parser 2.7.1.5 → 2.7.2.0

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (66) hide show
  1. checksums.yaml +4 -4
  2. data/lib/parser/current.rb +1 -1
  3. data/lib/parser/meta.rb +2 -2
  4. data/lib/parser/ruby28.rb +8047 -0
  5. data/lib/parser/version.rb +1 -1
  6. data/parser.gemspec +1 -20
  7. metadata +7 -96
  8. data/.travis.yml +0 -41
  9. data/.yardopts +0 -21
  10. data/CHANGELOG.md +0 -1137
  11. data/CONTRIBUTING.md +0 -17
  12. data/Gemfile +0 -10
  13. data/LICENSE.txt +0 -25
  14. data/README.md +0 -309
  15. data/Rakefile +0 -167
  16. data/ci/run_rubocop_specs +0 -14
  17. data/doc/AST_FORMAT.md +0 -2284
  18. data/doc/CUSTOMIZATION.md +0 -37
  19. data/doc/INTERNALS.md +0 -21
  20. data/doc/css/.gitkeep +0 -0
  21. data/doc/css/common.css +0 -68
  22. data/lib/parser/lexer.rl +0 -2550
  23. data/lib/parser/macruby.y +0 -2208
  24. data/lib/parser/ruby18.y +0 -1936
  25. data/lib/parser/ruby19.y +0 -2185
  26. data/lib/parser/ruby20.y +0 -2363
  27. data/lib/parser/ruby21.y +0 -2364
  28. data/lib/parser/ruby22.y +0 -2371
  29. data/lib/parser/ruby23.y +0 -2377
  30. data/lib/parser/ruby24.y +0 -2415
  31. data/lib/parser/ruby25.y +0 -2412
  32. data/lib/parser/ruby26.y +0 -2420
  33. data/lib/parser/ruby27.y +0 -2949
  34. data/lib/parser/ruby30.y +0 -3048
  35. data/lib/parser/rubymotion.y +0 -2192
  36. data/test/bug_163/fixtures/input.rb +0 -5
  37. data/test/bug_163/fixtures/output.rb +0 -5
  38. data/test/bug_163/rewriter.rb +0 -20
  39. data/test/helper.rb +0 -103
  40. data/test/parse_helper.rb +0 -328
  41. data/test/racc_coverage_helper.rb +0 -133
  42. data/test/test_ast_processor.rb +0 -32
  43. data/test/test_base.rb +0 -31
  44. data/test/test_current.rb +0 -31
  45. data/test/test_diagnostic.rb +0 -95
  46. data/test/test_diagnostic_engine.rb +0 -59
  47. data/test/test_encoding.rb +0 -99
  48. data/test/test_lexer.rb +0 -3617
  49. data/test/test_lexer_stack_state.rb +0 -78
  50. data/test/test_meta.rb +0 -12
  51. data/test/test_parse_helper.rb +0 -83
  52. data/test/test_parser.rb +0 -9986
  53. data/test/test_runner_parse.rb +0 -56
  54. data/test/test_runner_rewrite.rb +0 -47
  55. data/test/test_source_buffer.rb +0 -165
  56. data/test/test_source_comment.rb +0 -36
  57. data/test/test_source_comment_associator.rb +0 -399
  58. data/test/test_source_map.rb +0 -14
  59. data/test/test_source_range.rb +0 -192
  60. data/test/test_source_rewriter.rb +0 -541
  61. data/test/test_source_rewriter_action.rb +0 -46
  62. data/test/test_source_tree_rewriter.rb +0 -361
  63. data/test/test_static_environment.rb +0 -45
  64. data/test/using_tree_rewriter/fixtures/input.rb +0 -3
  65. data/test/using_tree_rewriter/fixtures/output.rb +0 -3
  66. data/test/using_tree_rewriter/using_tree_rewriter.rb +0 -9
@@ -1,37 +0,0 @@
1
- # Customizing Parsers
2
-
3
- While the default setup of the parsers provided by this Gem should be suitable
4
- for most some developers might want to change parts of it. An example would be
5
- the use of a custom class for nodes instead of `Parser::AST::Node`.
6
-
7
- Customizing the AST is done by creating a custom builder class and passing it
8
- to the constructor method of a parser. The default setup comes down to the
9
- following:
10
-
11
- builder = Parser::Builders::Default.new
12
- parser = Parser::Ruby19.new(builder)
13
-
14
- When creating your own builder class it's best to subclass the default one so
15
- that you don't have to redefine every used method again:
16
-
17
- class MyBuilder < Parser::Builders::Default
18
-
19
- end
20
-
21
- builder = MyBuilder.new
22
- parser = Parser::Ruby19.new(builder)
23
-
24
- ## Custom Node Classes
25
-
26
- To use a custom node class you have to override the method
27
- `Parser::Builders::Default#n`:
28
-
29
- class MyBuilder < Parser::Builders::Default
30
- def n(type, children, location)
31
- return MyNodeClass.new(type, children, :location => location)
32
- end
33
- end
34
-
35
- Note that the used class (and corresponding instance) must be compatible with
36
- `Parser::AST::Node` so it's best to subclass it and override/add code where
37
- needed.
@@ -1,21 +0,0 @@
1
- Entry points
2
- ------------
3
-
4
- Parser should be kept as slim as possible. This includes not loading
5
- any potentially large files when they are likely to be unused in practice.
6
-
7
- Parser has five main (classes of) `require` entry points:
8
-
9
- * `require 'parser'`. Main entry point, requires all classes which
10
- are used across the entire library.
11
- * `require 'parser/rubyXX'`. Version-specific entry point. Can raise
12
- a NotImplementedError if current Ruby runtime is unable to parse the
13
- requested Ruby version.
14
- * `require 'parser/all'`. Requires all available parsers for released
15
- versions of Ruby. Can raise NotImplementedError.
16
- * `require 'parser/runner'`. Requires all the stuff which is useful for
17
- command-line tools but not otherwise.
18
- * `require 'parser/runner/X'`. Runner-specific entry point.
19
-
20
- All non-main entry points internally `require 'parser'`. Additionally, all
21
- runner-specific entry points internally `requre 'parser/runner'`.
File without changes
@@ -1,68 +0,0 @@
1
- body
2
- {
3
- font-size: 14px;
4
- line-height: 1.6;
5
- margin: 0 auto;
6
- max-width: 960px;
7
- }
8
-
9
- p code
10
- {
11
- background: #f2f2f2;
12
- padding-left: 3px;
13
- padding-right: 3px;
14
- }
15
-
16
- pre.code
17
- {
18
- font-size: 13px;
19
- line-height: 1.4;
20
- }
21
-
22
- /**
23
- * YARD uses generic table styles, using a special class means those tables
24
- * don't get messed up.
25
- */
26
- .table
27
- {
28
- border: 1px solid #ccc;
29
- border-right: none;
30
- border-collapse: separate;
31
- border-spacing: 0;
32
- text-align: left;
33
- }
34
-
35
- .table.full
36
- {
37
- width: 100%;
38
- }
39
-
40
- .table .field_name
41
- {
42
- min-width: 160px;
43
- }
44
-
45
- .table thead tr th.no_sort:first-child
46
- {
47
- width: 25px;
48
- }
49
-
50
- .table thead tr th, .table tbody tr td
51
- {
52
- border-bottom: 1px solid #ccc;
53
- border-right: 1px solid #ccc;
54
- min-width: 20px;
55
- padding: 8px 5px;
56
- text-align: left;
57
- vertical-align: top;
58
- }
59
-
60
- .table tbody tr:last-child td
61
- {
62
- border-bottom: none;
63
- }
64
-
65
- .table tr:nth-child(odd) td
66
- {
67
- background: #f9f9f9;
68
- }
@@ -1,2550 +0,0 @@
1
- %%machine lex; # % fix highlighting
2
-
3
- #
4
- # === BEFORE YOU START ===
5
- #
6
- # Read the Ruby Hacking Guide chapter 11, available in English at
7
- # http://whitequark.org/blog/2013/04/01/ruby-hacking-guide-ch-11-finite-state-lexer/
8
- #
9
- # Remember two things about Ragel scanners:
10
- #
11
- # 1) Longest match wins.
12
- #
13
- # 2) If two matches have the same length, the first
14
- # in source code wins.
15
- #
16
- # General rules of making Ragel and Bison happy:
17
- #
18
- # * `p` (position) and `@te` contain the index of the character
19
- # they're pointing to ("current"), plus one. `@ts` contains the index
20
- # of the corresponding character. The code for extracting matched token is:
21
- #
22
- # @source_buffer.slice(@ts...@te)
23
- #
24
- # * If your input is `foooooooobar` and the rule is:
25
- #
26
- # 'f' 'o'+
27
- #
28
- # the result will be:
29
- #
30
- # foooooooobar
31
- # ^ ts=0 ^ p=te=9
32
- #
33
- # * A Ragel lexer action should not emit more than one token, unless
34
- # you know what you are doing.
35
- #
36
- # * All Ragel commands (fnext, fgoto, ...) end with a semicolon.
37
- #
38
- # * If an action emits the token and transitions to another state, use
39
- # these Ragel commands:
40
- #
41
- # emit($whatever)
42
- # fnext $next_state; fbreak;
43
- #
44
- # If you perform `fgoto` in an action which does not emit a token nor
45
- # rewinds the stream pointer, the parser's side-effectful,
46
- # context-sensitive lookahead actions will break in a hard to detect
47
- # and debug way.
48
- #
49
- # * If an action does not emit a token:
50
- #
51
- # fgoto $next_state;
52
- #
53
- # * If an action features lookbehind, i.e. matches characters with the
54
- # intent of passing them to another action:
55
- #
56
- # p = @ts - 1
57
- # fgoto $next_state;
58
- #
59
- # or, if the lookbehind consists of a single character:
60
- #
61
- # fhold; fgoto $next_state;
62
- #
63
- # * Ragel merges actions. So, if you have `e_lparen = '(' %act` and
64
- # `c_lparen = '('` and a lexer action `e_lparen | c_lparen`, the result
65
- # _will_ invoke the action `act`.
66
- #
67
- # e_something stands for "something with **e**mbedded action".
68
- #
69
- # * EOF is explicit and is matched by `c_eof`. If you want to introspect
70
- # the state of the lexer, add this rule to the state:
71
- #
72
- # c_eof => do_eof;
73
- #
74
- # * If you proceed past EOF, the lexer will complain:
75
- #
76
- # NoMethodError: undefined method `ord' for nil:NilClass
77
- #
78
-
79
- class Parser::Lexer
80
-
81
- %% write data nofinal;
82
- # %
83
-
84
- ESCAPES = {
85
- ?a.ord => "\a", ?b.ord => "\b", ?e.ord => "\e", ?f.ord => "\f",
86
- ?n.ord => "\n", ?r.ord => "\r", ?s.ord => "\s", ?t.ord => "\t",
87
- ?v.ord => "\v", ?\\.ord => "\\"
88
- }.freeze
89
-
90
- REGEXP_META_CHARACTERS = Regexp.union(*"\\$()*+.<>?[]^{|}".chars).freeze
91
-
92
- attr_reader :source_buffer
93
-
94
- attr_accessor :diagnostics
95
- attr_accessor :static_env
96
- attr_accessor :force_utf32
97
-
98
- attr_accessor :cond, :cmdarg, :in_kwarg, :context, :command_start
99
-
100
- attr_accessor :tokens, :comments
101
-
102
- def initialize(version)
103
- @version = version
104
- @static_env = nil
105
- @context = nil
106
-
107
- @tokens = nil
108
- @comments = nil
109
-
110
- reset
111
- end
112
-
113
- def reset(reset_state=true)
114
- # Ragel state:
115
- if reset_state
116
- # Unit tests set state prior to resetting lexer.
117
- @cs = self.class.lex_en_line_begin
118
-
119
- @cond = StackState.new('cond')
120
- @cmdarg = StackState.new('cmdarg')
121
- @cond_stack = []
122
- @cmdarg_stack = []
123
- end
124
-
125
- @force_utf32 = false # Set to true by some tests
126
-
127
- @source_pts = nil # @source as a codepoint array
128
-
129
- @p = 0 # stream position (saved manually in #advance)
130
- @ts = nil # token start
131
- @te = nil # token end
132
- @act = 0 # next action
133
-
134
- @stack = [] # state stack
135
- @top = 0 # state stack top pointer
136
-
137
- # Lexer state:
138
- @token_queue = []
139
- @literal_stack = []
140
-
141
- @eq_begin_s = nil # location of last encountered =begin
142
- @sharp_s = nil # location of last encountered #
143
-
144
- @newline_s = nil # location of last encountered newline
145
-
146
- @num_base = nil # last numeric base
147
- @num_digits_s = nil # starting position of numeric digits
148
- @num_suffix_s = nil # starting position of numeric suffix
149
- @num_xfrm = nil # numeric suffix-induced transformation
150
-
151
- @escape_s = nil # starting position of current sequence
152
- @escape = nil # last escaped sequence, as string
153
-
154
- @herebody_s = nil # starting position of current heredoc line
155
-
156
- # Ruby 1.9 ->() lambdas emit a distinct token if do/{ is
157
- # encountered after a matching closing parenthesis.
