Lexer.x 64.5 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
{-# OPTIONS -Wwarn -w #-}
35
36
37
-- The above warning supression flag is a temporary kludge.
-- While working on this module you are encouraged to remove it and fix
-- any warnings in the module. See
Ian Lynagh's avatar
Ian Lynagh committed
38
--     http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
39
-- for details
twanvl's avatar
twanvl committed
40
41
42
--
-- Note that Alex itself generates code with with some unused bindings and
-- without type signatures, so removing the flag might not be possible.
43

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

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

58
59
import Bag
import ErrUtils
60
61
62
63
64
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
65
import DynFlags
66
import Ctype
67
import Util		( maybePrefixMatch, readRational )
68

69
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
70
import Data.Bits
71
import Data.Char
Simon Marlow's avatar
Simon Marlow committed
72
import Data.Ratio
73
74
}

75
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
76
$whitechar   = [\ \n\r\f\v $unispace]
77
$white_no_nl = $whitechar # \n
78
$tab         = \t
79
80

$ascdigit  = 0-9
81
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
82
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
83
84
85
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
86
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
87
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
88
89
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

90
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
91
$asclarge  = [A-Z]
92
93
$large     = [$asclarge $unilarge]

94
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
95
$ascsmall  = [a-z]
96
97
$small     = [$ascsmall $unismall \_]

98
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
99
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
100
101

$octit	   = 0-7
102
$hexit     = [$decdigit A-F a-f]
103
104
105
106
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

107
108
$docsym    = [\| \^ \* \$]

109
110
111
112
113
114
@varid     = $small $idchar*
@conid     = $large $idchar*

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

115
@decimal     = $decdigit+
116
117
118
119
120
121
122
123
124
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

125
126
127
128
129
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

130
131
132
133
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
134
$tab+         { warn Opt_WarnTabs (text "Tab character") }
135
136
137
138
139

-- 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
140
141
142
143
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
144
145
146
147

-- 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.
148
149
150
151
152
153
154
155
156
157

-- 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
158
159
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
160
161
162

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

Jedai's avatar
Jedai committed
163
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
164
165
166
167
168
169
170

-- 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
171
"---"\-* [^$symbol :] .* { lineCommentToken }
172
173
174
175

-- 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
176
"--"\-* / { atEOL } { lineCommentToken }
177
178
179
180

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

Jedai's avatar
Jedai committed
181
"-- " / { atEOL } { lineCommentToken }
182
183
184
185
186
187
188
189
190
191
192
193
194

-- '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 }
195
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
196
  ^\# \! .* \n				; -- #!, for scripts
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
  ()					{ 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 }

219
<0,option_prags> \n				{ begin bol }
220

221
222
"{-#" $whitechar* (line|LINE) / { notFollowedByPragmaChar }
                            { begin line_prag2 }
223
224
225

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
226
<line_prag1> $decdigit+			{ setLine line_prag1a }
227
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
228
229
230
231
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
232
<line_prag2> $decdigit+			{ setLine line_prag2a }
233
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
234
235
236
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
237

