Lexer.x 94.5 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9
--
-- 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
Ian Lynagh's avatar
Ian Lynagh committed
10 11
-- file is maintained.  Every token has a start and end RealSrcLoc
-- attached to it.
12 13 14 15 16 17 18 19 20 21 22
--
-----------------------------------------------------------------------------

--   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

23
--   qualified operator NOTES.
Ian Lynagh's avatar
Ian Lynagh committed
24
--
25 26 27 28 29 30 31 32 33
--   - 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.

34
{
35
-- XXX The above flags turn off warnings in the generated code:
36
{-# LANGUAGE BangPatterns #-}
37 38 39 40 41 42 43 44
{-# 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 #-}
45

46 47
{-# OPTIONS_GHC -funbox-strict-fields #-}

48
module Lexer (
49
   Token(..), lexer, pragState, mkPState, PState(..),
Ian Lynagh's avatar
Ian Lynagh committed
50
   P(..), ParseResult(..), getSrcLoc,
51
   getPState, getDynFlags, withThisPackage,
52
   failLocMsgP, failSpanMsgP, srcParseFail,
Ian Lynagh's avatar
Ian Lynagh committed
53
   getMessages,
54
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
vivian's avatar
vivian committed
55
   activeContext, nextIsEOF,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
56
   getLexState, popLexState, pushLexState,
57
   extension, bangPatEnabled, datatypeContextsEnabled,
58
   traditionalRecordSyntaxEnabled,
59
   typeLiteralsEnabled,
60 61
   explicitForallEnabled,
   inRulePrag,
cactus's avatar
cactus committed
62 63 64
   explicitNamespacesEnabled,
   patternSynonymsEnabled,
   sccProfilingOn, hpcEnabled,
Jedai's avatar
Jedai committed
65 66
   addWarning,
   lexTokenStream
67 68
  ) where

69 70
import Bag
import ErrUtils
71 72 73 74 75
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
76
import DynFlags
77
import Module
78
import Ctype
Ian Lynagh's avatar
Ian Lynagh committed
79 80
import BasicTypes       ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) )
import Util             ( readRational )
81

82
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
83
import Data.Bits
84
import Data.ByteString (ByteString)
85
import Data.Char
86
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
87
import Data.Maybe
88 89
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
90
import Data.Ratio
91
import Data.Word
92 93
}

94
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
95
$whitechar   = [\ \n\r\f\v $unispace]
96
$white_no_nl = $whitechar # \n
97
$tab         = \t
98 99

$ascdigit  = 0-9
100
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
101
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
102 103 104
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
105
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
106
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
107 108
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

109
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
110
$asclarge  = [A-Z]
111 112
$large     = [$asclarge $unilarge]

113
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
114
$ascsmall  = [a-z]
115 116
$small     = [$ascsmall $unismall \_]

117
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
118
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
119

120
$binit     = 0-1
Ian Lynagh's avatar
Ian Lynagh committed
121
$octit     = 0-7
122
$hexit     = [$decdigit A-F a-f]
123 124 125 126
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

127 128
$pragmachar = [$small $large $digit]

129 130
$docsym    = [\| \^ \* \$]

131 132 133 134 135 136
@varid     = $small $idchar*
@conid     = $large $idchar*

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

137
@decimal     = $decdigit+
138
@binary      = $binit+
139 140 141 142 143 144 145 146 147
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

148 149 150 151 152
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

153 154 155
haskell :-

-- everywhere: skip whitespace and comments
Ian Lynagh's avatar
Ian Lynagh committed
156
$white_no_nl+ ;
157
$tab+         { warn Opt_WarnTabs (text "Tab character") }
158 159 160 161 162

-- 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
163 164 165 166
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
167 168 169 170

-- 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.
171 172

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
173
-- out here.
174 175 176 177 178 179 180

-- 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
181 182
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
183 184 185

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

Jedai's avatar
Jedai committed
186
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
187 188 189 190 191 192 193

-- 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
194
"---"\-* [^$symbol :] .* { lineCommentToken }
195 196 197 198

-- 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
199
"--"\-* / { atEOL } { lineCommentToken }
200 201 202 203

