Lexer.x 78.5 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
--
-- GHC's lexer.
--
-- This is a combination of an Alex-generated lexer from a regex
-- definition, with some hand-coded bits.
--
-- Completely accurate information about token-spans within the source
-- file is maintained.  Every token has a start and end SrcLoc attached to it.
--
-----------------------------------------------------------------------------

--   ToDo / known bugs:
--    - parsing integers is a bit slow
--    - readRational is a bit slow
--
--   Known bugs, that were also in the previous version:
--    - M... should be 3 tokens, not 1.
--    - pragma-end should be only valid in a pragma

22
23
24
25
26
27
28
29
30
31
32
--   qualified operator NOTES.
--   
--   - If M.(+) is a single lexeme, then..
--     - Probably (+) should be a single lexeme too, for consistency.
--       Otherwise ( + ) would be a prefix operator, but M.( + ) would not be.
--     - But we have to rule out reserved operators, otherwise (..) becomes
--       a different lexeme.
--     - Should we therefore also rule out reserved operators in the qualified
--       form?  This is quite difficult to achieve.  We don't do it for
--       qualified varids.

33
{
34
35
36
37
38
39
40
41
42
-- XXX The above flags turn off warnings in the generated code:
{-# OPTIONS_GHC -fno-warn-unused-matches #-}
{-# OPTIONS_GHC -fno-warn-unused-binds #-}
{-# OPTIONS_GHC -fno-warn-unused-imports #-}
{-# OPTIONS_GHC -fno-warn-missing-signatures #-}
-- But alex still generates some code that causes the "lazy unlifted bindings"
-- warning, and old compilers don't know about it so we can't easily turn
-- it off, so for now we use the sledge hammer:
{-# OPTIONS_GHC -w #-}
43

44
45
{-# OPTIONS_GHC -funbox-strict-fields #-}

46
module Lexer (
47
   Token(..), lexer, pragState, mkPState, PState(..),
48
   P(..), ParseResult(..), getSrcLoc, 
49
   getPState, getDynFlags, withThisPackage,
50
   failLocMsgP, failSpanMsgP, srcParseFail,
51
   getMessages, 
52
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
53
   getLexState, popLexState, pushLexState,
54
   extension, standaloneDerivingEnabled, bangPatEnabled,
Jedai's avatar
Jedai committed
55
56
   addWarning,
   lexTokenStream
57
58
  ) where

59
60
import Bag
import ErrUtils
61
62
63
64
65
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
66
import DynFlags
67
import Module
68
import Ctype
69
import Util		( readRational )
70

71
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
72
import Data.Bits
73
import Data.Char
74
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
75
import Data.Maybe
76
77
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
78
import Data.Ratio
79
80
}

81
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
82
$whitechar   = [\ \n\r\f\v $unispace]
83
$white_no_nl = $whitechar # \n
84
$tab         = \t
85
86

$ascdigit  = 0-9
87
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
88
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
89
90
91
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
92
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
93
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
94
95
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

96
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
97
$asclarge  = [A-Z]
98
99
$large     = [$asclarge $unilarge]

100
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
101
$ascsmall  = [a-z]
102
103
$small     = [$ascsmall $unismall \_]

104
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
105
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
106
107

$octit	   = 0-7
108
$hexit     = [$decdigit A-F a-f]
109
110
111
112
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

113
114
$pragmachar = [$small $large $digit]

115
116
$docsym    = [\| \^ \* \$]

117
118
119
120
121
122
@varid     = $small $idchar*
@conid     = $large $idchar*

@varsym    = $symbol $symchar*
@consym    = \: $symchar*

123
@decimal     = $decdigit+
124
125
126
127
128
129
130
131
132
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

-- we support the hierarchical module name extension:
@qual = (@conid \.)+

@floating_point = @decimal \. @decimal @exponent? | @decimal @exponent

133
134
135
136
137
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

138
139
140
141
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
142
$tab+         { warn Opt_WarnTabs (text "Tab character") }
143
144
145
146
147

-- Everywhere: deal with nested comments.  We explicitly rule out
-- pragmas, "{-#", so that we don't accidentally treat them as comments.
-- (this can happen even though pragmas will normally take precedence due to
-- longest-match, because pragmas aren't valid in every state, but comments
148
149
150
151
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
152
153
154
155

-- Single-line comments are a bit tricky.  Haskell 98 says that two or
-- more dashes followed by a symbol should be parsed as a varsym, so we
-- have to exclude those.
156
157
158
159
160
161
162
163
164
165

-- Since Haddock comments aren't valid in every state, we need to rule them
-- out here.  