238
<0,option_prags> {
239
240
241
242
  "{-#" $whitechar* (RULES|rules)  / { notFollowedByPragmaChar } { rulePrag }
  "{-#" $whitechar* (INLINE|inline)	 / { notFollowedByPragmaChar }
                    { token (ITinline_prag True) }
  "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline) / { notFollowedByPragmaChar }
243
  					{ token (ITinline_prag False) }
244
245
246
247
248
249
  "{-#" $whitechar* (INLINE|inline)
        $whitechar+ (CONLIKE|conlike) / { notFollowedByPragmaChar }
                                        { token (ITinline_conlike_prag True) }
  "{-#" $whitechar* (NO(T)?INLINE|no(t?)inline)
        $whitechar+ (CONLIKE|constructorlike) / { notFollowedByPragmaChar }
                                        { token (ITinline_conlike_prag False) }
250
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) / { notFollowedByPragmaChar }
251
  					{ token ITspec_prag }
252
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
253
254
	$whitechar+ (INLINE|inline) / { notFollowedByPragmaChar }
                    { token (ITspec_inline_prag True) }
255
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
256
	$whitechar+ (NO(T?)INLINE|no(t?)inline) / { notFollowedByPragmaChar }
257
					{ token (ITspec_inline_prag False) }
258
259
260
  "{-#" $whitechar* (SOURCE|source) / { notFollowedByPragmaChar }
                    { token ITsource_prag }
  "{-#" $whitechar* (WARNING|warning) / { notFollowedByPragmaChar }
Ian Lynagh's avatar
Ian Lynagh committed
261
  					{ token ITwarning_prag }
262
  "{-#" $whitechar* (DEPRECATED|deprecated) / { notFollowedByPragmaChar }
263
  					{ token ITdeprecated_prag }
264
265
266
  "{-#" $whitechar* (SCC|scc)  / { notFollowedByPragmaChar }
                    { token ITscc_prag }
  "{-#" $whitechar* (GENERATED|generated) / { notFollowedByPragmaChar }
andy@galois.com's avatar
andy@galois.com committed
267
  					{ token ITgenerated_prag }
268
269
270
271
272
273
  "{-#" $whitechar* (CORE|core) / { notFollowedByPragmaChar }
                    { token ITcore_prag }
  "{-#" $whitechar* (UNPACK|unpack) / { notFollowedByPragmaChar }
                    { token ITunpack_prag }
  "{-#" $whitechar* (ANN|ann) / { notFollowedByPragmaChar }
                    { token ITann_prag }
274

275
276
  -- We ignore all these pragmas, but don't generate a warning for them
  -- CFILES is a hugs-only thing.
277
  "{-#" $whitechar* (OPTIONS_(HUGS|hugs|NHC98|nhc98|JHC|jhc|YHC|yhc|CATCH|catch|DERIVE|derive)|CFILES|cfiles|CONTRACT|contract) / { notFollowedByPragmaChar }
278
                    { nested_comment lexToken }
279
280

  -- ToDo: should only be valid inside a pragma:
281
  "#-}" 				{ endPrag }
282
283
}

284
<option_prags> {
285
286
287
  "{-#"  $whitechar* (OPTIONS|options) / { notFollowedByPragmaChar }
                                        { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc) / { notFollowedByPragmaChar }
288
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
289
  "{-#"  $whitechar* (OPTIONS_HADDOCK|options_haddock)
290
                   / { notFollowedByPragmaChar }
David Waern's avatar
David Waern committed
291
292
                                         { lex_string_prag ITdocOptions }
  "-- #"                                 { multiline_doc_comment }
293
294
295
296
  "{-#"  $whitechar* (LANGUAGE|language) / { notFollowedByPragmaChar }
                                         { token ITlanguage_prag }
  "{-#"  $whitechar* (INCLUDE|include) / { notFollowedByPragmaChar }
                                         { lex_string_prag ITinclude_prag }
David Waern's avatar
David Waern committed
297
298
}

299
300
<0> {
  -- In the "0" mode we ignore these pragmas
301
  "{-#"  $whitechar* (OPTIONS|options|OPTIONS_GHC|options_ghc|OPTIONS_HADDOCK|options_haddock|LANGUAGE|language|INCLUDE|include) / { notFollowedByPragmaChar }
302
303
304
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
305
<0> {
Jedai's avatar
Jedai committed
306
  "-- #" .* { lineCommentToken }
307
}
308

309
<0,option_prags> {
310
311
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
312
313
}

314
315
-- '0' state: ordinary lexemes

316
317
-- Haddock comments

318
<0> {
David Waern's avatar
David Waern committed
319
320
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
321
322
}

323
324
-- "special" symbols

325
<0> {
326
327
328
329
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
330
<0> {
331
332
333
334
335
336
337
338
  "[|"	    / { 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 }
339
340
341

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
342
343
}

344
<0> {
ross's avatar
ross committed
345
346
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
347
348
349
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

350
<0> {
351
352
353
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

354
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
355
356
357
358
359
360
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

361
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
362
363
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
364
365
}

366
<0,option_prags> {
367
368
369
370
371
372
373
374
375
376
377
378
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