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

Jedai's avatar
Jedai committed
204
"-- " / { atEOL } { lineCommentToken }
205 206 207 208 209 210 211 212 213 214 215

-- '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> {
Ian Lynagh's avatar
Ian Lynagh committed
216 217 218 219 220
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
221 222 223 224 225
}

-- 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.
226
<layout, layout_do, layout_if> {
Ian Lynagh's avatar
Ian Lynagh committed
227 228 229 230
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
231 232
}

233 234 235 236 237 238
-- after an 'if', a vertical bar starts a layout context for MultiWayIf
<layout_if> {
  \| / { notFollowedBySymbol }          { new_layout_context True ITvbar }
  ()                                    { pop }
}

239
-- do is treated in a subtly different way, see new_layout_context
240 241
<layout>    ()                          { new_layout_context True  ITvocurly }
<layout_do> ()                          { new_layout_context False ITvocurly }
242 243 244 245

-- 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.
Ian Lynagh's avatar
Ian Lynagh committed
246
<layout_left>  ()                       { do_layout_left }
247

Ian Lynagh's avatar
Ian Lynagh committed
248
<0,option_prags> \n                     { begin bol }
249

250 251
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
252 253 254

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
Ian Lynagh's avatar
Ian Lynagh committed
255 256 257
<line_prag1> $decdigit+                 { setLine line_prag1a }
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
258 259 260

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
Ian Lynagh's avatar
Ian Lynagh committed
261 262 263
<line_prag2> $decdigit+                 { setLine line_prag2a }
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
264 265
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
266

267
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
268
  "{-#" $whitechar* $pragmachar+
269 270 271 272 273
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
274

275
  -- We ignore all these pragmas, but don't generate a warning for them
276 277
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
278 279

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
280
  "#-}"                          { endPrag }
281 282
}

283
<option_prags> {
284 285 286
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

Ian Lynagh's avatar
Ian Lynagh committed
287
  "-- #"                           { multiline_doc_comment }
David Waern's avatar
David Waern committed
288 289
}

290 291
<0> {
  -- In the "0" mode we ignore these pragmas
292
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
293 294 295
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
296
<0> {
Jedai's avatar
Jedai committed
297
  "-- #" .* { lineCommentToken }
298
}
299

300
<0,option_prags> {
301 302
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
303 304
}

305 306
-- '0' state: ordinary lexemes

307 308
-- Haddock comments

309
<0,option_prags> {
David Waern's avatar
David Waern committed
310 311
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
312 313
}

314 315
-- "special" symbols

316
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
317 318
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
319
}
Ian Lynagh's avatar
Ian Lynagh committed
320

321
<0> {
322 323 324
  "[|"        / { ifExtension thEnabled } { token ITopenExpQuote }
  "[||"       / { ifExtension thEnabled } { token ITopenTExpQuote }
  "[e|"       / { ifExtension thEnabled } { token ITopenExpQuote }
325
  "[e||"      / { ifExtension thEnabled } { token ITopenTExpQuote }
326 327 328 329 330 331 332 333 334
  "[p|"       / { ifExtension thEnabled } { token ITopenPatQuote }
  "[d|"       / { ifExtension thEnabled } { layout_token ITopenDecQuote }
  "[t|"       / { ifExtension thEnabled } { token ITopenTypQuote }
  "|]"        / { ifExtension thEnabled } { token ITcloseQuote }
  "||]"       / { ifExtension thEnabled } { token ITcloseTExpQuote }
  \$ @varid   / { ifExtension thEnabled } { skip_one_varid ITidEscape }
  "$$" @varid / { ifExtension thEnabled } { skip_two_varid ITidTyEscape }
  "$("        / { ifExtension thEnabled } { token ITparenEscape }
  "$$("       / { ifExtension thEnabled } { token ITparenTyEscape }
335

336 337 338 339
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

340
  "[" @varid "|"  / { ifExtension qqEnabled }
341
                     { lex_quasiquote_tok }
342 343 344 345