158
- @paren_nest = 0
159
- @lambda_stack = []
160
-
161
- # After encountering the closing line of <<~SQUIGGLY_HEREDOC,
162
- # we store the indentation level and give it out to the parser
163
- # on request. It is not possible to infer indentation level just
164
- # from the AST because escape sequences such as `\ ` or `\t` are
165
- # expanded inside the lexer, but count as non-whitespace for
166
- # indentation purposes.
167
- @dedent_level = nil
168
-
169
- # If the lexer is in `command state' (aka expr_value)
170
- # at the entry to #advance, it will transition to expr_cmdarg
171
- # instead of expr_arg at certain points.
172
- @command_start = true
173
-
174
- # True at the end of "def foo a:"
175
- @in_kwarg = false
176
-
177
- # State before =begin / =end block comment
178
- @cs_before_block_comment = self.class.lex_en_line_begin
179
- end
180
-
181
- def source_buffer=(source_buffer)
182
- @source_buffer = source_buffer
183
-
184
- if @source_buffer
185
- source = @source_buffer.source
186
-
187
- if source.encoding == Encoding::UTF_8
188
- @source_pts = source.unpack('U*')
189
- else
190
- @source_pts = source.unpack('C*')
191
- end
192
-
193
- if @source_pts[0] == 0xfeff
194
- # Skip byte order mark.
195
- @p = 1
196
- end
197
- else
198
- @source_pts = nil
199
- end
200
- end
201
-
202
- def encoding
203
- @source_buffer.source.encoding
204
- end
205
-
206
- LEX_STATES = {
207
- :line_begin => lex_en_line_begin,
208
- :expr_dot => lex_en_expr_dot,
209
- :expr_fname => lex_en_expr_fname,
210
- :expr_value => lex_en_expr_value,
211
- :expr_beg => lex_en_expr_beg,
212
- :expr_mid => lex_en_expr_mid,
213
- :expr_arg => lex_en_expr_arg,
214
- :expr_cmdarg => lex_en_expr_cmdarg,
215
- :expr_end => lex_en_expr_end,
216
- :expr_endarg => lex_en_expr_endarg,
217
- :expr_endfn => lex_en_expr_endfn,
218
- :expr_labelarg => lex_en_expr_labelarg,
219
-
220
- :interp_string => lex_en_interp_string,
221
- :interp_words => lex_en_interp_words,
222
- :plain_string => lex_en_plain_string,
223
- :plain_words => lex_en_plain_string,
224
- }
225
-
226
- def state
227
- LEX_STATES.invert.fetch(@cs, @cs)
228
- end
229
-
230
- def state=(state)
231
- @cs = LEX_STATES.fetch(state)
232
- end
233
-
234
- def push_cmdarg
235
- @cmdarg_stack.push(@cmdarg)
236
- @cmdarg = StackState.new("cmdarg.#{@cmdarg_stack.count}")
237
- end
238
-
239
- def pop_cmdarg
240
- @cmdarg = @cmdarg_stack.pop
241
- end
242
-
243
- def push_cond
244
- @cond_stack.push(@cond)
245
- @cond = StackState.new("cond.#{@cond_stack.count}")
246
- end
247
-
248
- def pop_cond
249
- @cond = @cond_stack.pop
250
- end
251
-
252
- def dedent_level
253
- # We erase @dedent_level as a precaution to avoid accidentally
254
- # using a stale value.
255
- dedent_level, @dedent_level = @dedent_level, nil
256
- dedent_level
257
- end
258
-
259
- # Return next token: [type, value].
260
- def advance
261
- if @token_queue.any?
262
- return @token_queue.shift
263
- end
264
-
265
- # Ugly, but dependent on Ragel output. Consider refactoring it somehow.
266
- klass = self.class
267
- _lex_trans_keys = klass.send :_lex_trans_keys
268
- _lex_key_spans = klass.send :_lex_key_spans
269
- _lex_index_offsets = klass.send :_lex_index_offsets
270
- _lex_indicies = klass.send :_lex_indicies
271
- _lex_trans_targs = klass.send :_lex_trans_targs
272
- _lex_trans_actions = klass.send :_lex_trans_actions
273
- _lex_to_state_actions = klass.send :_lex_to_state_actions
274
- _lex_from_state_actions = klass.send :_lex_from_state_actions
275
- _lex_eof_trans = klass.send :_lex_eof_trans
276
-
277
- pe = @source_pts.size + 2
278
- p, eof = @p, pe
279
-
280
- cmd_state = @command_start
281
- @command_start = false
282
-
283
- %% write exec;
284
- # %
285
-
286
- # Ragel creates a local variable called `testEof` but it doesn't use
287
- # it in any assignment. This dead code is here to swallow the warning.
288
- # It has no runtime cost because Ruby doesn't produce any instructions from it.
289
- if false
290
- testEof
291
- end
292
-
293
- @p = p
294
-
295
- if @token_queue.any?
296
- @token_queue.shift
297
- elsif @cs == klass.lex_error
298
- [ false, [ '$error'.freeze, range(p - 1, p) ] ]
299
- else
300
- eof = @source_pts.size
301
- [ false, [ '$eof'.freeze, range(eof, eof) ] ]
302
- end
303
- end
304
-
305
- protected
306
-
307
- def eof_codepoint?(point)
308
- [0x04, 0x1a, 0x00].include? point
309
- end
310
-
311
- def version?(*versions)
312
- versions.include?(@version)
313
- end
314
-
315
- def stack_pop
316
- @top -= 1
317
- @stack[@top]
318
- end
319
-
320
- def encode_escape(ord)
321
- ord.chr.force_encoding(@source_buffer.source.encoding)
322
- end
323
-
324
- def tok(s = @ts, e = @te)
325
- @source_buffer.slice(s...e)
326
- end
327
-
328
- def range(s = @ts, e = @te)
329
- Parser::Source::Range.new(@source_buffer, s, e)
330
- end
331
-
332
- def emit(type, value = tok, s = @ts, e = @te)
333
- token = [ type, [ value, range(s, e) ] ]
334
-
335
- @token_queue.push(token)
336
-
337
- @tokens.push(token) if @tokens
338
-
339
- token
340
- end
341
-
342
- def emit_table(table, s = @ts, e = @te)
343
- value = tok(s, e)
344
-
345
- emit(table[value], value, s, e)
346
- end
347
-
348
- def emit_do(do_block=false)
349
- if @cond.active?
350
- emit(:kDO_COND, 'do'.freeze)
351
- elsif @cmdarg.active? || do_block
352
- emit(:kDO_BLOCK, 'do'.freeze)
353
- else
354
- emit(:kDO, 'do'.freeze)
355
- end
356
- end
357
-
358
- def arg_or_cmdarg(cmd_state)
359
- if cmd_state
360
- self.class.lex_en_expr_cmdarg
361
- else
362
- self.class.lex_en_expr_arg
363
- end
364
- end
365
-
366
- def emit_comment(s = @ts, e = @te)
367
- if @comments
368
- @comments.push(Parser::Source::Comment.new(range(s, e)))
369
- end
370
-
371
- if @tokens
372
- @tokens.push([ :tCOMMENT, [ tok(s, e), range(s, e) ] ])
373
- end
374
-
375
- nil
376
- end
377
-
378
- def diagnostic(type, reason, arguments=nil, location=range, highlights=[])
379
- @diagnostics.process(
380
- Parser::Diagnostic.new(type, reason, arguments, location, highlights))
381
- end
382
-
383
- #
384
- # === LITERAL STACK ===
385
- #
386
-
387
- def push_literal(*args)
388
- new_literal = Literal.new(self, *args)
389
- @literal_stack.push(new_literal)
390
- next_state_for_literal(new_literal)
391
- end
392
-
393
- def next_state_for_literal(literal)
394
- if literal.words? && literal.backslash_delimited?
395
- if literal.interpolate?
396
- self.class.lex_en_interp_backslash_delimited_words
397
- else
398
- self.class.lex_en_plain_backslash_delimited_words
399
- end
400
- elsif literal.words? && !literal.backslash_delimited?
401
- if literal.interpolate?
402
- self.class.lex_en_interp_words
403
- else
404
- self.class.lex_en_plain_words
405
- end
406
- elsif !literal.words? && literal.backslash_delimited?
407
- if literal.interpolate?
408
- self.class.lex_en_interp_backslash_delimited
409
- else
410
- self.class.lex_en_plain_backslash_delimited
411
- end
412
- else
413
- if literal.interpolate?
414
- self.class.lex_en_interp_string
415
- else
416
- self.class.lex_en_plain_string
417
- end
418
- end
419
- end
420
-
421
- def literal
422
- @literal_stack.last
423
- end
424
-
425
- def pop_literal
426
- old_literal = @literal_stack.pop
427
-
428
- @dedent_level = old_literal.dedent_level
429
-
430
- if old_literal.type == :tREGEXP_BEG
431
- # Fetch modifiers.
432
- self.class.lex_en_regexp_modifiers
433
- else
434
- self.class.lex_en_expr_end
435
- end
436
- end
437
-
438
- # Mapping of strings to parser tokens.
439
-
440
- PUNCTUATION = {
441
- '=' => :tEQL, '&' => :tAMPER2, '|' => :tPIPE,
442
- '!' => :tBANG, '^' => :tCARET, '+' => :tPLUS,
443
- '-' => :tMINUS, '*' => :tSTAR2, '/' => :tDIVIDE,
444
- '%' => :tPERCENT, '~' => :tTILDE, ',' => :tCOMMA,
445
- ';' => :tSEMI, '.' => :tDOT, '..' => :tDOT2,
446
- '...' => :tDOT3, '[' => :tLBRACK2, ']' => :tRBRACK,
447
- '(' => :tLPAREN2, ')' => :tRPAREN, '?' => :tEH,
448
- ':' => :tCOLON, '&&' => :tANDOP, '||' => :tOROP,
449
- '-@' => :tUMINUS, '+@' => :tUPLUS, '~@' => :tTILDE,
450
- '**' => :tPOW, '->' => :tLAMBDA, '=~' => :tMATCH,
451
- '!~' => :tNMATCH, '==' => :tEQ, '!=' => :tNEQ,
452
- '>' => :tGT, '>>' => :tRSHFT, '>=' => :tGEQ,
453
- '<' => :tLT, '<<' => :tLSHFT, '<=' => :tLEQ,
454
- '=>' => :tASSOC, '::' => :tCOLON2, '===' => :tEQQ,
455
- '<=>' => :tCMP, '[]' => :tAREF, '[]=' => :tASET,
456
- '{' => :tLCURLY, '}' => :tRCURLY, '`' => :tBACK_REF2,
457
- '!@' => :tBANG, '&.' => :tANDDOT,
458
- }
459
-
460
- PUNCTUATION_BEGIN = {
461
- '&' => :tAMPER, '*' => :tSTAR, '**' => :tDSTAR,
462
- '+' => :tUPLUS, '-' => :tUMINUS, '::' => :tCOLON3,
463
- '(' => :tLPAREN, '{' => :tLBRACE, '[' => :tLBRACK,
464
- }
465
-
466
- KEYWORDS = {
467
- 'if' => :kIF_MOD, 'unless' => :kUNLESS_MOD,
468
- 'while' => :kWHILE_MOD, 'until' => :kUNTIL_MOD,
469
- 'rescue' => :kRESCUE_MOD, 'defined?' => :kDEFINED,
470
- 'BEGIN' => :klBEGIN, 'END' => :klEND,
471
- }
472
-
473
- KEYWORDS_BEGIN = {
474
- 'if' => :kIF, 'unless' => :kUNLESS,
475
- 'while' => :kWHILE, 'until' => :kUNTIL,
476
- 'rescue' => :kRESCUE, 'defined?' => :kDEFINED,
477
- 'BEGIN' => :klBEGIN, 'END' => :klEND,
478
- }
479
-
480
- %w(class module def undef begin end then elsif else ensure case when
481
- for break next redo retry in do return yield super self nil true
482
- false and or not alias __FILE__ __LINE__ __ENCODING__).each do |keyword|
483
- KEYWORDS_BEGIN[keyword] = KEYWORDS[keyword] = :"k#{keyword.upcase}"
484
- end
485
-
486
- %%{
487
- # %
488
-
489
- access @;
490
- getkey (@source_pts[p] || 0);
491
-
492
- # === CHARACTER CLASSES ===
493
- #
494
- # Pay close attention to the differences between c_any and any.