379
<0,option_prags> {
380
  @qual @varid			{ idtoken qvarid }
381
382
383
384
385
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

386
<0> {
387
388
389
390
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
391
392
}

393
394
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
395
<0> {
396
397
398
399
400
401
  @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 }
402
403
}

404
405
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
406
<0> {
407
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
408
409
410
  @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 }
411
412

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
413
  @floating_point    { strtoken tok_float }
414
415
}

416
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
417
  -- Unboxed ints (:: Int#) and words (:: Word#)
418
419
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
420
421
422
423
424
  @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 }
425
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
426

Ian Lynagh's avatar
Ian Lynagh committed
427
428
429
430
  @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 }

431
432
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
433
434
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
435
436
437
438
439
440
}

-- 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.
441
<0> {
442
443
444
445
446
447
448
449
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

450
data Token
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
  = 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
  | ITdotnet
  | ITmdo
489
  | ITfamily
490
491
492
  | ITgroup
  | ITby
  | ITusing
493

494
495
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
496
  | ITinline_conlike_prag Bool  -- same
497
498
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
499
500
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
501
  | ITwarning_prag
502
503
504
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
505
  | ITgenerated_prag
506
  | ITcore_prag                 -- hdaume: core annotations
507
  | ITunpack_prag
508
  | ITann_prag
509
  | ITclose_prag
510
511
512
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538

  | 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
539
540
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
  | 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)
559
560
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
561
562
563
564
565
566
567
568
569
570
571
572
573

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

  | ITpragma StringBuffer

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

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
574
  | ITprimword   Integer
575
576
577
578
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
579
580
581
582
583
584
585
586
587
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
588
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
589
590
591
592

  -- Arrow notation extension
  | ITproc
  | ITrec
593
594
595
596
597
598
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
599
600
601

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
602
603
604
605
606
607
608

  -- 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
609
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
610
611
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
612

613
614
615
616
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
617
{-
618
isSpecial :: Token -> Bool
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
-- 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
isSpecial ITmdo		= True
635
isSpecial ITfamily	= True
636
637
638
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
639
isSpecial _             = False
twanvl's avatar
twanvl committed
640
-}
641
642
643
644
645
646
647
648

-- 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
649
reservedWordsFM :: UniqFM (Token, Int)
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
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

679
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
680
	( "mdo",	ITmdo,		 bit recursiveDoBit),
681
	( "family",	ITfamily,	 bit tyFamBit),
682
683
684
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
685
686
687
688
689
690
691
692
693
694
695
696
697

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
	( "dotnet",     ITdotnet,	 bit ffiBit),

	( "rec",	ITrec,		 bit arrowsBit),
698
	( "proc",	ITproc,		 bit arrowsBit)
699
700
     ]

Ian Lynagh's avatar
Ian Lynagh committed
701
reservedSymsFM :: UniqFM (Token, Int -> Bool)
702
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
    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
720
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
721
        -- For 'forall a . t'
722
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
723
724
725
726
727

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
728

Ian Lynagh's avatar
Ian Lynagh committed
729
730
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
731
732
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
733
734
735
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
736
737
738
        -- 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).
739
740
741
742
743
       ]

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

744
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
745

746
special :: Token -> Action
twanvl's avatar
twanvl committed
747
special tok span _buf _len = return (L span tok)
748

749
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
750
751
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
752

753
754
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
755

756
757
758
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
759

760
761
762
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
763

764
init_strtoken :: Int -> (String -> Token) -> Action
765
-- like strtoken, but drops the last N character(s)
766
767
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
768
769

begin :: Int -> Action
770
begin code _span _str _len = do pushLexState code; lexToken
771
772

pop :: Action
773
pop _span _buf _len = do popLexState; lexToken
774
775

pop_and :: Action -> Action
776
pop_and act span buf len = do popLexState; act span buf len
777

778
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
779
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
780
781
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
782
notFollowedBy :: Char -> AlexAccPred Int
783
784
notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
785

Ian Lynagh's avatar
Ian Lynagh committed
786
notFollowedBySymbol :: AlexAccPred Int
787
notFollowedBySymbol _ _ _ (AI _ _ buf)
788
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
789