-- The following two rules match comments that begin with two dashes, but
-- continue with a different character. The rules test that this character
-- is not a symbol (in which case we'd have a varsym), and that it's not a
-- space followed by a Haddock comment symbol (docsym) (in which case we'd
-- have a Haddock comment). The rules then munch the rest of the line.

Jedai's avatar
Jedai committed
166
167
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
168
169
170

-- Next, match Haddock comments if no -haddock flag

Jedai's avatar
Jedai committed
171
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
172
173
174
175
176
177
178

-- Now, when we've matched comments that begin with 2 dashes and continue
-- with a different character, we need to match comments that begin with three
-- or more dashes (which clearly can't be Haddock comments). We only need to
-- make sure that the first non-dash character isn't a symbol, and munch the
-- rest of the line.

Jedai's avatar
Jedai committed
179
"---"\-* [^$symbol :] .* { lineCommentToken }
180
181
182
183

-- Since the previous rules all match dashes followed by at least one
-- character, we also need to match a whole line filled with just dashes.

Jedai's avatar
Jedai committed
184
"--"\-* / { atEOL } { lineCommentToken }
185
186
187
188

-- We need this rule since none of the other single line comment rules
-- actually match this case.

Jedai's avatar
Jedai committed
189
"-- " / { atEOL } { lineCommentToken }
190
191
192
193
194
195
196
197
198
199
200
201
202

-- 'bol' state: beginning of a line.  Slurp up all the whitespace (including
-- blank lines) until we find a non-whitespace character, then do layout
-- processing.
--
-- One slight wibble here: what if the line begins with {-#? In
-- theory, we have to lex the pragma to see if it's one we recognise,
-- and if it is, then we backtrack and do_bol, otherwise we treat it
-- as a nested comment.  We don't bother with this: if the line begins
-- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
<bol> {
  \n					;
  ^\# (line)?				{ begin line_prag1 }
203
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
204
  ^\# \! .* \n				; -- #!, for scripts
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
  ()					{ do_bol }
}

-- after a layout keyword (let, where, do, of), we begin a new layout
-- context if the curly brace is missing.
-- Careful! This stuff is quite delicate.
<layout, layout_do> {
  \{ / { notFollowedBy '-' }		{ pop_and open_brace }
	-- we might encounter {-# here, but {- has been handled already
  \n					;
  ^\# (line)?				{ begin line_prag1 }
}

-- do is treated in a subtly different way, see new_layout_context
<layout>    ()				{ new_layout_context True }
<layout_do> ()				{ new_layout_context False }

-- after a new layout context which was found to be to the left of the
-- previous context, we have generated a '{' token, and we now need to
-- generate a matching '}' token.
<layout_left>  ()			{ do_layout_left }

227
<0,option_prags> \n				{ begin bol }
228

229
230
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
231
232
233

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
234
<line_prag1> $decdigit+			{ setLine line_prag1a }
235
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
236
237
238
239
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
240
<line_prag2> $decdigit+			{ setLine line_prag2a }
241
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
242
243
244
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
245

246
<0,option_prags> {
247
248
249
250
251
252
  "{-#" $whitechar* $pragmachar+ 
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
253

254
  -- We ignore all these pragmas, but don't generate a warning for them
255
256
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
257
258

  -- ToDo: should only be valid inside a pragma:
259
  "#-}" 				{ endPrag }
260
261
}

262
<option_prags> {
263
264
265
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

David Waern's avatar
David Waern committed
266
267
268
  "-- #"                                 { multiline_doc_comment }
}

269
270
<0> {
  -- In the "0" mode we ignore these pragmas
271
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
272
273
274
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
275
<0> {
Jedai's avatar
Jedai committed
276
  "-- #" .* { lineCommentToken }
277
}
278

279
<0,option_prags> {
280
281
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
282
283
}

284
285
-- '0' state: ordinary lexemes

286
287
-- Haddock comments

288
<0> {
David Waern's avatar
David Waern committed
289
290
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
291
292
}