495
- # c_any does not include EOF and so will cause incorrect behavior
496
- # for machine subtraction (any-except rules) and default transitions
497
- # for scanners.
498
-
499
- action do_nl {
500
- # Record position of a newline for precise location reporting on tNL
501
- # tokens.
502
- #
503
- # This action is embedded directly into c_nl, as it is idempotent and
504
- # there are no cases when we need to skip it.
505
- @newline_s = p
506
- }
507
-
508
- c_nl = '\n' $ do_nl;
509
- c_space = [ \t\r\f\v];
510
- c_space_nl = c_space | c_nl;
511
-
512
- c_eof = 0x04 | 0x1a | 0 | zlen; # ^D, ^Z, \0, EOF
513
- c_eol = c_nl | c_eof;
514
- c_any = any - c_eof;
515
-
516
- c_nl_zlen = c_nl | zlen;
517
- c_line = any - c_nl_zlen;
518
-
519
- c_unicode = c_any - 0x00..0x7f;
520
- c_upper = [A-Z];
521
- c_lower = [a-z_] | c_unicode;
522
- c_alpha = c_lower | c_upper;
523
- c_alnum = c_alpha | [0-9];
524
-
525
- action do_eof {
526
- # Sit at EOF indefinitely. #advance would return $eof each time.
527
- # This allows to feed the lexer more data if needed; this is only used
528
- # in tests.
529
- #
530
- # Note that this action is not embedded into e_eof like e_heredoc_nl and e_bs
531
- # below. This is due to the fact that scanner state at EOF is observed
532
- # by tests, and encapsulating it in a rule would break the introspection.
533
- fhold; fbreak;
534
- }
535
-
536
- #
537
- # === TOKEN DEFINITIONS ===
538
- #
539
-
540
- # All operators are punctuation. There is more to punctuation
541
- # than just operators. Operators can be overridden by user;
542
- # punctuation can not.
543
-
544
- # A list of operators which are valid in the function name context, but
545
- # have different semantics in others.
546
- operator_fname = '[]' | '[]=' | '`' | '-@' | '+@' | '~@' | '!@' ;
547
-
548
- # A list of operators which can occur within an assignment shortcut (+ → +=).
549
- operator_arithmetic = '&' | '|' | '&&' | '||' | '^' | '+' | '-' |
550
- '*' | '/' | '**' | '~' | '<<' | '>>' | '%' ;
551
-
552
- # A list of all user-definable operators not covered by groups above.
553
- operator_rest = '=~' | '!~' | '==' | '!=' | '!' | '===' |
554
- '<' | '<=' | '>' | '>=' | '<=>' | '=>' ;
555
-
556
- # Note that `{` and `}` need to be referred to as e_lbrace and e_rbrace,
557
- # as they are ambiguous with interpolation `#{}` and should be counted.
558
- # These braces are not present in punctuation lists.
559
-
560
- # A list of punctuation which has different meaning when used at the
561
- # beginning of expression.
562
- punctuation_begin = '-' | '+' | '::' | '(' | '[' |
563
- '*' | '**' | '&' ;
564
-
565
- # A list of all punctuation except punctuation_begin.
566
- punctuation_end = ',' | '=' | '->' | '(' | '[' | ']' |
567
- '::' | '?' | ':' | '.' | '..' | '...' ;
568
-
569
- # A list of keywords which have different meaning at the beginning of expression.
570
- keyword_modifier = 'if' | 'unless' | 'while' | 'until' | 'rescue' ;
571
-
572
- # A list of keywords which accept an argument-like expression, i.e. have the
573
- # same post-processing as method calls or commands. Example: `yield 1`,
574
- # `yield (1)`, `yield(1)`, are interpreted as if `yield` was a function.
575
- keyword_with_arg = 'yield' | 'super' | 'not' | 'defined?' ;
576
-
577
- # A list of keywords which accept a literal function name as an argument.
578
- keyword_with_fname = 'def' | 'undef' | 'alias' ;
579
-
580
- # A list of keywords which accept an expression after them.
581
- keyword_with_value = 'else' | 'case' | 'ensure' | 'module' | 'elsif' | 'then' |
582
- 'for' | 'in' | 'do' | 'when' | 'begin' | 'class' |
583
- 'and' | 'or' ;
584
-
585
- # A list of keywords which accept a value, and treat the keywords from
586
- # `keyword_modifier` list as modifiers.
587
- keyword_with_mid = 'rescue' | 'return' | 'break' | 'next' ;
588
-
589
- # A list of keywords which do not accept an expression after them.
590
- keyword_with_end = 'end' | 'self' | 'true' | 'false' | 'retry' |
591
- 'redo' | 'nil' | 'BEGIN' | 'END' | '__FILE__' |
592
- '__LINE__' | '__ENCODING__';
593
-
594
- # All keywords.
595
- keyword = keyword_with_value | keyword_with_mid |
596
- keyword_with_end | keyword_with_arg |
597
- keyword_with_fname | keyword_modifier ;
598
-
599
- constant = c_upper c_alnum*;
600
- bareword = c_alpha c_alnum*;
601
-
602
- call_or_var = c_lower c_alnum*;
603
- class_var = '@@' bareword;
604
- instance_var = '@' bareword;
605
- global_var = '$'
606
- ( bareword | digit+
607
- | [`'+~*$&?!@/\\;,.=:<>"] # `
608
- | '-' c_alnum
609
- )
610
- ;
611
-
612
- # Ruby accepts (and fails on) variables with leading digit
613
- # in literal context, but not in unquoted symbol body.
614
- class_var_v = '@@' c_alnum+;
615
- instance_var_v = '@' c_alnum+;
616
-
617
- label = bareword [?!]? ':';
618
-
619
- #
620
- # === NUMERIC PARSING ===
621
- #
622
-
623
- int_hex = ( xdigit+ '_' )* xdigit* '_'? ;
624
- int_dec = ( digit+ '_' )* digit* '_'? ;
625
- int_bin = ( [01]+ '_' )* [01]* '_'? ;
626
-
627
- flo_int = [1-9] [0-9]* ( '_' digit+ )* | '0';
628
- flo_frac = '.' ( digit+ '_' )* digit+;
629
- flo_pow = [eE] [+\-]? ( digit+ '_' )* digit+;
630
-
631
- int_suffix =
632
- '' % { @num_xfrm = lambda { |chars| emit(:tINTEGER, chars) } }
633
- | 'r' % { @num_xfrm = lambda { |chars| emit(:tRATIONAL, Rational(chars)) } }
634
- | 'i' % { @num_xfrm = lambda { |chars| emit(:tIMAGINARY, Complex(0, chars)) } }
635
- | 'ri' % { @num_xfrm = lambda { |chars| emit(:tIMAGINARY, Complex(0, Rational(chars))) } }
636
- | 're' % { @num_xfrm = lambda { |chars| emit(:tINTEGER, chars, @ts, @te - 2); p -= 2 } }
637
- | 'if' % { @num_xfrm = lambda { |chars| emit(:tINTEGER, chars, @ts, @te - 2); p -= 2 } }
638
- | 'rescue' % { @num_xfrm = lambda { |chars| emit(:tINTEGER, chars, @ts, @te - 6); p -= 6 } };
639
-
640
- flo_pow_suffix =
641
- '' % { @num_xfrm = lambda { |chars| emit(:tFLOAT, Float(chars)) } }
642
- | 'i' % { @num_xfrm = lambda { |chars| emit(:tIMAGINARY, Complex(0, Float(chars))) } }
643
- | 'if' % { @num_xfrm = lambda { |chars| emit(:tFLOAT, Float(chars), @ts, @te - 2); p -= 2 } };
644
-
645
- flo_suffix =
646
- flo_pow_suffix
647
- | 'r' % { @num_xfrm = lambda { |chars| emit(:tRATIONAL, Rational(chars)) } }
648
- | 'ri' % { @num_xfrm = lambda { |chars| emit(:tIMAGINARY, Complex(0, Rational(chars))) } }
649
- | 'rescue' % { @num_xfrm = lambda { |chars| emit(:tFLOAT, Float(chars), @ts, @te - 6); p -= 6 } };
650
-
651
- #
652
- # === ESCAPE SEQUENCE PARSING ===
653
- #
654
-
655
- # Escape parsing code is a Ragel pattern, not a scanner, and therefore
656
- # it shouldn't directly raise errors or perform other actions with side effects.
657
- # In reality this would probably just mess up error reporting in pathological
658
- # cases, through.
659
-
660
- # The amount of code required to parse \M\C stuff correctly is ridiculous.
661
-
662
- escaped_nl = "\\" c_nl;
663
-
664
- action unicode_points {
665
- @escape = ""
666
-
667
- codepoints = tok(@escape_s + 2, p - 1)
668
- codepoint_s = @escape_s + 2
669
-
670
- if @version < 24
671
- if codepoints.start_with?(" ") || codepoints.start_with?("\t")
672
- diagnostic :fatal, :invalid_unicode_escape, nil,
673
- range(@escape_s + 2, @escape_s + 3)
674
- end
675
-
676
- if spaces_p = codepoints.index(/[ \t]{2}/)
677
- diagnostic :fatal, :invalid_unicode_escape, nil,
678
- range(codepoint_s + spaces_p + 1, codepoint_s + spaces_p + 2)
679
- end
680
-
681
- if codepoints.end_with?(" ") || codepoints.end_with?("\t")
682
- diagnostic :fatal, :invalid_unicode_escape, nil, range(p - 1, p)
683
- end
684
- end
685
-
686
- codepoints.scan(/([0-9a-fA-F]+)|([ \t]+)/).each do |(codepoint_str, spaces)|
687
- if spaces
688
- codepoint_s += spaces.length
689
- else
690
- codepoint = codepoint_str.to_i(16)
691
-
692
- if codepoint >= 0x110000
693
- diagnostic :error, :unicode_point_too_large, nil,
694
- range(codepoint_s, codepoint_s + codepoint_str.length)
695
- break
696
- end
697
-
698
- @escape += codepoint.chr(Encoding::UTF_8)
699
- codepoint_s += codepoint_str.length
700
- end
701
- end
702
- }
703
-
704
- action unescape_char {
705
- codepoint = @source_pts[p - 1]
706
- if (@escape = ESCAPES[codepoint]).nil?
707
- @escape = encode_escape(@source_buffer.slice(p - 1))
708
- end
709
- }
710
-
711
- action invalid_complex_escape {
712
- diagnostic :fatal, :invalid_escape
713
- }
714
-
715
- action read_post_meta_or_ctrl_char {
716
- @escape = @source_buffer.slice(p - 1).chr
717
-
718
- if @version >= 27 && ((0..8).include?(@escape.ord) || (14..31).include?(@escape.ord))
719
- diagnostic :fatal, :invalid_escape
720
- end
721
- }
722
-
723
- action slash_c_char {
724
- @escape = encode_escape(@escape[0].ord & 0x9f)
725
- }
726
-
727
- action slash_m_char {
728
- @escape = encode_escape(@escape[0].ord | 0x80)
729
- }
730
-
731
- maybe_escaped_char = (
732
- '\\' c_any %unescape_char
733
- | ( c_any - [\\] ) %read_post_meta_or_ctrl_char
734
- );
735
-
736
- maybe_escaped_ctrl_char = ( # why?!
737
- '\\' c_any %unescape_char %slash_c_char
738
- | '?' % { @escape = "\x7f" }
739
- | ( c_any - [\\?] ) %read_post_meta_or_ctrl_char %slash_c_char
740
- );
741
-
742
- escape = (
743
- # \377
744
- [0-7]{1,3}
745
- % { @escape = encode_escape(tok(@escape_s, p).to_i(8) % 0x100) }
746
-
747
- # \xff
748
- | 'x' xdigit{1,2}
749
- % { @escape = encode_escape(tok(@escape_s + 1, p).to_i(16)) }
750
-
751
- # %q[\x]
752
- | 'x' ( c_any - xdigit )
753
- % {
754
- diagnostic :fatal, :invalid_hex_escape, nil, range(@escape_s - 1, p + 2)
755
- }
756
-
757
- # \u263a
758
- | 'u' xdigit{4}
759
- % { @escape = tok(@escape_s + 1, p).to_i(16).chr(Encoding::UTF_8) }
760
-
761
- # \u123
762
- | 'u' xdigit{0,3}
763
- % {
764
- diagnostic :fatal, :invalid_unicode_escape, nil, range(@escape_s - 1, p)
765
- }
766
-
767
- # u{not hex} or u{}
768
- | 'u{' ( c_any - xdigit - [ \t}] )* '}'
769
- % {
770
- diagnostic :fatal, :invalid_unicode_escape, nil, range(@escape_s - 1, p)
771
- }
772
-
773
- # \u{ \t 123 \t 456 \t\t }
774
- | 'u{' [ \t]* ( xdigit{1,6} [ \t]+ )*
775
- (
776
- ( xdigit{1,6} [ \t]* '}'
777
- %unicode_points
778
- )
779
- |
780
- ( xdigit* ( c_any - xdigit - [ \t}] )+ '}'
781
- | ( c_any - [ \t}] )* c_eof
782
- | xdigit{7,}
783
- ) % {
784
- diagnostic :fatal, :unterminated_unicode, nil, range(p - 1, p)
785
- }
786
- )
787
-
788
- # \C-\a \cx
789
- | ( 'C-' | 'c' ) escaped_nl?