Ian Lynagh's avatar
Ian Lynagh committed
790
notFollowedByPragmaChar :: AlexAccPred Int
791
792
793
notFollowedByPragmaChar _ _ _ (AI _ _ buf)
  = nextCharIs buf (\c -> not (isAlphaNum c || c == '_'))

794
795
796
797
798
-- 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
799
isNormalComment :: AlexAccPred Int
800
isNormalComment bits _ _ (AI _ _ buf)
801
802
803
804
805
806
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
807
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
808
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
809

twanvl's avatar
twanvl committed
810
{-
811
812
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
813
-}
814

Ian Lynagh's avatar
Ian Lynagh committed
815
atEOL :: AlexAccPred Int
816
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
817

Ian Lynagh's avatar
Ian Lynagh committed
818
ifExtension :: (Int -> Bool) -> AlexAccPred Int
819
820
ifExtension pred bits _ _ _ = pred bits

821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
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
849
850
851
852
853
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

854
855
856
857
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
858
859
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
860
  input <- getInput
Jedai's avatar
Jedai committed
861
  go "" (1::Int) input
862
  where
Jedai's avatar
Jedai committed
863
864
865
866
867
868
    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
869
870
871
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
872
873
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
874
875
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
876
877
878
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
879
880
881
882
883
884
885
886

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
887
        Just ('\125',input) ->
888
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
889
        Just (_,_) -> go ('-':commentAcc) input docType False
890
891
892
893
894
895
      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
896
        Just (_,_) -> go ('\123':commentAcc) input docType False
897
898
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
899
900
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
                 -> P (Located Token)
901
withLexedDocType lexDocComment = do
902
903
904
905
906
  input@(AI _ _ buf) <- getInput
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
907
908
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
909
    _ -> panic "withLexedDocType: Bad doc type"
910
911
912
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
913
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
914
915
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

916
917
918
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
Ian Lynagh's avatar
Ian Lynagh committed
919
rulePrag span _ _ = do
920
  setExts (.|. bit inRulePragBit)
921
922
923
  return (L span ITrules_prag)

endPrag :: Action
Ian Lynagh's avatar
Ian Lynagh committed
924
endPrag span _ _ = do
925
926
927
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
-- 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. 

-- This is complicated by the fact that Haddock tokens can span multiple lines, 
-- which is something that the original lexer didn't account for. 
-- I have added last_line_len in the parser state which represents the length 
-- of the part of the token that is on the last line. It is now used for layout 
-- calculation in pushCurrentContext instead of last_len. last_len is, like it 
-- was before, the full length of the token, and it is now only used for error
-- messages. /Waern 

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
                 SrcSpan -> P (Located Token) 
docCommentEnd input commentAcc docType buf span = do
  setInput input
  let (AI loc last_offs nextBuf) = input
      comment = reverse commentAcc
      span' = mkSrcSpan (srcSpanStart span) loc
      last_len = byteDiff buf nextBuf
      
      last_line_len = if (last_offs - last_len < 0) 
        then last_offs
        else last_len  
  
  span `seq` setLastToken span' last_len last_line_len
  return (L span' (docType comment))
 
Ian Lynagh's avatar
Ian Lynagh committed
960
errBrace :: AlexInput -> SrcSpan -> P a
961
errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
962

963
open_brace, close_brace :: Action
964
open_brace span _str _len = do 
965
966
  ctx <- getContext
  setContext (NoLayout:ctx)
967
968
  return (L span ITocurly)
close_brace span _str _len = do 
969
  popContext
970
  return (L span ITccurly)
971

Ian Lynagh's avatar
Ian Lynagh committed
972
qvarid, qconid :: StringBuffer -> Int -> Token
973
974
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
975

976
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
977
978
979
-- 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.
980
splitQualName orig_buf len parens = split orig_buf orig_buf
981
  where
982
983
984
985
986
987
    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
988
989
990
991
  
    -- 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.
992
993
994
995
996
997
998
999
    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),
1000
1001
1002
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
1003
1004
      where
	qual_size = orig_buf `byteDiff` dot_buf
1005

