1
0
Fork 0
mirror of https://github.com/jashkenas/coffeescript.git synced 2022-11-09 12:23:24 -05:00
jashkenas--coffeescript/test/strings.coffee
Michal Srb dc0fb85fd3 [CS2] Support for CSX - equivalent of JSX (#4551)
* CSX implementation

* fixed comment, used toJS, added error tests, fixed error in identifier regex, fixed interpolation inside attributes value and added test

* added missing test for bare attributes, split attribute and indentifier regex, fixed checking for closing tags closing angle bracket

* Refactor tests that compare expected generated JavaScript with actual generated JavaScript to use common helper; add colors to error message to make differences easier to read

* Better match the style of the rest of the codebase

* Remove unused function

* More style fixes

* Allow unspaced less-than operator when not using CSX

* Replace includesCSX with a counter and simplify the unspaced operator logic

* Fixed indexing and realized that I completely enabled the tight spacing, added a test for it too

* Style fixes
2017-06-06 23:33:46 -07:00

430 lines
8.5 KiB
CoffeeScript
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

# String Literals
# ---------------
# TODO: refactor string literal tests
# TODO: add indexing and method invocation tests: "string"["toString"] is String::toString, "string".toString() is "string"
# * Strings
# * Heredocs
test "backslash escapes", ->
eq "\\/\\\\", /\/\\/.source
eq '(((dollars)))', '\(\(\(dollars\)\)\)'
eq 'one two three', "one
two
three"
eq "four five", 'four
five'
test "#3229, multiline strings", ->
# Separate lines by default by a single space in literal strings.
eq 'one
two', 'one two'
eq "one
two", 'one two'
eq '
a
b
', 'a b'
eq "
a
b
", 'a b'
eq 'one
two', 'one two'
eq "one
two", 'one two'
eq '
indentation
doesn\'t
matter', 'indentation doesn\'t matter'
eq 'trailing ws
doesn\'t matter', 'trailing ws doesn\'t matter'
# Use backslashes at the end of a line to specify whitespace between lines.
eq 'a \
b\
c \
d', 'a bc d'
eq "a \
b\
c \
d", 'a bc d'
eq 'ignore \
trailing whitespace', 'ignore trailing whitespace'
# Backslash at the beginning of a literal string.
eq '\
ok', 'ok'
eq ' \
ok', ' ok'
# #1273, empty strings.
eq '\
', ''
eq '
', ''
eq '
', ''
eq ' ', ' '
# Same behavior in interpolated strings.
eq "interpolation #{1}
follows #{2} \
too #{3}\
!", 'interpolation 1 follows 2 too 3!'
eq "a #{
'string ' + "inside
interpolation"
}", "a string inside interpolation"
eq "
#{1}
", '1'
# Handle escaped backslashes correctly.
eq '\\', `'\\'`
eq 'escaped backslash at EOL\\
next line', 'escaped backslash at EOL\\ next line'
eq '\\
next line', '\\ next line'
eq '\\
', '\\'
eq '\\\\\\
', '\\\\\\'
eq "#{1}\\
after interpolation", '1\\ after interpolation'
eq 'escaped backslash before slash\\ \
next line', 'escaped backslash before slash\\ next line'
eq 'triple backslash\\\
next line', 'triple backslash\\next line'
eq 'several escaped backslashes\\\\\\
ok', 'several escaped backslashes\\\\\\ ok'
eq 'several escaped backslashes slash\\\\\\\
ok', 'several escaped backslashes slash\\\\\\ok'
eq 'several escaped backslashes with trailing ws \\\\\\
ok', 'several escaped backslashes with trailing ws \\\\\\ ok'
# Backslashes at beginning of lines.
eq 'first line
\ backslash at BOL', 'first line \ backslash at BOL'
eq 'first line\
\ backslash at BOL', 'first line\ backslash at BOL'
# Backslashes at end of strings.
eq 'first line \ ', 'first line '
eq 'first line
second line \
', 'first line second line '
eq 'first line
second line
\
', 'first line second line'
eq 'first line
second line
\
', 'first line second line'
# Edge case.
eq 'lone
\
backslash', 'lone backslash'
test "#3249, escape newlines in heredocs with backslashes", ->
# Ignore escaped newlines
eq '''
Set whitespace \
<- this is ignored\
none
normal indentation
''', 'Set whitespace <- this is ignorednone\n normal indentation'
eq """
Set whitespace \
<- this is ignored\
none
normal indentation
""", 'Set whitespace <- this is ignorednone\n normal indentation'
# Changed from #647, trailing backslash.
eq '''
Hello, World\
''', 'Hello, World'
eq '''
\\
''', '\\'
# Backslash at the beginning of a literal string.
eq '''\
ok''', 'ok'
eq ''' \
ok''', ' ok'
# Same behavior in interpolated strings.
eq """