790
- maybe_escaped_ctrl_char
791
-
792
- # \M-a
793
- | 'M-' escaped_nl?
794
- maybe_escaped_char
795
- %slash_m_char
796
-
797
- # \C-\M-f \M-\cf \c\M-f
798
- | ( ( 'C-' | 'c' ) escaped_nl? '\\M-'
799
- | 'M-\\' escaped_nl? ( 'C-' | 'c' ) ) escaped_nl?
800
- maybe_escaped_ctrl_char
801
- %slash_m_char
802
-
803
- | 'C' c_any %invalid_complex_escape
804
- | 'M' c_any %invalid_complex_escape
805
- | ( 'M-\\C' | 'C-\\M' ) c_any %invalid_complex_escape
806
-
807
- | ( c_any - [0-7xuCMc] ) %unescape_char
808
-
809
- | c_eof % {
810
- diagnostic :fatal, :escape_eof, nil, range(p - 1, p)
811
- }
812
- );
813
-
814
- # Use rules in form of `e_bs escape' when you need to parse a sequence.
815
- e_bs = '\\' % {
816
- @escape_s = p
817
- @escape = nil
818
- };
819
-
820
- #
821
- # === STRING AND HEREDOC PARSING ===
822
- #
823
-
824
- # Heredoc parsing is quite a complex topic. First, consider that heredocs
825
- # can be arbitrarily nested. For example:
826
- #
827
- # puts <<CODE
828
- # the result is: #{<<RESULT.inspect
829
- # i am a heredoc
830
- # RESULT
831
- # }
832
- # CODE
833
- #
834
- # which, incidentally, evaluates to:
835
- #
836
- # the result is: " i am a heredoc\n"
837
- #
838
- # To parse them, lexer refers to two kinds (remember, nested heredocs)
839
- # of positions in the input stream, namely heredoc_e
840
- # (HEREDOC declaration End) and @herebody_s (HEREdoc BODY line Start).
841
- #
842
- # heredoc_e is simply contained inside the corresponding Literal, and
843
- # when the heredoc is closed, the lexing is restarted from that position.
844
- #
845
- # @herebody_s is quite more complex. First, @herebody_s changes after each
846
- # heredoc line is lexed. This way, at '\n' tok(@herebody_s, @te) always
847
- # contains the current line, and also when a heredoc is started, @herebody_s
848
- # contains the position from which the heredoc will be lexed.
849
- #
850
- # Second, as (insanity) there are nested heredocs, we need to maintain a
851
- # stack of these positions. Each time #push_literal is called, it saves current
852
- # @heredoc_s to literal.saved_herebody_s, and after an interpolation (possibly
853
- # containing another heredocs) is closed, the previous value is restored.
854
-
855
- e_heredoc_nl = c_nl % {
856
- # After every heredoc was parsed, @herebody_s contains the
857
- # position of next token after all heredocs.
858
- if @herebody_s
859
- p = @herebody_s
860
- @herebody_s = nil
861
- end
862
- };
863
-
864
- action extend_string {
865
- string = tok
866
-
867
- # tLABEL_END is only possible in non-cond context on >= 2.2
868
- if @version >= 22 && !@cond.active?
869
- lookahead = @source_buffer.slice(@te...@te+2)
870
- end
871
-
872
- current_literal = literal
873
- if !current_literal.heredoc? &&
874
- (token = current_literal.nest_and_try_closing(string, @ts, @te, lookahead))
875
- if token[0] == :tLABEL_END
876
- p += 1
877
- pop_literal
878
- fnext expr_labelarg;
879
- else
880
- fnext *pop_literal;
881
- end
882
- fbreak;
883
- else
884
- current_literal.extend_string(string, @ts, @te)
885
- end
886
- }
887
-
888
- action extend_string_escaped {
889
- current_literal = literal
890
- # Get the first character after the backslash.
891
- escaped_char = @source_buffer.slice(@escape_s).chr
892
-
893
- if current_literal.munge_escape? escaped_char
894
- # If this particular literal uses this character as an opening
895
- # or closing delimiter, it is an escape sequence for that
896
- # particular character. Write it without the backslash.
897
-
898
- if current_literal.regexp? && REGEXP_META_CHARACTERS.match(escaped_char)
899
- # Regular expressions should include escaped delimiters in their
900
- # escaped form, except when the escaped character is
901
- # a closing delimiter but not a regexp metacharacter.
902
- #
903
- # The backslash itself cannot be used as a closing delimiter
904
- # at the same time as an escape symbol, but it is always munged,
905
- # so this branch also executes for the non-closing-delimiter case
906
- # for the backslash.
907
- current_literal.extend_string(tok, @ts, @te)
908
- else
909
- current_literal.extend_string(escaped_char, @ts, @te)
910
- end
911
- else
912
- # It does not. So this is an actual escape sequence, yay!
913
- if current_literal.squiggly_heredoc? && escaped_char == "\n".freeze
914
- # Squiggly heredocs like
915
- # <<~-HERE
916
- # 1\
917
- # 2
918
- # HERE
919
- # treat '\' as a line continuation, but still dedent the body, so the heredoc above becomes "12\n".
920
- # This information is emitted as is, without escaping,
921
- # later this escape sequence (\\\n) gets handled manually in the Lexer::Dedenter
922
- current_literal.extend_string(tok, @ts, @te)
923
- elsif current_literal.supports_line_continuation_via_slash? && escaped_char == "\n".freeze
924
- # Heredocs, regexp and a few other types of literals support line
925
- # continuation via \\\n sequence. The code like
926
- # "a\
927
- # b"
928
- # must be parsed as "ab"
929
- current_literal.extend_string(tok.gsub("\\\n".freeze, ''.freeze), @ts, @te)
930
- elsif current_literal.regexp?
931
- # Regular expressions should include escape sequences in their
932
- # escaped form. On the other hand, escaped newlines are removed (in cases like "\\C-\\\n\\M-x")
933
- current_literal.extend_string(tok.gsub("\\\n".freeze, ''.freeze), @ts, @te)
934
- else
935
- current_literal.extend_string(@escape || tok, @ts, @te)
936
- end
937
- end
938
- }
939
-
940
- # Extend a string with a newline or a EOF character.
941
- # As heredoc closing line can immediately precede EOF, this action
942
- # has to handle such case specially.
943
- action extend_string_eol {
944
- current_literal = literal
945
- if @te == pe
946
- diagnostic :fatal, :string_eof, nil,
947
- range(current_literal.str_s, current_literal.str_s + 1)
948
- end
949
-
950
- if current_literal.heredoc?
951
- line = tok(@herebody_s, @ts).gsub(/\r+$/, ''.freeze)
952
-
953
- if version?(18, 19, 20)
954
- # See ruby:c48b4209c
955
- line = line.gsub(/\r.*$/, ''.freeze)
956
- end
957
-
958
- # Try ending the heredoc with the complete most recently
959
- # scanned line. @herebody_s always refers to the start of such line.
960
- if current_literal.nest_and_try_closing(line, @herebody_s, @ts)
961
- # Adjust @herebody_s to point to the next line.
962
- @herebody_s = @te
963
-
964
- # Continue regular lexing after the heredoc reference (<<END).
965
- p = current_literal.heredoc_e - 1
966
- fnext *pop_literal; fbreak;
967
- else
968
- # Calculate indentation level for <<~HEREDOCs.
969
- current_literal.infer_indent_level(line)
970
-
971
- # Ditto.
972
- @herebody_s = @te
973
- end
974
- else
975
- # Try ending the literal with a newline.
976
- if current_literal.nest_and_try_closing(tok, @ts, @te)
977
- fnext *pop_literal; fbreak;
978
- end
979
-
980
- if @herebody_s
981
- # This is a regular literal intertwined with a heredoc. Like:
982
- #
983
- # p <<-foo+"1
984
- # bar
985
- # foo
986
- # 2"
987
- #
988
- # which, incidentally, evaluates to "bar\n1\n2".
989
- p = @herebody_s - 1
990
- @herebody_s = nil
991
- end
992
- end
993
-
994
- if current_literal.words? && !eof_codepoint?(@source_pts[p])
995
- current_literal.extend_space @ts, @te
996
- else
997
- # A literal newline is appended if the heredoc was _not_ closed
998
- # this time (see fbreak above). See also Literal#nest_and_try_closing
999
- # for rationale of calling #flush_string here.
1000
- current_literal.extend_string tok, @ts, @te
1001
- current_literal.flush_string
1002
- end
1003
- }
1004
-
1005
- action extend_string_space {
1006
- literal.extend_space @ts, @te
1007
- }
1008
-
1009
- #
1010
- # === INTERPOLATION PARSING ===
1011
- #
1012
-
1013
- # Interpolations with immediate variable names simply call into
1014
- # the corresponding machine.
1015
-
1016
- interp_var = '#' ( global_var | class_var_v | instance_var_v );
1017
-
1018
- action extend_interp_var {
1019
- current_literal = literal
1020
- current_literal.flush_string
1021
- current_literal.extend_content
1022
-
1023
- emit(:tSTRING_DVAR, nil, @ts, @ts + 1)
1024
-
1025
- p = @ts
1026
- fcall expr_variable;
1027
- }
1028
-
1029
- # Special case for Ruby > 2.7
1030
- # If interpolated instance/class variable starts with a digit we parse it as a plain substring
1031
- # However, "#$1" is still a regular interpolation
1032
- interp_digit_var = '#' ('@' | '@@') digit c_alpha*;
1033
-
1034
- action extend_interp_digit_var {
1035
- if @version >= 27
1036
- literal.extend_string(tok, @ts, @te)
1037
- else
1038
- message = tok.start_with?('#@@') ? :cvar_name : :ivar_name
1039
- diagnostic :error, message, { :name => tok(@ts + 1, @te) }, range(@ts + 1, @te)
1040
- end
1041
- }
1042
-
1043
- # Interpolations with code blocks must match nested curly braces, as
1044
- # interpolation ending is ambiguous with a block ending. So, every
1045
- # opening and closing brace should be matched with e_[lr]brace rules,
1046
- # which automatically perform the counting.
1047
- #
1048
- # Note that interpolations can themselves be nested, so brace balance
1049
- # is tied to the innermost literal.
1050
- #
1051
- # Also note that literals themselves should not use e_[lr]brace rules
1052
- # when matching their opening and closing delimiters, as the amount of
1053
- # braces inside the characters of a string literal is independent.
1054
-
1055
- interp_code = '#{';
1056
-
1057
- e_lbrace = '{' % {
1058
- @cond.push(false); @cmdarg.push(false)
1059
-
1060
- current_literal = literal
1061
- if current_literal
1062
- current_literal.start_interp_brace
1063
- end
1064
- };
1065
-
1066
- e_rbrace = '}' % {
1067
- current_literal = literal
1068
- if current_literal
1069
- if current_literal.end_interp_brace_and_try_closing
1070
- if version?(18, 19)
1071
- emit(:tRCURLY, '}'.freeze, p - 1, p)
1072
- @cond.lexpop
1073
- @cmdarg.lexpop
1074
- else
1075
- emit(:tSTRING_DEND, '}'.freeze, p - 1, p)
1076
- end
1077
-
1078
- if current_literal.saved_herebody_s
1079
- @herebody_s = current_literal.saved_herebody_s
1080
- end
1081
-
1082
-
1083
- fhold;
1084
- fnext *next_state_for_literal(current_literal);
1085
- fbreak;
1086
- end
1087
- end
1088
-
1089
- @paren_nest -= 1
1090
- };
1091
-
1092
- action extend_interp_code {
1093
- current_literal = literal
1094
- current_literal.flush_string
1095
- current_literal.extend_content
1096
-
1097
- emit(:tSTRING_DBEG, '#{'.freeze)
1098
-
1099
- if current_literal.heredoc?
1100
- current_literal.saved_herebody_s = @herebody_s
1101
- @herebody_s = nil
1102
- end
1103
-
1104
- current_literal.start_interp_brace
1105
- @command_start = true
1106
- fnext expr_value;
1107
- fbreak;
1108
- }
1109
-
1110
- # Actual string parsers are simply combined from the primitives defined
1111
- # above.