Ian Lynagh's avatar
Ian Lynagh committed
1006
1007
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
1008
  fs `seq`
1009
1010
1011
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
1012
		return (L span keyword)
1013
1014
1015
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
1016
1017
1018
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
1019
1020
1021
  where
	fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
1022
conid :: StringBuffer -> Int -> Token
1023
1024
1025
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
1026
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
1027
1028
1029
1030
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
1031

Ian Lynagh's avatar
Ian Lynagh committed
1032
varsym, consym :: Action
1033
1034
1035
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
1036
1037
sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
    -> P (Located Token)
1038
sym con span buf len = 
1039
1040
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
1041
		b <- extension exts
1042
1043
1044
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
1045
1046
1047
  where
	fs = lexemeToFastString buf len

1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
-- Variations on the integral numeric literal.
tok_integral :: (Integer -> Token)
     -> (Integer -> Integer)
 --    -> (StringBuffer -> StringBuffer) -> (Int -> Int)
     -> Int -> Int
     -> (Integer, (Char->Int)) -> Action
tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
  return $ L span $ itint $! transint $ parseUnsignedInteger
     (offsetBytes transbuf buf) (subtract translen len) radix char_to_int

-- some conveniences for use with tok_integral
Ian Lynagh's avatar
Ian Lynagh committed
1059
1060
1061
tok_num :: (Integer -> Integer)
        -> Int -> Int
        -> (Integer, (Char->Int)) -> Action
1062
tok_num = tok_integral ITinteger
Ian Lynagh's avatar
Ian Lynagh committed
1063
1064
1065
tok_primint :: (Integer -> Integer)
            -> Int -> Int
            -> (Integer, (Char->Int)) -> Action
1066
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1067
1068
tok_primword :: Int -> Int
             -> (Integer, (Char->Int)) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
1069
tok_primword = tok_integral ITprimword positive
Ian Lynagh's avatar
Ian Lynagh committed
1070
positive, negative :: (Integer -> Integer)
1071
1072
positive = id
negative = negate
Ian Lynagh's avatar
Ian Lynagh committed
1073
decimal, octal, hexadecimal :: (Integer, Char -> Int)
1074
1075
1076
1077
1078
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
Ian Lynagh's avatar
Ian Lynagh committed
1079
tok_float, tok_primfloat, tok_primdouble :: String -> Token
1080
tok_float        str = ITrational   $! readRational str
1081
1082
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
1083
1084
1085
1086
1087
1088

-- -----------------------------------------------------------------------------
-- Layout processing

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1089
do_bol span _str _len = do
1090
	pos <- getOffside
1091
1092
1093
1094
1095
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1096
		return (L span ITvccurly)
1097
1098
1099
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1100
		return (L span ITsemi)
1101
1102
1103
1104
1105
1106
	    GT -> do
		popLexState
		lexToken

-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
Ian Lynagh's avatar
Ian Lynagh committed
1107
maybe_layout :: Token -> P ()
1108
maybe_layout ITdo	= pushLexState layout_do
1109
maybe_layout ITmdo	= pushLexState layout_do
1110
1111
1112
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1113
maybe_layout ITrec	= pushLexState layout
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
maybe_layout _	        = return ()

-- Pushing a new implicit layout context.  If the indentation of the
-- next token is not greater than the previous layout context, then
-- Haskell 98 says that the new layout context should be empty; that is
-- the lexer must generate {}.
--
-- We are slightly more lenient than this: when the new context is started
-- by a 'do', then we allow the new context to be at the same indentation as
-- the previous context.  This is what the 'strict' argument is for.
--
Ian Lynagh's avatar
Ian Lynagh committed
1125
new_layout_context :: Bool -> Action
1126
new_layout_context strict span _buf _len = do
1127
    popLexState
1128
    (AI _ offset _) <- getInput
1129
1130
1131
1132
1133
1134
1135
1136
    ctx <- getContext
    case ctx of
	Layout prev_off : _  | 
	   (strict     && prev_off >= offset  ||
	    not strict && prev_off > offset) -> do
		-- token is indented to the left of the previous context.
		-- we must generate a {} sequence now.
		pushLexState layout_left
1137
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1138
	_ -> do
1139
		setContext (Layout offset : ctx)
1140
		return (L span ITvocurly)
1141