interpolation #{1}
follows #{2} \
too #{3}\
!
""", 'interpolation 1\n follows 2 too 3!'
eq """
#{1} #{2}
""", '\n1 2\n'
# Handle escaped backslashes correctly.
eq '''
escaped backslash at EOL\\
next line
''', 'escaped backslash at EOL\\\n next line'
eq '''\\
''', '\\\n'
# Backslashes at beginning of lines.
eq '''first line
\ backslash at BOL''', 'first line\n\ backslash at BOL'
eq """first line\
\ backslash at BOL""", 'first line\ backslash at BOL'
# Backslashes at end of strings.
eq '''first line \ ''', 'first line '
eq '''
first line
second line \
''', 'first line\nsecond line '
eq '''
first line
second line
\
''', 'first line\nsecond line'
eq '''
first line
second line
\
''', 'first line\nsecond line\n'
# Edge cases.
eq '''lone
\
backslash''', 'lone\n\n backslash'
eq '''\
''', ''
test '#2388: `"""` in heredoc interpolations', ->
eq """a heredoc #{
"inside \
interpolation"
}""", "a heredoc inside interpolation"
eq """a#{"""b"""}c""", 'abc'
eq """#{""""""}""", ''
test "trailing whitespace", ->
testTrailing = (str, expected) ->
eq CoffeeScript.eval(str.replace /\|$/gm, ''), expected
testTrailing '''" |
|
a |
|
"''', 'a'
testTrailing """''' |
|
a |
|
'''""", ' \na \n '
#647
eq "''Hello, World\\''", '''
'\'Hello, World\\\''
'''
eq '""Hello, World\\""', """
"\"Hello, World\\\""
"""
test "#1273, escaping quotes at the end of heredocs.", ->
# """\""" no longer compiles
eq """\\""", '\\'
eq """\\\"""", '\\\"'
a = """
basic heredoc
on two lines
"""
ok a is "basic heredoc\non two lines"
a = '''
a
"b
c
'''
ok a is "a\n \"b\nc"
a = """
a
b
c
"""
ok a is "a\n b\n c"
a = '''one-liner'''
ok a is 'one-liner'
a = """
out
here
"""
ok a is "out\nhere"
a = '''
a
b
c
'''
ok a is " a\n b\nc"
a = '''
a
b c
'''
ok a is "a\n\n\nb c"
a = '''more"than"one"quote'''
ok a is 'more"than"one"quote'
a = '''here's an apostrophe'''
ok a is "here's an apostrophe"
a = """""surrounded by two quotes"\""""
ok a is '""surrounded by two quotes""'
a = '''''surrounded by two apostrophes'\''''
ok a is "''surrounded by two apostrophes''"
# The indentation detector ignores blank lines without trailing whitespace
a = """
one
two
"""
ok a is "one\ntwo\n"
eq ''' line 0
should not be relevant
to the indent level
''', ' line 0\nshould not be relevant\n to the indent level'
eq """
interpolation #{
"contents"
}
should not be relevant
to the indent level
""", 'interpolation contents\nshould not be relevant\n to the indent level'
eq ''' '\\\' ''', " '\\' "
eq """ "\\\" """, ' "\\" '
eq ''' <- keep these spaces -> ''', ' <- keep these spaces -> '
eq '''undefined''', 'undefined'
eq """undefined""", 'undefined'
test "#1046, empty string interpolations", ->
eq "#{ }", ''
test "strings are not callable", ->
throws -> CoffeeScript.compile '"a"()'
throws -> CoffeeScript.compile '"a#{b}"()'
throws -> CoffeeScript.compile '"a" 1'
throws -> CoffeeScript.compile '"a#{b}" 1'
throws -> CoffeeScript.compile '''
"a"
k: v
'''
throws -> CoffeeScript.compile '''
"a#{b}"
k: v
'''
test "#3795: Escape otherwise invalid characters", ->
eq '', '\u2028'
eq '', '\u2029'
eq '\0\
1', '\x001'
eq "", '\u2028'
eq "", '\u2029'
eq "\0\
1", '\x001'
eq '''''', '\u2028'
eq '''''', '\u2029'
eq '''\0\
1''', '\x001'
eq """""", '\u2028'
eq """""", '\u2029'
eq """\0\
1""", '\x001'
a = 'a'
eq "#{a}", 'a\u2028'
eq "#{a}", 'a\u2029'
eq "#{a}\0\
1", 'a\0' + '1'
eq """#{a}""", 'a\u2028'
eq """#{a}""", 'a\u2029'
eq """#{a}\0\
1""", 'a\0' + '1'
test "#4314: Whitespace less than or equal to stripped indentation", ->
# The odd indentation is intentional here, to test 1-space indentation.
eq ' ', """
#{} #{}
"""
eq '1 2 3 4 5 end\na 0 b', """
#{1} #{2} #{3} #{4} #{5} end
a #{0} b"""
test "#4248: Unicode code point escapes", ->
eq '\u01ab\u00cd', '\u{1ab}\u{cd}'
eq '\u01ab', '\u{000001ab}'
eq 'a\u01ab', "#{ 'a' }\u{1ab}"
eq '\u01abc', '''\u{01ab}c'''
eq '\u01abc', """\u{1ab}#{ 'c' }"""
eq '\udab3\uddef', '\u{bcdef}'
eq '\udab3\uddef', '\u{0000bcdef}'
eq 'a\udab3\uddef', "#{ 'a' }\u{bcdef}"
eq '\udab3\uddefc', '''\u{0bcdef}c'''
eq '\udab3\uddefc', """\u{bcdef}#{ 'c' }"""
eq '\\u{123456}', "#{'\\'}#{'u{123456}'}"
# don't rewrite code point escapes
eqJS """
'\\u{bcdef}\\u{abc}'
""",
"""
'\\u{bcdef}\\u{abc}';
"""
eqJS """
"#{ 'a' }\\u{bcdef}"
""",
"""
"a\\u{bcdef}";
"""