  -- qualified quasi-quote (#5555)
  "[" @qual @varid "|"  / { ifExtension qqEnabled }
                          { lex_qquasiquote_tok }
346 347
}

348
<0> {
ross's avatar
ross committed
349
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
Ian Lynagh's avatar
Ian Lynagh committed
350
                                        { special IToparenbar }
351 352 353
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

354
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
355
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
356 357
}

358
<0> {
359
  "(#" / { ifExtension unboxedTuplesEnabled }
Ian Lynagh's avatar
Ian Lynagh committed
360 361 362 363 364
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

365
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
366 367 368 369 370 371 372 373 374 375
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
376 377
}

378
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
379 380 381
  @qual @varid                  { idtoken qvarid }
  @qual @conid                  { idtoken qconid }
  @varid                        { varid }
382
  @conid                        { idtoken conid }
383 384
}

385
<0> {
386 387 388
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
389
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
390 391
}

392 393
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
394
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
395 396
  @qual @varsym                                    { idtoken qvarsym }
  @qual @consym                                    { idtoken qconsym }
397 398
  @varsym                                          { varsym }
  @consym                                          { consym }
399 400
}

401 402
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
403
<0> {
404
  -- Normal integral literals (:: Num a => a, from Integer)
405
  @decimal                                                               { tok_num positive 0 0 decimal }
406
  0[bB] @binary                / { ifExtension binaryLiteralsEnabled }   { tok_num positive 2 2 binary }
407 408 409
  0[oO] @octal                                                           { tok_num positive 2 2 octal }
  0[xX] @hexadecimal                                                     { tok_num positive 2 2 hexadecimal }
  @negative @decimal           / { ifExtension negativeLiteralsEnabled } { tok_num negative 1 1 decimal }
410 411
  @negative 0[bB] @binary      / { ifExtension negativeLiteralsEnabled `alexAndPred`
                                   ifExtension binaryLiteralsEnabled }   { tok_num negative 3 3 binary }
412 413
  @negative 0[oO] @octal       / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 octal }
  @negative 0[xX] @hexadecimal / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 hexadecimal }
414 415

  -- Normal rational literals (:: Fractional a => a, from Rational)
416 417
  @floating_point                                                        { strtoken tok_float }
  @negative @floating_point    / { ifExtension negativeLiteralsEnabled } { strtoken tok_float }
418 419
}

420
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
421
  -- Unboxed ints (:: Int#) and words (:: Word#)
422 423
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
424
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
425 426
  0[bB] @binary                \# / { ifExtension magicHashEnabled `alexAndPred`
                                      ifExtension binaryLiteralsEnabled } { tok_primint positive 2 3 binary }
Ian Lynagh's avatar
Ian Lynagh committed
427 428 429
  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 }
430 431
  @negative 0[bB] @binary      \# / { ifExtension magicHashEnabled `alexAndPred`
                                      ifExtension binaryLiteralsEnabled } { tok_primint negative 3 4 binary }
Ian Lynagh's avatar
Ian Lynagh committed
432
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
433
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
434

Ian Lynagh's avatar
Ian Lynagh committed
435
  @decimal                     \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
436 437
  0[bB] @binary                \# \# / { ifExtension magicHashEnabled `alexAndPred`
                                         ifExtension binaryLiteralsEnabled } { tok_primword 2 4 binary }
Ian Lynagh's avatar
Ian Lynagh committed
438 439 440
  0[oO] @octal                 \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
  0[xX] @hexadecimal           \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }

441 442
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
443 444
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
445 446 447 448 449 450
}

-- 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.
451
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
452 453
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
454 455 456 457 458 459
}

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

460
data Token
Ian Lynagh's avatar
Ian Lynagh committed
461
  = ITas                        -- Haskell keywords
462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
  | ITdo
  | ITelse
  | IThiding
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere

Ian Lynagh's avatar
Ian Lynagh committed
486
  | ITforall                    -- GHC extension keywords
487 488 489 490 491
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
492
  | ITinterruptible
493 494 495
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
496
  | ITcapiconv
497
  | ITprimcallconv
thoughtpolice's avatar
thoughtpolice committed
498
  | ITjavascriptcallconv
499
  | ITmdo
500
  | ITfamily
501
  | ITrole
502 503 504
  | ITgroup
  | ITby
  | ITusing
cactus's avatar
cactus committed
505
  | ITpattern
506