293
294
-- "special" symbols

295
<0> {
296
297
298
299
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
300
<0> {
301
302
303
304
305
306
307
308
  "[|"	    / { ifExtension thEnabled }	{ token ITopenExpQuote }
  "[e|"	    / { ifExtension thEnabled }	{ token ITopenExpQuote }
  "[p|"	    / { ifExtension thEnabled }	{ token ITopenPatQuote }
  "[d|"	    / { ifExtension thEnabled }	{ layout_token ITopenDecQuote }
  "[t|"	    / { ifExtension thEnabled }	{ token ITopenTypQuote }
  "|]"	    / { ifExtension thEnabled }	{ token ITcloseQuote }
  \$ @varid / { ifExtension thEnabled }	{ skip_one_varid ITidEscape }
  "$("	    / { ifExtension thEnabled }	{ token ITparenEscape }
309

310
  "[" @varid "|"  / { ifExtension qqEnabled }
311
                     { lex_quasiquote_tok }
312
313
}

314
<0> {
ross's avatar
ross committed
315
316
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
317
318
319
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

320
<0> {
321
322
323
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

324
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
325
326
327
328
329
330
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

331
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
332
333
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
334
335
}

336
<0,option_prags> {
337
338
339
340
341
342
343
344
345
346
347
348
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

349
<0,option_prags> {
350
  @qual @varid			{ idtoken qvarid }
351
352
353
354
355
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

356
<0> {
357
358
359
360
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
361
362
}

363
364
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
365
<0> {
366
367
368
369
370
371
  @qual @varsym       / { ifExtension oldQualOps } { idtoken qvarsym }
  @qual @consym       / { ifExtension oldQualOps } { idtoken qconsym }
  @qual \( @varsym \) / { ifExtension newQualOps } { idtoken prefixqvarsym }
  @qual \( @consym \) / { ifExtension newQualOps } { idtoken prefixqconsym }
  @varsym                                          { varsym }
  @consym                                          { consym }
372
373
}

374
375
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
376
<0> {
377
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
378
379
380
  @decimal           { tok_num positive 0 0 decimal }
  0[oO] @octal       { tok_num positive 2 2 octal }
  0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal }
381
382

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
383
  @floating_point    { strtoken tok_float }
384
385
}

386
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
387
  -- Unboxed ints (:: Int#) and words (:: Word#)
388
389
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
390
391
392
393
394
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
  0[oO] @octal                 \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
  0[xX] @hexadecimal           \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
  @negative @decimal           \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
395
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
396

Ian Lynagh's avatar
Ian Lynagh committed
397
398
399
400
  @decimal                     \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
  0[oO] @octal                 \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
  0[xX] @hexadecimal           \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }

401
402
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
403
404
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
405
406
407
408
409
410
}

-- Strings and chars are lexed by hand-written code.  The reason is
-- that even if we recognise the string or char here in the regex
-- lexer, we would still have to parse the string afterward in order
-- to convert it to a String.
411
<0> {
412
413
414
415
416
417
418
419
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

{
-- -----------------------------------------------------------------------------
-- The token type

420
data Token
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
  = ITas  			-- Haskell keywords
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
  | ITdo
  | ITelse
  | IThiding
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere
  | ITscc			-- ToDo: remove (we use {-# SCC "..." #-} now)

  | ITforall			-- GHC extension keywords
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
  | ITthreadsafe
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
457
  | ITprimcallconv
458
  | ITmdo
459
  | ITfamily
460
461
462
  | ITgroup
  | ITby
  | ITusing
463

464
465
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
466
  | ITinline_conlike_prag Bool  -- same
467
468
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
469
470
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
471
  | ITwarning_prag
472
473
474
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
475
  | ITgenerated_prag
476
  | ITcore_prag                 -- hdaume: core annotations
477
  | ITunpack_prag
478
  | ITann_prag
479
  | ITclose_prag
480
481
482
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508

  | ITdotdot  			-- reserved symbols
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

  | ITbiglam			-- GHC-extension symbols

  | ITocurly  			-- special symbols
  | ITccurly
  | ITocurlybar                 -- {|, for type applications
  | ITccurlybar                 -- |}, for type applications
  | ITvocurly
  | ITvccurly
  | ITobrack
509
510
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote

  | ITvarid   FastString	-- identifiers
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)
529
530
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
531
532
533
534
535
536
537
538
539
540
541

  | ITdupipvarid   FastString	-- GHC extension: implicit param: ?x

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
  | ITrational   Rational

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
542
  | ITprimword   Integer
543
544
545
  | ITprimfloat  Rational
  | ITprimdouble Rational

546
  -- Template Haskell extension tokens
547
548
549
550
551
552
553
554
555
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
556
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
557
558
559
560