1112
-
1113
- interp_words := |*
1114
- interp_code => extend_interp_code;
1115
- interp_digit_var => extend_interp_digit_var;
1116
- interp_var => extend_interp_var;
1117
- e_bs escape => extend_string_escaped;
1118
- c_space+ => extend_string_space;
1119
- c_eol => extend_string_eol;
1120
- c_any => extend_string;
1121
- *|;
1122
-
1123
- interp_string := |*
1124
- interp_code => extend_interp_code;
1125
- interp_digit_var => extend_interp_digit_var;
1126
- interp_var => extend_interp_var;
1127
- e_bs escape => extend_string_escaped;
1128
- c_eol => extend_string_eol;
1129
- c_any => extend_string;
1130
- *|;
1131
-
1132
- plain_words := |*
1133
- e_bs c_any => extend_string_escaped;
1134
- c_space+ => extend_string_space;
1135
- c_eol => extend_string_eol;
1136
- c_any => extend_string;
1137
- *|;
1138
-
1139
- plain_string := |*
1140
- '\\' c_nl => extend_string_eol;
1141
- e_bs c_any => extend_string_escaped;
1142
- c_eol => extend_string_eol;
1143
- c_any => extend_string;
1144
- *|;
1145
-
1146
- interp_backslash_delimited := |*
1147
- interp_code => extend_interp_code;
1148
- interp_digit_var => extend_interp_digit_var;
1149
- interp_var => extend_interp_var;
1150
- c_eol => extend_string_eol;
1151
- c_any => extend_string;
1152
- *|;
1153
-
1154
- plain_backslash_delimited := |*
1155
- c_eol => extend_string_eol;
1156
- c_any => extend_string;
1157
- *|;
1158
-
1159
- interp_backslash_delimited_words := |*
1160
- interp_code => extend_interp_code;
1161
- interp_digit_var => extend_interp_digit_var;
1162
- interp_var => extend_interp_var;
1163
- c_space+ => extend_string_space;
1164
- c_eol => extend_string_eol;
1165
- c_any => extend_string;
1166
- *|;
1167
-
1168
- plain_backslash_delimited_words := |*
1169
- c_space+ => extend_string_space;
1170
- c_eol => extend_string_eol;
1171
- c_any => extend_string;
1172
- *|;
1173
-
1174
- regexp_modifiers := |*
1175
- [A-Za-z]+
1176
- => {
1177
- unknown_options = tok.scan(/[^imxouesn]/)
1178
- if unknown_options.any?
1179
- diagnostic :error, :regexp_options,
1180
- { :options => unknown_options.join }
1181
- end
1182
-
1183
- emit(:tREGEXP_OPT)
1184
- fnext expr_end;
1185
- fbreak;
1186
- };
1187
-
1188
- any
1189
- => {
1190
- emit(:tREGEXP_OPT, tok(@ts, @te - 1), @ts, @te - 1)
1191
- fhold;
1192
- fgoto expr_end;
1193
- };
1194
- *|;
1195
-
1196
- #
1197
- # === WHITESPACE HANDLING ===
1198
- #
1199
-
1200
- # Various contexts in Ruby allow various kinds of whitespace
1201
- # to be used. They are grouped to clarify the lexing machines
1202
- # and ease collection of comments.
1203
-
1204
- # A line of code with inline #comment at end is always equivalent
1205
- # to a line of code ending with just a newline, so an inline
1206
- # comment is deemed equivalent to non-newline whitespace
1207
- # (c_space character class).
1208
-
1209
- w_space =
1210
- c_space+
1211
- | '\\' e_heredoc_nl
1212
- ;
1213
-
1214
- w_comment =
1215
- '#' %{ @sharp_s = p - 1 }
1216
- # The (p == pe) condition compensates for added "\0" and
1217
- # the way Ragel handles EOF.
1218
- c_line* %{ emit_comment(@sharp_s, p == pe ? p - 2 : p) }
1219
- ;
1220
-
1221
- w_space_comment =
1222
- w_space
1223
- | w_comment
1224
- ;
1225
-
1226
- # A newline in non-literal context always interoperates with
1227
- # here document logic and can always be escaped by a backslash,
1228
- # still interoperating with here document logic in the same way,
1229
- # yet being invisible to anything else.
1230
- #
1231
- # To demonstrate:
1232
- #
1233
- # foo = <<FOO \
1234
- # bar
1235
- # FOO
1236
- # + 2
1237
- #
1238
- # is equivalent to `foo = "bar\n" + 2`.
1239
-
1240
- w_newline =
1241
- e_heredoc_nl;
1242
-
1243
- w_any =
1244
- w_space
1245
- | w_comment
1246
- | w_newline
1247
- ;
1248
-
1249
-
1250
- #
1251
- # === EXPRESSION PARSING ===
1252
- #
1253
-
1254
- # These rules implement a form of manually defined lookahead.
1255
- # The default longest-match scanning does not work here due
1256
- # to sheer ambiguity.
1257
-
1258
- ambiguous_fid_suffix = # actual parsed
1259
- [?!] %{ tm = p } | # a? a?
1260
- [?!]'=' %{ tm = p - 2 } # a!=b a != b
1261
- ;
1262
-
1263
- ambiguous_ident_suffix = # actual parsed
1264
- ambiguous_fid_suffix |
1265
- '=' %{ tm = p } | # a= a=
1266
- '==' %{ tm = p - 2 } | # a==b a == b
1267
- '=~' %{ tm = p - 2 } | # a=~b a =~ b
1268
- '=>' %{ tm = p - 2 } | # a=>b a => b
1269
- '===' %{ tm = p - 3 } # a===b a === b
1270
- ;
1271
-
1272
- ambiguous_symbol_suffix = # actual parsed
1273
- ambiguous_ident_suffix |
1274
- '==>' %{ tm = p - 2 } # :a==>b :a= => b
1275
- ;
1276
-
1277
- # Ambiguous with 1.9 hash labels.
1278
- ambiguous_const_suffix = # actual parsed
1279
- '::' %{ tm = p - 2 } # A::B A :: B
1280
- ;
1281
-
1282
- # Resolving kDO/kDO_COND/kDO_BLOCK ambiguity requires embedding
1283
- # @cond/@cmdarg-related code to e_lbrack, e_lparen and e_lbrace.
1284
-
1285
- e_lbrack = '[' % {
1286
- @cond.push(false); @cmdarg.push(false)
1287
-
1288
- @paren_nest += 1
1289
- };
1290
-
1291
- e_rbrack = ']' % {
1292
- @paren_nest -= 1
1293
- };
1294
-
1295
- # Ruby 1.9 lambdas require parentheses counting in order to
1296
- # emit correct opening kDO/tLBRACE.
1297
-
1298
- e_lparen = '(' % {
1299
- @cond.push(false); @cmdarg.push(false)
1300
-
1301
- @paren_nest += 1
1302
-
1303
- if version?(18)
1304
- @command_start = true
1305
- end
1306
- };
1307
-
1308
- e_rparen = ')' % {
1309
- @paren_nest -= 1
1310
- };
1311
-
1312
- # Ruby is context-sensitive wrt/ local identifiers.
1313
- action local_ident {
1314
- emit(:tIDENTIFIER)
1315
-
1316
- if !@static_env.nil? && @static_env.declared?(tok)
1317
- fnext expr_endfn; fbreak;
1318
- else
1319
- fnext *arg_or_cmdarg(cmd_state); fbreak;
1320
- end
1321
- }
1322
-
1323
- # Variable lexing code is accessed from both expressions and
1324
- # string interpolation related code.
1325
- #
1326
- expr_variable := |*
1327
- global_var
1328
- => {
1329
- if tok =~ /^\$([1-9][0-9]*)$/
1330
- emit(:tNTH_REF, tok(@ts + 1).to_i)
1331
- elsif tok =~ /^\$([&`'+])$/
1332
- emit(:tBACK_REF)
1333
- else
1334
- emit(:tGVAR)
1335
- end
1336
-
1337
- fnext *stack_pop; fbreak;
1338
- };
1339
-
1340
- class_var_v
1341
- => {
1342
- if tok =~ /^@@[0-9]/
1343
- diagnostic :error, :cvar_name, { :name => tok }
1344
- end
1345
-
1346
- emit(:tCVAR)
1347
- fnext *stack_pop; fbreak;
1348
- };
1349
-
1350
- instance_var_v
1351
- => {
1352
- if tok =~ /^@[0-9]/
1353
- diagnostic :error, :ivar_name, { :name => tok }
1354
- end
1355
-
1356
- emit(:tIVAR)
1357
- fnext *stack_pop; fbreak;
1358
- };
1359
- *|;
1360
-
1361
- # Literal function name in definition (e.g. `def class`).
1362
- # Keywords are returned as their respective tokens; this is used
1363
- # to support singleton def `def self.foo`. Global variables are
1364
- # returned as `tGVAR`; this is used in global variable alias
1365
- # statements `alias $a $b`. Symbols are returned verbatim; this
1366
- # is used in `alias :a :"b#{foo}"` and `undef :a`.
1367
- #
1368
- # Transitions to `expr_endfn` afterwards.
1369
- #
1370
- expr_fname := |*
1371
- keyword
1372
- => { emit_table(KEYWORDS_BEGIN);
1373
- fnext expr_endfn; fbreak; };
1374
-
1375
- constant
1376
- => { emit(:tCONSTANT)
1377
- fnext expr_endfn; fbreak; };
1378
-
1379
- bareword [?=!]?
1380
- => { emit(:tIDENTIFIER)
1381
- fnext expr_endfn; fbreak; };
1382
-
1383
- global_var
1384
- => { p = @ts - 1
1385
- fnext expr_end; fcall expr_variable; };
1386
-
1387
- # If the handling was to be delegated to expr_end,
1388
- # these cases would transition to something else than
1389
- # expr_endfn, which is incorrect.
1390
- operator_fname |
1391
- operator_arithmetic |
1392
- operator_rest
1393
- => { emit_table(PUNCTUATION)
1394
- fnext expr_endfn; fbreak; };
1395
-
1396
- '::'
1397
- => { fhold; fhold; fgoto expr_end; };
1398
-
1399
- ':'
1400
- => { fhold; fgoto expr_beg; };
1401
-
1402
- '%s' c_any
1403
- => {
1404
- if version?(23)
1405
- type, delimiter = tok[0..-2], tok[-1].chr
1406
- fgoto *push_literal(type, delimiter, @ts);
1407
- else
1408
- p = @ts - 1
1409
- fgoto expr_end;
1410
- end
1411
- };
1412
-
1413
- w_any;
1414
-
1415
- c_any
1416
- => { fhold; fgoto expr_end; };
1417
-
1418
- c_eof => do_eof;
1419
- *|;
1420
-
1421
- # After literal function name in definition. Behaves like `expr_end`,
1422
- # but allows a tLABEL.
1423
- #
1424
- # Transitions to `expr_end` afterwards.
1425
- #
1426
- expr_endfn := |*
1427
- label ( any - ':' )
1428
- => { emit(:tLABEL, tok(@ts, @te - 2), @ts, @te - 1)
1429
- fhold; fnext expr_labelarg; fbreak; };
1430
-
1431
- w_space_comment;
1432
-
1433
- c_any
1434
- => { fhold; fgoto expr_end; };
1435
-
1436
- c_eof => do_eof;
1437
- *|;
1438
-
1439
- # Literal function name in method call (e.g. `a.class`).
1440
- #
1441
- # Transitions to `expr_arg` afterwards.
1442
- #
1443
- expr_dot := |*
1444
- constant
1445
- => { emit(:tCONSTANT)
1446
- fnext *arg_or_cmdarg(cmd_state); fbreak; };
1447
-
1448
- call_or_var
1449
- => { emit(:tIDENTIFIER)
1450
- fnext *arg_or_cmdarg(cmd_state); fbreak; };
1451
-
1452
- bareword ambiguous_fid_suffix
1453
- => { emit(:tFID, tok(@ts, tm), @ts, tm)
1454
- fnext *arg_or_cmdarg(cmd_state); p = tm - 1; fbreak; };
1455
-
1456
- # See the comment in `expr_fname`.
1457
- operator_fname |
1458
- operator_arithmetic |
1459
- operator_rest
1460
- => { emit_table(PUNCTUATION)
1461
- fnext expr_arg; fbreak; };
1462
-
1463
- w_any;
1464
-
1465
- c_any
1466
- => { fhold; fgoto expr_end; };
1467
-
1468
- c_eof => do_eof;
1469
- *|;
1470
-
1471
- # The previous token emitted was a `tIDENTIFIER` or `tFID`; no space
1472
- # is consumed; the current expression is a command or method call.
1473
- #
1474
- expr_arg := |*
1475
- #
1476
- # COMMAND MODE SPECIFIC TOKENS
1477
- #
1478
-
1479
- # cmd (1 + 2)
1480
- # See below the rationale about expr_endarg.