Ian Lynagh's avatar
Ian Lynagh committed
1142
do_layout_left :: Action
1143
do_layout_left span _buf _len = do
1144
1145
    popLexState
    pushLexState bol  -- we must be at the start of a line
1146
    return (L span ITvccurly)
1147
1148
1149
1150

-- -----------------------------------------------------------------------------
-- LINE pragmas

1151
1152
setLine :: Int -> Action
setLine code span buf len = do
1153
  let line = parseUnsignedInteger buf len 10 octDecDigit
1154
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1155
1156
1157
1158
1159
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1160
1161
setFile :: Int -> Action
setFile code span buf len = do
1162
  let file = lexemeToFastString (stepOn buf) (len-2)
1163
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1164
1165
1166
1167
  popLexState
  pushLexState code
  lexToken

1168
1169
1170
1171
1172

-- -----------------------------------------------------------------------------
-- Options, includes and language pragmas.

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1173
lex_string_prag mkTok span _buf _len
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
    = do input <- getInput
         start <- getSrcLoc
         tok <- go [] input
         end <- getSrcLoc
         return (L (mkSrcSpan start end) tok)
    where go acc input
              = if isString input "#-}"
                   then do setInput input
                           return (mkTok (reverse acc))
                   else case alexGetChar input of
                          Just (c,i) -> go (c:acc) i
                          Nothing -> err input
twanvl's avatar
twanvl committed
1186
          isString _ [] = True
1187
1188
1189
1190
1191
1192
1193
          isString i (x:xs)
              = case alexGetChar i of
                  Just (c,i') | c == x    -> isString i' xs
                  _other -> False
          err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"


1194
1195
1196
1197
1198
1199
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1200
lex_string_tok span _buf _len = do
1201
1202
  tok <- lex_string ""
  end <- getSrcLoc 
1203
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1204

1205
lex_string :: String -> P Token
1206
1207
lex_string s = do
  i <- getInput
1208
  case alexGetChar' i of
1209
1210
1211
1212
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1213
1214
	magicHash <- extension magicHashEnabled
	if magicHash
1215
1216
	  then do
	    i <- getInput
1217
	    case alexGetChar' i of
1218
1219
1220
1221
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1222
                    else let s' = mkZFastString (reverse s) in
1223
			 return (ITprimstring s')
1224
1225
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
	      _other ->
		return (ITstring (mkFastString (reverse s)))
	  else
		return (ITstring (mkFastString (reverse s)))

    Just ('\\',i)
	| Just ('&',i) <- next -> do 
		setInput i; lex_string s
	| Just (c,i) <- next, is_space c -> do 
		setInput i; lex_stringgap s
1236
	where next = alexGetChar' i
1237

1238
1239
1240
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1241

Ian Lynagh's avatar
Ian Lynagh committed
1242
lex_stringgap :: String -> P Token
1243
1244
1245
1246
1247
1248
1249
1250
1251
lex_stringgap s = do
  c <- getCharOrFail
  case c of
    '\\' -> lex_string s
    c | is_space c -> lex_stringgap s
    _other -> lit_error


lex_char_tok :: Action
1252
1253
1254
1255
1256
1257
-- Here we are basically parsing character literals, such as 'x' or '\n'
-- but, when Template Haskell is on, we additionally spot
-- 'x and ''T, returning ITvarQuote and ITtyQuote respectively, 
-- but WIHTOUT CONSUMING the x or T part  (the parser does that).
-- So we have to do two characters of lookahead: when we see 'x we need to
-- see if there's a trailing quote
twanvl's avatar
twanvl committed
1258
lex_char_tok span _buf _len = do	-- We've seen '
1259
   i1 <- getInput	-- Look ahead to first character
1260
   let loc = srcSpanStart span
1261
   case alexGetChar' i1 of
1262
1263
	Nothing -> lit_error 

1264
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1265
1266
1267
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1268
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1269
1270
		   else lit_error

twanvl's avatar
twanvl committed