  -- Arrow notation extension
  | ITproc
  | ITrec
561
562
563
564
565
566
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
567
568
569

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
570
571
572
573
574
575
576

  -- Documentation annotations
  | ITdocCommentNext  String     -- something beginning '-- |'
  | ITdocCommentPrev  String     -- something beginning '-- ^'
  | ITdocCommentNamed String     -- something beginning '-- $'
  | ITdocSection      Int String -- a section heading
  | ITdocOptions      String     -- doc options (prune, ignore-exports, etc)
David Waern's avatar
David Waern committed
577
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
578
579
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
580

581
582
583
584
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
585
{-
586
isSpecial :: Token -> Bool
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
-- If we see M.x, where x is a keyword, but
-- is special, we treat is as just plain M.x, 
-- not as a keyword.
isSpecial ITas        	= True
isSpecial IThiding    	= True
isSpecial ITqualified 	= True
isSpecial ITforall    	= True
isSpecial ITexport    	= True
isSpecial ITlabel     	= True
isSpecial ITdynamic   	= True
isSpecial ITsafe    	= True
isSpecial ITthreadsafe 	= True
isSpecial ITunsafe    	= True
isSpecial ITccallconv   = True
isSpecial ITstdcallconv = True
602
isSpecial ITprimcallconv = True
603
isSpecial ITmdo		= True
604
isSpecial ITfamily	= True
605
606
607
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
608
isSpecial _             = False
twanvl's avatar
twanvl committed
609
-}
610
611
612
613
614
615
616
617

-- the bitmap provided as the third component indicates whether the
-- corresponding extension keyword is valid under the extension options
-- provided to the compiler; if the extension corresponding to *any* of the
-- bits set in the bitmap is enabled, the keyword is valid (this setup
-- facilitates using a keyword in two different extensions that can be
-- activated independently)
--
Ian Lynagh's avatar
Ian Lynagh committed
618
reservedWordsFM :: UniqFM (Token, Int)
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
reservedWordsFM = listToUFM $
	map (\(x, y, z) -> (mkFastString x, (y, z)))
       [( "_",		ITunderscore, 	0 ),
	( "as",		ITas, 		0 ),
	( "case",	ITcase, 	0 ),     
	( "class",	ITclass, 	0 ),    
	( "data",	ITdata, 	0 ),     
	( "default",	ITdefault, 	0 ),  
	( "deriving",	ITderiving, 	0 ), 
	( "do",		ITdo, 		0 ),       
	( "else",	ITelse, 	0 ),     
	( "hiding",	IThiding, 	0 ),
	( "if",		ITif, 		0 ),       
	( "import",	ITimport, 	0 ),   
	( "in",		ITin, 		0 ),       
	( "infix",	ITinfix, 	0 ),    
	( "infixl",	ITinfixl, 	0 ),   
	( "infixr",	ITinfixr, 	0 ),   
	( "instance",	ITinstance, 	0 ), 
	( "let",	ITlet, 		0 ),      
	( "module",	ITmodule, 	0 ),   
	( "newtype",	ITnewtype, 	0 ),  
	( "of",		ITof, 		0 ),       
	( "qualified",	ITqualified, 	0 ),
	( "then",	ITthen, 	0 ),     
	( "type",	ITtype, 	0 ),     
	( "where",	ITwhere, 	0 ),
	( "_scc_",	ITscc, 		0 ),		-- ToDo: remove

648
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
649
	( "mdo",	ITmdo,		 bit recursiveDoBit),
650
	( "family",	ITfamily,	 bit tyFamBit),
651
652
653
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
654
655
656
657
658
659

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
660
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),  -- ToDo: remove
661
662
663
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
664
	( "prim",       ITprimcallconv,	 bit ffiBit),
665

666
	( "rec",	ITrec,		 bit recBit),
667
	( "proc",	ITproc,		 bit arrowsBit)
668
669
     ]

Ian Lynagh's avatar
Ian Lynagh committed
670
reservedSymsFM :: UniqFM (Token, Int -> Bool)
671
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
    map (\ (x,y,z) -> (mkFastString x,(y,z)))
      [ ("..",  ITdotdot,   always)
        -- (:) is a reserved op, meaning only list cons
       ,(":",   ITcolon,    always)
       ,("::",  ITdcolon,   always)
       ,("=",   ITequal,    always)
       ,("\\",  ITlam,      always)
       ,("|",   ITvbar,     always)
       ,("<-",  ITlarrow,   always)
       ,("->",  ITrarrow,   always)
       ,("@",   ITat,       always)
       ,("~",   ITtilde,    always)
       ,("=>",  ITdarrow,   always)
       ,("-",   ITminus,    always)
       ,("!",   ITbang,     always)

