
],
// The CoffeeScript switch/when/else block replaces the JavaScript
// switch/case/default by compiling into an if-else chain.
Switch: [o("SWITCH Expression INDENT Whens OUTDENT", function() {
return $4.rewrite_condition($2);
}), o("SWITCH Expression INDENT Whens ELSE Block OUTDENT", function() {
return $4.rewrite_condition($2).add_else($6, true);
})
],
// The inner list of whens is left recursive. At code-generation time, the
// IfNode will rewrite them into a proper chain.
Whens: [o("When"), o("Whens When", function() {
return $1.push($2);
})
],
// An individual **When** clause, with action.
When: [o("LEADING_WHEN SimpleArgs Block", function() {
return new IfNode($2, $3, null, {
statement: true
});
}), o("LEADING_WHEN SimpleArgs Block TERMINATOR", function() {
return new IfNode($2, $3, null, {
statement: true
});
}), o("Comment TERMINATOR When", function() {
$3.comment = $1;
return $3;
})
],
// The most basic form of *if* is a condition and an action. The following
// if-related rules are broken up along these lines in order to avoid
// ambiguity.
IfStart: [o("IF Expression Block", function() {
return new IfNode($2, $3);
}), o("IfStart ElsIf", function() {
return $1.add_else($2);
})
],
// An **IfStart** can optionally be followed by an else block.
IfBlock: [o("IfStart"), o("IfStart ELSE Block", function() {
return $1.add_else($3);
})
],
// An *else if* continuation of the *if* expression.
ElsIf: [o("ELSE IF Expression Block", function() {
return (new IfNode($3, $4)).force_statement();
})
],
// The full complement of *if* expressions, including postfix one-liner
// *if* and *unless*.
If: [o("IfBlock"), o("Expression IF Expression", function() {
return new IfNode($3, Expressions.wrap([$1]), null, {
statement: true
});
}), o("Expression UNLESS Expression", function() {
return new IfNode($3, Expressions.wrap([$1]), null, {
statement: true,
invert: true
});
})
],
// Arithmetic and logical operators, working on one or more operands.
// Here they are grouped by order of precedence. The actual precedence rules
// are defined at the bottom of the page. It would be shorter if we could
// combine most of these rules into a single generic *Operand OpSymbol Operand*
// -type rule, but in order to make the precedence binding possible, separate
// rules are necessary.
Operation: [o("! Expression", function() {
return new OpNode('!', $2);
}), o("!! Expression", function() {
return new OpNode('!!', $2);
}), o("- Expression", (function() {
return new OpNode('-', $2);
}), {
prec: 'UMINUS'
}), o("+ Expression", (function() {
return new OpNode('+', $2);
}), {
prec: 'UPLUS'
}), o("NOT Expression", function() {
return new OpNode('not', $2);
}), o("~ Expression", function() {
return new OpNode('~', $2);
}), o("-- Expression", function() {
return new OpNode('--', $2);
}), o("++ Expression", function() {
return new OpNode('++', $2);
}), o("DELETE Expression", function() {
return new OpNode('delete', $2);
}), o("TYPEOF Expression", function() {
return new OpNode('typeof', $2);
}), o("Expression --", function() {
return new OpNode('--', $1, null, true);
}), o("Expression ++", function() {
return new OpNode('++', $1, null, true);
}), o("Expression * Expression", function() {
return new OpNode('*', $1, $3);
}), o("Expression / Expression", function() {
return new OpNode('/', $1, $3);
}), o("Expression % Expression", function() {
return new OpNode('%', $1, $3);
}), o("Expression + Expression", function() {
return new OpNode('+', $1, $3);
}), o("Expression - Expression", function() {
return new OpNode('-', $1, $3);
}), o("Expression << Expression", function() {
return new OpNode('<<', $1, $3);
}), o("Expression >> Expression", function() {
return new OpNode('>>', $1, $3);
}), o("Expression >>> Expression", function() {
return new OpNode('>>>', $1, $3);
}), o("Expression & Expression", function() {
return new OpNode('&', $1, $3);
}), o("Expression | Expression", function() {
return new OpNode('|', $1, $3);
}), o("Expression ^ Expression", function() {