1481
- w_space+ e_lparen
1482
- => {
1483
- if version?(18)
1484
- emit(:tLPAREN2, '('.freeze, @te - 1, @te)
1485
- fnext expr_value; fbreak;
1486
- else
1487
- emit(:tLPAREN_ARG, '('.freeze, @te - 1, @te)
1488
- fnext expr_beg; fbreak;
1489
- end
1490
- };
1491
-
1492
- # meth(1 + 2)
1493
- # Regular method call.
1494
- e_lparen
1495
- => { emit(:tLPAREN2, '('.freeze)
1496
- fnext expr_beg; fbreak; };
1497
-
1498
- # meth [...]
1499
- # Array argument. Compare with indexing `meth[...]`.
1500
- w_space+ e_lbrack
1501
- => { emit(:tLBRACK, '['.freeze, @te - 1, @te)
1502
- fnext expr_beg; fbreak; };
1503
-
1504
- # cmd {}
1505
- # Command: method call without parentheses.
1506
- w_space* e_lbrace
1507
- => {
1508
- if @lambda_stack.last == @paren_nest
1509
- @lambda_stack.pop
1510
- emit(:tLAMBEG, '{'.freeze, @te - 1, @te)
1511
- else
1512
- emit(:tLCURLY, '{'.freeze, @te - 1, @te)
1513
- end
1514
- @command_start = true
1515
- @paren_nest += 1
1516
- fnext expr_value; fbreak;
1517
- };
1518
-
1519
- #
1520
- # AMBIGUOUS TOKENS RESOLVED VIA EXPR_BEG
1521
- #
1522
-
1523
- # a??
1524
- # Ternary operator
1525
- '?' c_space_nl
1526
- => {
1527
- # Unlike expr_beg as invoked in the next rule, do not warn
1528
- p = @ts - 1
1529
- fgoto expr_end;
1530
- };
1531
-
1532
- # a ?b, a? ?
1533
- # Character literal or ternary operator
1534
- w_space* '?'
1535
- => { fhold; fgoto expr_beg; };
1536
-
1537
- # a %{1}, a %[1] (but not "a %=1=" or "a % foo")
1538
- # a /foo/ (but not "a / foo" or "a /=foo")
1539
- # a <<HEREDOC
1540
- w_space+ %{ tm = p }
1541
- ( [%/] ( c_any - c_space_nl - '=' ) # /
1542
- | '<<'
1543
- )
1544
- => {
1545
- if tok(tm, tm + 1) == '/'.freeze
1546
- # Ambiguous regexp literal.
1547
- diagnostic :warning, :ambiguous_literal, nil, range(tm, tm + 1)
1548
- end
1549
-
1550
- p = tm - 1
1551
- fgoto expr_beg;
1552
- };
1553
-
1554
- # x *1
1555
- # Ambiguous splat, kwsplat or block-pass.
1556
- w_space+ %{ tm = p } ( '+' | '-' | '*' | '&' | '**' )
1557
- => {
1558
- diagnostic :warning, :ambiguous_prefix, { :prefix => tok(tm, @te) },
1559
- range(tm, @te)
1560
-
1561
- p = tm - 1
1562
- fgoto expr_beg;
1563
- };
1564
-
1565
- # x ::Foo
1566
- # Ambiguous toplevel constant access.
1567
- w_space+ '::'
1568
- => { fhold; fhold; fgoto expr_beg; };
1569
-
1570
- # x:b
1571
- # Symbol.
1572
- w_space* ':'
1573
- => { fhold; fgoto expr_beg; };
1574
-
1575
- w_space+ label
1576
- => { p = @ts - 1; fgoto expr_beg; };
1577
-
1578
- #
1579
- # AMBIGUOUS TOKENS RESOLVED VIA EXPR_END
1580
- #
1581
-
1582
- # a ? b
1583
- # Ternary operator.
1584
- w_space+ %{ tm = p } '?' c_space_nl
1585
- => { p = tm - 1; fgoto expr_end; };
1586
-
1587
- # x + 1: Binary operator or operator-assignment.
1588
- w_space* operator_arithmetic
1589
- ( '=' | c_space_nl )? |
1590
- # x rescue y: Modifier keyword.
1591
- w_space* keyword_modifier |
1592
- # a &. b: Safe navigation operator.
1593
- w_space* '&.' |
1594
- # Miscellanea.
1595
- w_space* punctuation_end
1596
- => {
1597
- p = @ts - 1
1598
- fgoto expr_end;
1599
- };
1600
-
1601
- w_space;
1602
-
1603
- w_comment
1604
- => { fgoto expr_end; };
1605
-
1606
- w_newline
1607
- => { fhold; fgoto expr_end; };
1608
-
1609
- c_any
1610
- => { fhold; fgoto expr_beg; };
1611
-
1612
- c_eof => do_eof;
1613
- *|;
1614
-
1615
- # The previous token was an identifier which was seen while in the
1616
- # command mode (that is, the state at the beginning of #advance was
1617
- # expr_value). This state is very similar to expr_arg, but disambiguates
1618
- # two very rare and specific condition:
1619
- # * In 1.8 mode, "foo (lambda do end)".
1620
- # * In 1.9+ mode, "f x: -> do foo do end end".
1621
- expr_cmdarg := |*
1622
- w_space+ e_lparen
1623
- => {
1624
- emit(:tLPAREN_ARG, '('.freeze, @te - 1, @te)
1625
- if version?(18)
1626
- fnext expr_value; fbreak;
1627
- else
1628
- fnext expr_beg; fbreak;
1629
- end
1630
- };
1631
-
1632
- w_space* 'do'
1633
- => {
1634
- if @cond.active?
1635
- emit(:kDO_COND, 'do'.freeze, @te - 2, @te)
1636
- else
1637
- emit(:kDO, 'do'.freeze, @te - 2, @te)
1638
- end
1639
- fnext expr_value; fbreak;
1640
- };
1641
-
1642
- c_any |
1643
- # Disambiguate with the `do' rule above.
1644
- w_space* bareword |
1645
- w_space* label
1646
- => { p = @ts - 1
1647
- fgoto expr_arg; };
1648
-
1649
- c_eof => do_eof;
1650
- *|;
1651
-
1652
- # The rationale for this state is pretty complex. Normally, if an argument
1653
- # is passed to a command and then there is a block (tLCURLY...tRCURLY),
1654
- # the block is attached to the innermost argument (`f` in `m f {}`), or it
1655
- # is a parse error (`m 1 {}`). But there is a special case for passing a single
1656
- # primary expression grouped with parentheses: if you write `m (1) {}` or
1657
- # (2.0 only) `m () {}`, then the block is attached to `m`.
1658
- #
1659
- # Thus, we recognize the opening `(` of a command (remember, a command is
1660
- # a method call without parens) as a tLPAREN_ARG; then, in parser, we recognize
1661
- # `tLPAREN_ARG expr rparen` as a `primary_expr` and before rparen, set the
1662
- # lexer's state to `expr_endarg`, which makes it emit the possibly following
1663
- # `{` as `tLBRACE_ARG`.
1664
- #
1665
- # The default post-`expr_endarg` state is `expr_end`, so this state also handles
1666
- # `do` (as `kDO_BLOCK` in `expr_beg`).
1667
- expr_endarg := |*
1668
- e_lbrace
1669
- => {
1670
- if @lambda_stack.last == @paren_nest
1671
- @lambda_stack.pop
1672
- emit(:tLAMBEG, '{'.freeze)
1673
- else
1674
- emit(:tLBRACE_ARG, '{'.freeze)
1675
- end
1676
- @paren_nest += 1
1677
- @command_start = true
1678
- fnext expr_value; fbreak;
1679
- };
1680
-
1681
- 'do'
1682
- => { emit_do(true)
1683
- fnext expr_value; fbreak; };
1684
-
1685
- w_space_comment;
1686
-
1687
- c_any
1688
- => { fhold; fgoto expr_end; };
1689
-
1690
- c_eof => do_eof;
1691
- *|;
1692
-
1693
- # The rationale for this state is that several keywords accept value
1694
- # (i.e. should transition to `expr_beg`), do not accept it like a command
1695
- # (i.e. not an `expr_arg`), and must behave like a statement, that is,
1696
- # accept a modifier if/while/etc.
1697
- #
1698
- expr_mid := |*
1699
- keyword_modifier
1700
- => { emit_table(KEYWORDS)
1701
- fnext expr_beg; fbreak; };
1702
-
1703
- bareword
1704
- => { p = @ts - 1; fgoto expr_beg; };
1705
-
1706
- w_space_comment;
1707
-
1708
- w_newline
1709
- => { fhold; fgoto expr_end; };
1710
-
1711
- c_any
1712
- => { fhold; fgoto expr_beg; };
1713
-
1714
- c_eof => do_eof;
1715
- *|;
1716
-
1717
- # Beginning of an expression.
1718
- #
1719
- # Don't fallthrough to this state from `c_any`; make sure to handle
1720
- # `c_space* c_nl` and let `expr_end` handle the newline.
1721
- # Otherwise code like `f\ndef x` gets glued together and the parser
1722
- # explodes.
1723
- #
1724
- expr_beg := |*
1725
- # +5, -5, - 5
1726
- [+\-] w_any* [0-9]
1727
- => {
1728
- emit(:tUNARY_NUM, tok(@ts, @ts + 1), @ts, @ts + 1)
1729
- fhold; fnext expr_end; fbreak;
1730
- };
1731
-
1732
- # splat *a
1733
- '*'
1734
- => { emit(:tSTAR, '*'.freeze)
1735
- fbreak; };
1736
-
1737
- #
1738
- # STRING AND REGEXP LITERALS
1739
- #
1740
-
1741
- # /regexp/oui
1742
- # /=/ (disambiguation with /=)
1743
- '/' c_any
1744
- => {
1745
- type = delimiter = tok[0].chr
1746
- fhold; fgoto *push_literal(type, delimiter, @ts);
1747
- };
1748
-
1749
- # %<string>
1750
- '%' ( any - [A-Za-z] )
1751
- => {
1752
- type, delimiter = @source_buffer.slice(@ts).chr, tok[-1].chr
1753
- fgoto *push_literal(type, delimiter, @ts);
1754
- };
1755
-
1756
- # %w(we are the people)
1757
- '%' [A-Za-z]+ c_any
1758
- => {
1759
- type, delimiter = tok[0..-2], tok[-1].chr
1760
- fgoto *push_literal(type, delimiter, @ts);
1761
- };
1762
-
1763
- '%' c_eof
1764
- => {
1765
- diagnostic :fatal, :string_eof, nil, range(@ts, @ts + 1)
1766
- };
1767
-
1768
- # Heredoc start.
1769
- # <<END | <<'END' | <<"END" | <<`END` |
1770
- # <<-END | <<-'END' | <<-"END" | <<-`END` |
1771
- # <<~END | <<~'END' | <<~"END" | <<~`END`
1772
- '<<' [~\-]?
1773
- ( '"' ( any - '"' )* '"'
1774
- | "'" ( any - "'" )* "'"
1775
- | "`" ( any - "`" )* "`"
1776
- | bareword ) % { heredoc_e = p }
1777
- c_line* c_nl % { new_herebody_s = p }
1778
- => {
1779
- tok(@ts, heredoc_e) =~ /^<<(-?)(~?)(["'`]?)(.*)\3$/m
1780
-
1781
- indent = !$1.empty? || !$2.empty?
1782
- dedent_body = !$2.empty?
1783
- type = $3.empty? ? '<<"'.freeze : ('<<'.freeze + $3)
1784
- delimiter = $4
1785
-
1786
- if @version >= 27
1787
- if delimiter.count("\n") > 0 || delimiter.count("\r") > 0
1788
- diagnostic :error, :unterminated_heredoc_id, nil, range(@ts, @ts + 1)
1789
- end
1790
- elsif @version >= 24
1791
- if delimiter.count("\n") > 0
1792
- if delimiter.end_with?("\n")
1793
- diagnostic :warning, :heredoc_id_ends_with_nl, nil, range(@ts, @ts + 1)
1794
- delimiter = delimiter.rstrip
1795
- else
1796
- diagnostic :fatal, :heredoc_id_has_newline, nil, range(@ts, @ts + 1)
1797
- end
1798
- end
1799
- end
1800
-
1801
- if dedent_body && version?(18, 19, 20, 21, 22)
1802
- emit(:tLSHFT, '<<'.freeze, @ts, @ts + 2)
1803
- p = @ts + 1
1804
- fnext expr_beg; fbreak;
1805
- else
1806
- fnext *push_literal(type, delimiter, @ts, heredoc_e, indent, dedent_body);
1807
-
1808
- @herebody_s ||= new_herebody_s
1809
- p = @herebody_s - 1
1810
- end
1811
- };
1812
-
1813
- # Escaped unterminated heredoc start
1814
- # <<'END | <<"END | <<`END |
1815
- # <<-'END | <<-"END | <<-`END |
1816
- # <<~'END | <<~"END | <<~`END
1817
- #
1818
- # If the heredoc is terminated the rule above should handle it
1819
- '<<' [~\-]?