        -- For data T (a::*) = MkT
689
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
690
        -- For 'forall a . t'
691
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
692
693
694
695
696

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
697

Ian Lynagh's avatar
Ian Lynagh committed
698
699
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
700
701
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
702
703
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
704
705
706
707
708
709
710
711

       ,("⤙",   ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
       ,("⤚",   ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
       ,("⤛",   ITLarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
       ,("⤜",   ITRarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)

       ,("★", ITstar, unicodeSyntaxEnabled)

Simon Marlow's avatar
Simon Marlow committed
712
713
714
        -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
        -- form part of a large operator.  This would let us have a better
        -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
715
716
717
718
719
       ]

-- -----------------------------------------------------------------------------
-- Lexer actions

720
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
721

722
special :: Token -> Action
twanvl's avatar
twanvl committed
723
special tok span _buf _len = return (L span tok)
724

725
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
726
727
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
728

729
730
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
731

732
733
734
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
735

736
737
738
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
739

740
init_strtoken :: Int -> (String -> Token) -> Action
741
-- like strtoken, but drops the last N character(s)
742
743
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
744
745

begin :: Int -> Action
746
begin code _span _str _len = do pushLexState code; lexToken
747
748

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
749
750
pop _span _buf _len = do _ <- popLexState
                         lexToken
751
752

pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
753
754
pop_and act span buf len = do _ <- popLexState
                              act span buf len
755

756
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
757
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
758
759
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
760
notFollowedBy :: Char -> AlexAccPred Int
761
notFollowedBy char _ _ _ (AI _ buf) 
762
  = nextCharIs buf (/=char)
763

Ian Lynagh's avatar
Ian Lynagh committed
764
notFollowedBySymbol :: AlexAccPred Int
765
notFollowedBySymbol _ _ _ (AI _ buf)
766
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
767

768
769
770
771
772
-- We must reject doc comments as being ordinary comments everywhere.
-- In some cases the doc comment will be selected as the lexeme due to
-- maximal munch, but not always, because the nested comment rule is
-- valid in all states, but the doc-comment rules are only valid in
-- the non-layout states.
Ian Lynagh's avatar
Ian Lynagh committed
773
isNormalComment :: AlexAccPred Int
774
isNormalComment bits _ _ (AI _ buf)
775
776
777
778
779
780
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
781
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
782
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
783

twanvl's avatar
twanvl committed
784
{-
785
haddockDisabledAnd p bits _ _ (AI _ buf)
786
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
787
-}
788

Ian Lynagh's avatar
Ian Lynagh committed
789
atEOL :: AlexAccPred Int
790
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
791

Ian Lynagh's avatar
Ian Lynagh committed
792
ifExtension :: (Int -> Bool) -> AlexAccPred Int
793
794
ifExtension pred bits _ _ _ = pred bits

795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
    worker commentAcc input docType oneLine = case alexGetChar input of
      Just ('\n', input') 
        | oneLine -> docCommentEnd input commentAcc docType buf span
        | otherwise -> case checkIfCommentLine input' of
          Just input -> worker ('\n':commentAcc) input docType False
          Nothing -> docCommentEnd input commentAcc docType buf span
      Just (c, input) -> worker (c:commentAcc) input docType oneLine
      Nothing -> docCommentEnd input commentAcc docType buf span
      
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
        check input = case alexGetChar input of
          Just ('-', input) -> case alexGetChar input of
            Just ('-', input) -> case alexGetChar input of
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

        dropNonNewlineSpace input = case alexGetChar input of
          Just (c, input') 
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
823
824
825
826
827
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

828
829
830
831
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
832
833
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
834
  input <- getInput
Jedai's avatar
Jedai committed
835
  go "" (1::Int) input
836
  where
Jedai's avatar
Jedai committed
837
838
839
840
841
842
    go commentAcc 0 input = do setInput input
                               b <- extension rawTokenStreamEnabled
                               if b
                                 then docCommentEnd input commentAcc ITblockComment _str span
                                 else cont
    go commentAcc n input = case alexGetChar input of
843
844
845
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
846
847
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
848
849
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
850
851
852
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
853
854
855
856
857
858
859
860