return new OpNode('^', $1, $3);
}), o("Expression <= Expression", function() {
return new OpNode('<=', $1, $3);
}), o("Expression < Expression", function() {
return new OpNode('<', $1, $3);
}), o("Expression > Expression", function() {
return new OpNode('>', $1, $3);
}), o("Expression >= Expression", function() {
return new OpNode('>=', $1, $3);
}), o("Expression == Expression", function() {
return new OpNode('==', $1, $3);
}), o("Expression != Expression", function() {
return new OpNode('!=', $1, $3);
}), o("Expression IS Expression", function() {
return new OpNode('is', $1, $3);
}), o("Expression ISNT Expression", function() {
return new OpNode('isnt', $1, $3);
}), o("Expression && Expression", function() {
return new OpNode('&&', $1, $3);
}), o("Expression || Expression", function() {
return new OpNode('||', $1, $3);
}), o("Expression AND Expression", function() {
return new OpNode('and', $1, $3);
}), o("Expression OR Expression", function() {
return new OpNode('or', $1, $3);
}), o("Expression ? Expression", function() {
return new OpNode('?', $1, $3);
}), o("Expression -= Expression", function() {
return new OpNode('-=', $1, $3);
}), o("Expression += Expression", function() {
return new OpNode('+=', $1, $3);
}), o("Expression /= Expression", function() {
return new OpNode('/=', $1, $3);
}), o("Expression *= Expression", function() {
return new OpNode('*=', $1, $3);
}), o("Expression %= Expression", function() {
return new OpNode('%=', $1, $3);
}), o("Expression ||= Expression", function() {
return new OpNode('||=', $1, $3);
}), o("Expression &&= Expression", function() {
return new OpNode('&&=', $1, $3);
}), o("Expression ?= Expression", function() {
return new OpNode('?=', $1, $3);
}), o("Expression INSTANCEOF Expression", function() {
return new OpNode('instanceof', $1, $3);
}), o("Expression IN Expression", function() {
return new OpNode('in', $1, $3);
})
]
};
// Precedence
// ----------
// Operators at the top of this list have higher precedence than the ones lower
// down. Following these rules is what makes `2 + 3 * 4` parse as:
// 2 + (3 * 4)
// And not:
// (2 + 3) * 4
operators = [["left", '?'], ["nonassoc", 'UMINUS', 'UPLUS', 'NOT', '!', '!!', '~', '++', '--'], ["left", '*', '/', '%'], ["left", '+', '-'], ["left", '<<', '>>', '>>>'], ["left", '&', '|', '^'], ["left", '<=', '<', '>', '>='], ["right", 'DELETE', 'INSTANCEOF', 'TYPEOF'], ["left", '==', '!=', 'IS', 'ISNT'], ["left", '&&', '||', 'AND', 'OR'], ["right", '-=', '+=', '/=', '*=', '%=', '||=', '&&=', '?='], ["left", '.'], ["right", 'INDENT'], ["left", 'OUTDENT'], ["right", 'WHEN', 'LEADING_WHEN', 'IN', 'OF', 'BY', 'THROW'], ["right", 'FOR', 'NEW', 'SUPER', 'CLASS'], ["left", 'EXTENDS'], ["right", 'ASSIGN', 'RETURN'], ["right", '->', '=>', '<-', 'UNLESS', 'IF', 'ELSE', 'WHILE']];
// Wrapping Up
// -----------
// Finally, now what we have our **grammar** and our **operators**, we can create
// our **Jison.Parser**. We do this by processing all of our rules, recording all
// terminals (every symbol which does not appear as the name of a rule above)
// as "tokens".
tokens = [];
_a = grammar;
for (name in _a) { if (__hasProp.call(_a, name)) {
alternatives = _a[name];
grammar[name] = (function() {
_b = []; _c = alternatives;
for (_d = 0, _e = _c.length; _d < _e; _d++) {
alt = _c[_d];
_b.push((function() {
_f = alt[0].split(' ');
for (_g = 0, _h = _f.length; _g < _h; _g++) {
token = _f[_g];
if (!(grammar[token])) {
tokens.push(token);
}
}
if (name === 'Root') {
alt[1] = "return " + (alt[1]);
}
return alt;
}).call(this));
}
return _b;
return null;
}).call(this);
}}
// Initialize the **Parser** with our list of terminal **tokens**, our **grammar**
// rules, and the name of the root. Reverse the operators because Jison orders
// precedence from low to high, and we have it high to low
// (as in [Yacc](http://dinosaur.compilertools.net/yacc/index.html)).
exports.parser = new Parser({
tokens: tokens.join(' '),
bnf: grammar,
operators: operators.reverse(),
startSymbol: 'Root'
});
})();