1820
- ('"' (any - c_nl - '"')*
1821
- |"'" (any - c_nl - "'")*
1822
- |"`" (any - c_nl - "`")
1823
- )
1824
- => {
1825
- diagnostic :error, :unterminated_heredoc_id, nil, range(@ts, @ts + 1)
1826
- };
1827
-
1828
- #
1829
- # SYMBOL LITERALS
1830
- #
1831
-
1832
- # :&&, :||
1833
- ':' ('&&' | '||') => {
1834
- fhold; fhold;
1835
- emit(:tSYMBEG, tok(@ts, @ts + 1), @ts, @ts + 1)
1836
- fgoto expr_fname;
1837
- };
1838
-
1839
- # :"bar", :'baz'
1840
- ':' ['"] # '
1841
- => {
1842
- type, delimiter = tok, tok[-1].chr
1843
- fgoto *push_literal(type, delimiter, @ts);
1844
- };
1845
-
1846
- # :!@ is :!
1847
- # :~@ is :~
1848
- ':' [!~] '@'
1849
- => {
1850
- emit(:tSYMBOL, tok(@ts + 1, @ts + 2))
1851
- fnext expr_end; fbreak;
1852
- };
1853
-
1854
- ':' bareword ambiguous_symbol_suffix
1855
- => {
1856
- emit(:tSYMBOL, tok(@ts + 1, tm), @ts, tm)
1857
- p = tm - 1
1858
- fnext expr_end; fbreak;
1859
- };
1860
-
1861
- ':' ( bareword | global_var | class_var | instance_var |
1862
- operator_fname | operator_arithmetic | operator_rest )
1863
- => {
1864
- emit(:tSYMBOL, tok(@ts + 1), @ts)
1865
- fnext expr_end; fbreak;
1866
- };
1867
-
1868
- ':' ( '@' %{ tm = p - 1; diag_msg = :ivar_name }
1869
- | '@@' %{ tm = p - 2; diag_msg = :cvar_name }
1870
- ) [0-9]*
1871
- => {
1872
- if @version >= 27
1873
- diagnostic :error, diag_msg, { name: tok(tm, @te) }, range(tm, @te)
1874
- else
1875
- emit(:tCOLON, tok(@ts, @ts + 1), @ts, @ts + 1)
1876
- p = @ts
1877
- end
1878
-
1879
- fnext expr_end; fbreak;
1880
- };
1881
-
1882
- #
1883
- # AMBIGUOUS TERNARY OPERATOR
1884
- #
1885
-
1886
- # Character constant, like ?a, ?\n, ?\u1000, and so on
1887
- # Don't accept \u escape with multiple codepoints, like \u{1 2 3}
1888
- '?' ( e_bs ( escape - ( '\u{' (xdigit+ [ \t]+)+ xdigit+ '}' ))
1889
- | (c_any - c_space_nl - e_bs) % { @escape = nil }
1890
- )
1891
- => {
1892
- value = @escape || tok(@ts + 1)
1893
-
1894
- if version?(18)
1895
- emit(:tINTEGER, value.getbyte(0))
1896
- else
1897
- emit(:tCHARACTER, value)
1898
- end
1899
-
1900
- fnext expr_end; fbreak;
1901
- };
1902
-
1903
- '?' c_space_nl
1904
- => {
1905
- escape = { " " => '\s', "\r" => '\r', "\n" => '\n', "\t" => '\t',
1906
- "\v" => '\v', "\f" => '\f' }[@source_buffer.slice(@ts + 1)]
1907
- diagnostic :warning, :invalid_escape_use, { :escape => escape }, range
1908
-
1909
- p = @ts - 1
1910
- fgoto expr_end;
1911
- };
1912
-
1913
- '?' c_eof
1914
- => {
1915
- diagnostic :fatal, :incomplete_escape, nil, range(@ts, @ts + 1)
1916
- };
1917
-
1918
- # f ?aa : b: Disambiguate with a character literal.
1919
- '?' [A-Za-z_] bareword
1920
- => {
1921
- p = @ts - 1
1922
- fgoto expr_end;
1923
- };
1924
-
1925
- #
1926
- # AMBIGUOUS EMPTY BLOCK ARGUMENTS
1927
- #
1928
-
1929
- # Ruby >= 2.7 emits it as two tPIPE terminals
1930
- # while Ruby < 2.7 as a single tOROP (like in `a || b`)
1931
- '||'
1932
- => {
1933
- if @version >= 27
1934
- emit(:tPIPE, tok(@ts, @ts + 1), @ts, @ts + 1)
1935
- fhold;
1936
- fnext expr_beg; fbreak;
1937
- else
1938
- p -= 2
1939
- fgoto expr_end;
1940
- end
1941
- };
1942
-
1943
- #
1944
- # KEYWORDS AND PUNCTUATION
1945
- #
1946
-
1947
- # a({b=>c})
1948
- e_lbrace
1949
- => {
1950
- if @lambda_stack.last == @paren_nest
1951
- @lambda_stack.pop
1952
- @command_start = true
1953
- emit(:tLAMBEG, '{'.freeze)
1954
- else
1955
- emit(:tLBRACE, '{'.freeze)
1956
- end
1957
- @paren_nest += 1
1958
- fbreak;
1959
- };
1960
-
1961
- # a([1, 2])
1962
- e_lbrack
1963
- => { emit(:tLBRACK, '['.freeze)
1964
- fbreak; };
1965
-
1966
- # a()
1967
- e_lparen
1968
- => { emit(:tLPAREN, '('.freeze)
1969
- fbreak; };
1970
-
1971
- # a(+b)
1972
- punctuation_begin
1973
- => { emit_table(PUNCTUATION_BEGIN)
1974
- fbreak; };
1975
-
1976
- # rescue Exception => e: Block rescue.
1977
- # Special because it should transition to expr_mid.
1978
- 'rescue' %{ tm = p } '=>'?
1979
- => { emit(:kRESCUE, 'rescue'.freeze, @ts, tm)
1980
- p = tm - 1
1981
- fnext expr_mid; fbreak; };
1982
-
1983
- # if a: Statement if.
1984
- keyword_modifier
1985
- => { emit_table(KEYWORDS_BEGIN)
1986
- @command_start = true
1987
- fnext expr_value; fbreak; };
1988
-
1989
- #
1990
- # RUBY 1.9 HASH LABELS
1991
- #
1992
-
1993
- label ( any - ':' )
1994
- => {
1995
- fhold;
1996
-
1997
- if version?(18)
1998
- ident = tok(@ts, @te - 2)
1999
-
2000
- emit((@source_buffer.slice(@ts) =~ /[A-Z]/) ? :tCONSTANT : :tIDENTIFIER,
2001
- ident, @ts, @te - 2)
2002
- fhold; # continue as a symbol
2003
-
2004
- if !@static_env.nil? && @static_env.declared?(ident)
2005
- fnext expr_end;
2006
- else
2007
- fnext *arg_or_cmdarg(cmd_state);
2008
- end
2009
- else
2010
- emit(:tLABEL, tok(@ts, @te - 2), @ts, @te - 1)
2011
- fnext expr_labelarg;
2012
- end
2013
-
2014
- fbreak;
2015
- };
2016
-
2017
- #
2018
- # RUBY 2.7 BEGINLESS RANGE
2019
-
2020
- '..'
2021
- => {
2022
- if @version >= 27
2023
- emit(:tBDOT2)
2024
- else
2025
- emit(:tDOT2)
2026
- end
2027
-
2028
- fnext expr_beg; fbreak;
2029
- };
2030
-
2031
- '...'
2032
- => {
2033
- if @version >= 30
2034
- if @lambda_stack.any? && @lambda_stack.last + 1 == @paren_nest
2035
- # To reject `->(...)` like `->...`
2036
- emit(:tDOT3)
2037
- else
2038
- emit(:tBDOT3)
2039
- end
2040
- elsif @version >= 27
2041
- emit(:tBDOT3)
2042
- else
2043
- emit(:tDOT3)
2044
- end
2045
-
2046
- fnext expr_beg; fbreak;
2047
- };
2048
-
2049
- #
2050
- # CONTEXT-DEPENDENT VARIABLE LOOKUP OR COMMAND INVOCATION
2051
- #
2052
-
2053
- # foo= bar: Disambiguate with bareword rule below.
2054
- bareword ambiguous_ident_suffix |
2055
- # def foo: Disambiguate with bareword rule below.
2056
- keyword
2057
- => { p = @ts - 1
2058
- fgoto expr_end; };
2059
-
2060
- # a = 42; a [42]: Indexing.
2061
- # def a; end; a [42]: Array argument.
2062
- call_or_var
2063
- => local_ident;
2064
-
2065
- (call_or_var - keyword)
2066
- % { ident_tok = tok; ident_ts = @ts; ident_te = @te; }
2067
- w_space+ '('
2068
- => {
2069
- emit(:tIDENTIFIER, ident_tok, ident_ts, ident_te)
2070
- p = ident_te - 1
2071
-
2072
- if !@static_env.nil? && @static_env.declared?(ident_tok) && @version < 25
2073
- fnext expr_endfn;
2074
- else
2075
- fnext expr_cmdarg;
2076
- end
2077
- fbreak;
2078
- };
2079
-
2080
- #
2081
- # WHITESPACE
2082
- #
2083
-
2084
- w_any;
2085
-
2086
- e_heredoc_nl '=begin' ( c_space | c_nl_zlen )
2087
- => {
2088
- p = @ts - 1
2089
- @cs_before_block_comment = @cs
2090
- fgoto line_begin;
2091
- };
2092
-
2093
- #
2094
- # DEFAULT TRANSITION
2095
- #
2096
-
2097
- # The following rules match most binary and all unary operators.
2098
- # Rules for binary operators provide better error reporting.
2099
- operator_arithmetic '=' |
2100
- operator_rest |
2101
- punctuation_end |
2102
- c_any
2103
- => { p = @ts - 1; fgoto expr_end; };
2104
-
2105
- c_eof => do_eof;
2106
- *|;
2107
-
2108
- # Special newline handling for "def a b:"
2109
- #
2110
- expr_labelarg := |*
2111
- w_space_comment;
2112
-
2113
- w_newline
2114
- => {
2115
- if @in_kwarg
2116
- fhold; fgoto expr_end;
2117
- else
2118
- fgoto line_begin;
2119
- end
2120
- };
2121
-
2122
- c_any
2123
- => { fhold; fgoto expr_beg; };
2124
-
2125
- c_eof => do_eof;
2126
- *|;
2127
-
2128
- # Like expr_beg, but no 1.9 label or 2.2 quoted label possible.
2129
- #
2130
- expr_value := |*
2131
- # a:b: a(:b), a::B, A::B
2132
- label (any - ':')
2133
- => { p = @ts - 1
2134
- fgoto expr_end; };
2135
-
2136
- # "bar", 'baz'
2137
- ['"] # '
2138
- => {
2139
- fgoto *push_literal(tok, tok, @ts);
2140
- };
2141
-
2142
- w_space_comment;
2143
-
2144
- w_newline
2145
- => { fgoto line_begin; };
2146
-
2147
- c_any
2148
- => { fhold; fgoto expr_beg; };
2149
-
2150
- c_eof => do_eof;
2151
- *|;
2152
-
2153
- expr_end := |*
2154
- #
2155
- # STABBY LAMBDA
2156
- #
2157
-
2158
- '->'
2159
- => {
2160
- emit(:tLAMBDA, '->'.freeze, @ts, @ts + 2)
2161
-
2162
- @lambda_stack.push @paren_nest
2163
- fnext expr_endfn; fbreak;
2164
- };
2165
-
2166
- e_lbrace | 'do'
2167
- => {
2168
- if @lambda_stack.last == @paren_nest
2169
- @lambda_stack.pop
2170
-
2171
- if tok == '{'.freeze
2172
- emit(:tLAMBEG, '{'.freeze)
2173
- else # 'do'
2174
- emit(:kDO_LAMBDA, 'do'.freeze)
2175
- end
2176
- else
2177
- if tok == '{'.freeze
2178
- emit(:tLCURLY, '{'.freeze)
2179
- else # 'do'
2180
- emit_do
2181
- end
2182
- end
2183
- if tok == '{'.freeze
2184
- @paren_nest += 1
2185
- end
2186
- @command_start = true
2187
-
2188
- fnext expr_value; fbreak;
2189
- };
2190
-
2191
- #
2192
- # KEYWORDS
2193
- #
2194
-
2195
- keyword_with_fname
2196
- => { emit_table(KEYWORDS)
2197
- fnext expr_fname; fbreak; };
2198
-
2199
- 'class' w_any* '<<'
2200
- => { emit(:kCLASS, 'class'.freeze, @ts, @ts + 5)
2201
- emit(:tLSHFT, '<<'.freeze, @te - 2, @te)
2202
- fnext expr_value; fbreak; };
2203
-
2204
- # a if b:c: Syntax error.