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
    go commentAcc input docType _ = case alexGetChar input of
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
861
        Just ('\125',input) ->
862
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
863
        Just (_,_) -> go ('-':commentAcc) input docType False
864
865
866
867
868
869
      Just ('\123', input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> do
          setInput input
          let cont = do input <- getInput; go commentAcc input docType False
          nested_comment cont span buf _len
twanvl's avatar
twanvl committed
870
        Just (_,_) -> go ('\123':commentAcc) input docType False
871
872
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
873
874
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
                 -> P (Located Token)
875
withLexedDocType lexDocComment = do
876
  input@(AI _ buf) <- getInput
877
878
879
880
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
881
882
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
883
    _ -> panic "withLexedDocType: Bad doc type"
884
885
886
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
887
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
888
889
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

890
891
892
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
893
rulePrag span _buf _len = do
894
  setExts (.|. bit inRulePragBit)
895
896
897
  return (L span ITrules_prag)

endPrag :: Action
898
endPrag span _buf _len = do
899
900
901
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

902
903
904
905
906
907
908
909
910
911
912
913
-- docCommentEnd
-------------------------------------------------------------------------------
-- This function is quite tricky. We can't just return a new token, we also
-- need to update the state of the parser. Why? Because the token is longer
-- than what was lexed by Alex, and the lexToken function doesn't know this, so 
-- it writes the wrong token length to the parser state. This function is
-- called afterwards, so it can just update the state. 

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
                 SrcSpan -> P (Located Token) 
docCommentEnd input commentAcc docType buf span = do
  setInput input
914
  let (AI loc nextBuf) = input
915
916
917
918
      comment = reverse commentAcc
      span' = mkSrcSpan (srcSpanStart span) loc
      last_len = byteDiff buf nextBuf
      
919
  span `seq` setLastToken span' last_len
920
921
  return (L span' (docType comment))
 
Ian Lynagh's avatar
Ian Lynagh committed
922
errBrace :: AlexInput -> SrcSpan -> P a
923
errBrace (AI end _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
924

925
open_brace, close_brace :: Action
926
open_brace span _str _len = do 
927
928
  ctx <- getContext
  setContext (NoLayout:ctx)
929
930
  return (L span ITocurly)
close_brace span _str _len = do 
931
  popContext
932
  return (L span ITccurly)
933

Ian Lynagh's avatar
Ian Lynagh committed
934
qvarid, qconid :: StringBuffer -> Int -> Token
935
936
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
937

938
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
939
940
941
-- takes a StringBuffer and a length, and returns the module name
-- and identifier parts of a qualified name.  Splits at the *last* dot,
-- because of hierarchical module names.
942
splitQualName orig_buf len parens = split orig_buf orig_buf
943
  where
944
945
946
947
948
949
    split buf dot_buf
	| orig_buf `byteDiff` buf >= len  = done dot_buf
	| c == '.'                	  = found_dot buf'
	| otherwise               	  = split buf' dot_buf
      where
       (c,buf') = nextChar buf
950
951
952
953
  
    -- careful, we might get names like M....
    -- so, if the character after the dot is not upper-case, this is
    -- the end of the qualifier part.
954
955
956
957
958
959
960
961
    found_dot buf -- buf points after the '.'
	| isUpper c    = split buf' buf
	| otherwise    = done buf
      where
       (c,buf') = nextChar buf

    done dot_buf =
	(lexemeToFastString orig_buf (qual_size - 1),
962
963
964
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
965
966
      where
	qual_size = orig_buf `byteDiff` dot_buf
967

Ian Lynagh's avatar
Ian Lynagh committed
968
969
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
970
  fs `seq`
971
972
973
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
974
		return (L span keyword)
975
976
977
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
978
979
980
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
981
982
983
  where
	fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
984
conid :: StringBuffer -> Int -> Token
985
986
987
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
988
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
989
990
991
992
qvarsym buf len = ITqvarsym $! splitQualName buf len False
qconsym buf len = ITqconsym $! splitQualName buf len False
prefixqvarsym buf len = ITprefixqvarsym $! splitQualName buf len True
prefixqconsym buf len = ITprefixqconsym $! splitQualName buf len True
993

Ian Lynagh's avatar
Ian Lynagh committed
994
varsym, consym :: Action
995
996
997
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
998
999
sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
    -> P (Located Token)
1000
sym con span buf len = 
For faster browsing, not all history is shown. View entire blame