2205
- keyword_modifier
2206
- => { emit_table(KEYWORDS)
2207
- fnext expr_beg; fbreak; };
2208
-
2209
- # elsif b:c: elsif b(:c)
2210
- keyword_with_value
2211
- => { emit_table(KEYWORDS)
2212
- @command_start = true
2213
- fnext expr_value; fbreak; };
2214
-
2215
- keyword_with_mid
2216
- => { emit_table(KEYWORDS)
2217
- fnext expr_mid; fbreak; };
2218
-
2219
- keyword_with_arg
2220
- => {
2221
- emit_table(KEYWORDS)
2222
-
2223
- if version?(18) && tok == 'not'.freeze
2224
- fnext expr_beg; fbreak;
2225
- else
2226
- fnext expr_arg; fbreak;
2227
- end
2228
- };
2229
-
2230
- '__ENCODING__'
2231
- => {
2232
- if version?(18)
2233
- emit(:tIDENTIFIER)
2234
-
2235
- unless !@static_env.nil? && @static_env.declared?(tok)
2236
- fnext *arg_or_cmdarg(cmd_state);
2237
- end
2238
- else
2239
- emit(:k__ENCODING__, '__ENCODING__'.freeze)
2240
- end
2241
- fbreak;
2242
- };
2243
-
2244
- keyword_with_end
2245
- => { emit_table(KEYWORDS)
2246
- fbreak; };
2247
-
2248
- #
2249
- # NUMERIC LITERALS
2250
- #
2251
-
2252
- ( '0' [Xx] %{ @num_base = 16; @num_digits_s = p } int_hex
2253
- | '0' [Dd] %{ @num_base = 10; @num_digits_s = p } int_dec
2254
- | '0' [Oo] %{ @num_base = 8; @num_digits_s = p } int_dec
2255
- | '0' [Bb] %{ @num_base = 2; @num_digits_s = p } int_bin
2256
- | [1-9] digit* '_'? %{ @num_base = 10; @num_digits_s = @ts } int_dec
2257
- | '0' digit* '_'? %{ @num_base = 8; @num_digits_s = @ts } int_dec
2258
- ) %{ @num_suffix_s = p } int_suffix
2259
- => {
2260
- digits = tok(@num_digits_s, @num_suffix_s)
2261
-
2262
- if digits.end_with? '_'.freeze
2263
- diagnostic :error, :trailing_in_number, { :character => '_'.freeze },
2264
- range(@te - 1, @te)
2265
- elsif digits.empty? && @num_base == 8 && version?(18)
2266
- # 1.8 did not raise an error on 0o.
2267
- digits = '0'.freeze
2268
- elsif digits.empty?
2269
- diagnostic :error, :empty_numeric
2270
- elsif @num_base == 8 && (invalid_idx = digits.index(/[89]/))
2271
- invalid_s = @num_digits_s + invalid_idx
2272
- diagnostic :error, :invalid_octal, nil,
2273
- range(invalid_s, invalid_s + 1)
2274
- end
2275
-
2276
- if version?(18, 19, 20)
2277
- emit(:tINTEGER, digits.to_i(@num_base), @ts, @num_suffix_s)
2278
- p = @num_suffix_s - 1
2279
- else
2280
- @num_xfrm.call(digits.to_i(@num_base))
2281
- end
2282
- fbreak;
2283
- };
2284
-
2285
- flo_frac flo_pow?
2286
- => {
2287
- diagnostic :error, :no_dot_digit_literal
2288
- };
2289
-
2290
- flo_int [eE]
2291
- => {
2292
- if version?(18, 19, 20)
2293
- diagnostic :error,
2294
- :trailing_in_number, { :character => tok(@te - 1, @te) },
2295
- range(@te - 1, @te)
2296
- else
2297
- emit(:tINTEGER, tok(@ts, @te - 1).to_i, @ts, @te - 1)
2298
- fhold; fbreak;
2299
- end
2300
- };
2301
-
2302
- flo_int flo_frac [eE]
2303
- => {
2304
- if version?(18, 19, 20)
2305
- diagnostic :error,
2306
- :trailing_in_number, { :character => tok(@te - 1, @te) },
2307
- range(@te - 1, @te)
2308
- else
2309
- emit(:tFLOAT, tok(@ts, @te - 1).to_f, @ts, @te - 1)
2310
- fhold; fbreak;
2311
- end
2312
- };
2313
-
2314
- flo_int
2315
- ( flo_frac? flo_pow %{ @num_suffix_s = p } flo_pow_suffix
2316
- | flo_frac %{ @num_suffix_s = p } flo_suffix
2317
- )
2318
- => {
2319
- digits = tok(@ts, @num_suffix_s)
2320
-
2321
- if version?(18, 19, 20)
2322
- emit(:tFLOAT, Float(digits), @ts, @num_suffix_s)
2323
- p = @num_suffix_s - 1
2324
- else
2325
- @num_xfrm.call(digits)
2326
- end
2327
- fbreak;
2328
- };
2329
-
2330
- #
2331
- # STRING AND XSTRING LITERALS
2332
- #
2333
-
2334
- # `echo foo`, "bar", 'baz'
2335
- '`' | ['"] # '
2336
- => {
2337
- type, delimiter = tok, tok[-1].chr
2338
- fgoto *push_literal(type, delimiter, @ts, nil, false, false, true);
2339
- };
2340
-
2341
- #
2342
- # CONSTANTS AND VARIABLES
2343
- #
2344
-
2345
- constant
2346
- => { emit(:tCONSTANT)
2347
- fnext *arg_or_cmdarg(cmd_state); fbreak; };
2348
-
2349
- constant ambiguous_const_suffix
2350
- => { emit(:tCONSTANT, tok(@ts, tm), @ts, tm)
2351
- p = tm - 1; fbreak; };
2352
-
2353
- global_var | class_var_v | instance_var_v
2354
- => { p = @ts - 1; fcall expr_variable; };
2355
-
2356
- #
2357
- # METHOD CALLS
2358
- #
2359
-
2360
- '.' | '&.' | '::'
2361
- => { emit_table(PUNCTUATION)
2362
- fnext expr_dot; fbreak; };
2363
-
2364
- call_or_var
2365
- => local_ident;
2366
-
2367
- bareword ambiguous_fid_suffix
2368
- => {
2369
- if tm == @te
2370
- # Suffix was consumed, e.g. foo!
2371
- emit(:tFID)
2372
- else
2373
- # Suffix was not consumed, e.g. foo!=
2374
- emit(:tIDENTIFIER, tok(@ts, tm), @ts, tm)
2375
- p = tm - 1
2376
- end
2377
- fnext expr_arg; fbreak;
2378
- };
2379
-
2380
- #
2381
- # OPERATORS
2382
- #
2383
-
2384
- '*' | '=>'
2385
- => {
2386
- emit_table(PUNCTUATION)
2387
- fgoto expr_value;
2388
- };
2389
-
2390
- # When '|', '~', '!', '=>' are used as operators
2391
- # they do not accept any symbols (or quoted labels) after.
2392
- # Other binary operators accept it.
2393
- ( operator_arithmetic | operator_rest ) - ( '|' | '~' | '!' | '*' )
2394
- => {
2395
- emit_table(PUNCTUATION);
2396
- fnext expr_value; fbreak;
2397
- };
2398
-
2399
- ( e_lparen | '|' | '~' | '!' )
2400
- => { emit_table(PUNCTUATION)
2401
- fnext expr_beg; fbreak; };
2402
-
2403
- e_rbrace | e_rparen | e_rbrack
2404
- => {
2405
- emit_table(PUNCTUATION)
2406
-
2407
- if @version < 24
2408
- @cond.lexpop
2409
- @cmdarg.lexpop
2410
- else
2411
- @cond.pop
2412
- @cmdarg.pop
2413
- end
2414
-
2415
- if tok == '}'.freeze || tok == ']'.freeze
2416
- if @version >= 25
2417
- fnext expr_end;
2418
- else
2419
- fnext expr_endarg;
2420
- end
2421
- else # )
2422
- # fnext expr_endfn; ?
2423
- end
2424
-
2425
- fbreak;
2426
- };
2427
-
2428
- operator_arithmetic '='
2429
- => { emit(:tOP_ASGN, tok(@ts, @te - 1))
2430
- fnext expr_beg; fbreak; };
2431
-
2432
- '?'
2433
- => { emit(:tEH, '?'.freeze)
2434
- fnext expr_value; fbreak; };
2435
-
2436
- e_lbrack
2437
- => { emit(:tLBRACK2, '['.freeze)
2438
- fnext expr_beg; fbreak; };
2439
-
2440
- '...' c_nl
2441
- => {
2442
- if @paren_nest == 0
2443
- diagnostic :warning, :triple_dot_at_eol, nil, range(@ts, @te - 1)
2444
- end
2445
-
2446
- emit(:tDOT3, '...'.freeze, @ts, @te - 1)
2447
- fhold;
2448
- fnext expr_beg; fbreak;
2449
- };
2450
-
2451
- punctuation_end
2452
- => { emit_table(PUNCTUATION)
2453
- fnext expr_beg; fbreak; };
2454
-
2455
- #
2456
- # WHITESPACE
2457
- #
2458
-
2459
- w_space_comment;
2460
-
2461
- w_newline
2462
- => { fgoto leading_dot; };
2463
-
2464
- ';'
2465
- => { emit(:tSEMI, ';'.freeze)
2466
- @command_start = true
2467
- fnext expr_value; fbreak; };
2468
-
2469
- '\\' c_line {
2470
- diagnostic :error, :bare_backslash, nil, range(@ts, @ts + 1)
2471
- fhold;
2472
- };
2473
-
2474
- c_any
2475
- => {
2476
- diagnostic :fatal, :unexpected, { :character => tok.inspect[1..-2] }
2477
- };
2478
-
2479
- c_eof => do_eof;
2480
- *|;
2481
-
2482
- leading_dot := |*
2483
- # Insane leading dots:
2484
- # a #comment
2485
- # # post-2.7 comment
2486
- # .b: a.b
2487
-
2488
- # Here we use '\n' instead of w_newline to not modify @newline_s
2489
- # and eventually properly emit tNL
2490
- (c_space* w_space_comment '\n')+
2491
- => {
2492
- if @version < 27
2493
- # Ruby before 2.7 doesn't support comments before leading dot.
2494
- # If a line after "a" starts with a comment then "a" is a self-contained statement.
2495
- # So in that case we emit a special tNL token and start reading the
2496
- # next line as a separate statement.
2497
- #
2498
- # Note: block comments before leading dot are not supported on any version of Ruby.
2499
- emit(:tNL, nil, @newline_s, @newline_s + 1)
2500
- fhold; fnext line_begin; fbreak;
2501
- end
2502
- };
2503
-
2504
- c_space* %{ tm = p } ('.' | '&.')
2505
- => { p = tm - 1; fgoto expr_end; };
2506
-
2507
- any
2508
- => { emit(:tNL, nil, @newline_s, @newline_s + 1)
2509
- fhold; fnext line_begin; fbreak; };
2510
- *|;
2511
-
2512
- #
2513
- # === EMBEDDED DOCUMENT (aka BLOCK COMMENT) PARSING ===
2514
- #
2515
-
2516
- line_comment := |*
2517
- '=end' c_line* c_nl_zlen
2518
- => {
2519
- emit_comment(@eq_begin_s, @te)
2520
- fgoto *@cs_before_block_comment;
2521
- };
2522
-
2523
- c_line* c_nl;
2524
-
2525
- c_line* zlen
2526
- => {
2527
- diagnostic :fatal, :embedded_document, nil,
2528
- range(@eq_begin_s, @eq_begin_s + '=begin'.length)
2529
- };
2530
- *|;
2531
-
2532
- line_begin := |*
2533
- w_any;
2534
-
2535
- '=begin' ( c_space | c_nl_zlen )
2536
- => { @eq_begin_s = @ts
2537
- fgoto line_comment; };
2538
-
2539
- '__END__' ( c_eol - zlen )
2540
- => { p = pe - 3 };
2541
-
2542
- c_any
2543
- => { cmd_state = true; fhold; fgoto expr_value; };
2544
-
2545
- c_eof => do_eof;
2546
- *|;
2547
-
2548
- }%%
2549
- # %
2550
- end