Ian Lynagh's avatar
Ian Lynagh committed
507
  -- Pragmas
508
  | ITinline_prag InlineSpec RuleMatchInfo
Ian Lynagh's avatar
Ian Lynagh committed
509 510
  | ITspec_prag                 -- SPECIALISE
  | ITspec_inline_prag Bool     -- SPECIALISE INLINE (or NOINLINE)
511 512
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
513
  | ITwarning_prag
514 515 516
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
517
  | ITgenerated_prag
518
  | ITcore_prag                 -- hdaume: core annotations
519
  | ITunpack_prag
520
  | ITnounpack_prag
521
  | ITann_prag
522
  | ITclose_prag
523 524 525
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
526 527
  | ITvect_prag
  | ITvect_scalar_prag
528
  | ITnovect_prag
529
  | ITminimal_prag
530 531 532
  | IToverlappable_prag         -- instance overlap mode
  | IToverlapping_prag          -- instance overlap mode
  | IToverlaps_prag             -- instance overlap mode
533
  | ITincoherent_prag           -- instance overlap mode
534
  | ITctype
535

Ian Lynagh's avatar
Ian Lynagh committed
536
  | ITdotdot                    -- reserved symbols
537 538 539 540
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
541
  | ITlcase
542 543 544 545 546
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
batterseapower's avatar
batterseapower committed
547
  | ITtildehsh
548 549 550 551 552 553
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
554
  | ITbiglam                    -- GHC-extension symbols
555

Ian Lynagh's avatar
Ian Lynagh committed
556
  | ITocurly                    -- special symbols
557 558 559 560
  | ITccurly
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
561 562
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
563 564 565 566 567 568 569 570 571
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote
dreixel's avatar
dreixel committed
572
  | ITsimpleQuote               --  '
573

Ian Lynagh's avatar
Ian Lynagh committed
574
  | ITvarid   FastString        -- identifiers
575 576 577 578 579 580 581
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)
582 583
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
584

Ian Lynagh's avatar
Ian Lynagh committed
585
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
586 587 588 589

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
590
  | ITrational   FractionalLit
591 592

  | ITprimchar   Char
593
  | ITprimstring ByteString
594
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
595
  | ITprimword   Integer
596 597
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
598

599
  -- Template Haskell extension tokens
Ian Lynagh's avatar
Ian Lynagh committed
600 601 602 603 604
  | ITopenExpQuote              --  [| or [e|
  | ITopenPatQuote              --  [p|
  | ITopenDecQuote              --  [d|
  | ITopenTypQuote              --  [t|
  | ITcloseQuote                --  |]
605 606
  | ITopenTExpQuote             --  [||
  | ITcloseTExpQuote            --  ||]
Ian Lynagh's avatar
Ian Lynagh committed
607 608
  | ITidEscape   FastString     --  $x
  | ITparenEscape               --  $(
609 610
  | ITidTyEscape   FastString   --  $$x
  | ITparenTyEscape             --  $$(
Ian Lynagh's avatar
Ian Lynagh committed
611
  | ITtyQuote                   --  ''
612 613 614 615 616 617 618 619
  | ITquasiQuote (FastString,FastString,RealSrcSpan)
    -- ITquasiQuote(quoter, quote, loc)
    -- represents a quasi-quote of the form
    -- [quoter| quote |]
  | ITqQuasiQuote (FastString,FastString,FastString,RealSrcSpan)
    -- ITqQuasiQuote(Qual, quoter, quote, loc)
    -- represents a qualified quasi-quote of the form
    -- [Qual.quoter| quote |]
620 621 622 623

  -- Arrow notation extension
  | ITproc
  | ITrec
Ian Lynagh's avatar
Ian Lynagh committed
624 625 626 627 628 629
  | IToparenbar                 --  (|
  | ITcparenbar                 --  |)
  | ITlarrowtail                --  -<
  | ITrarrowtail                --  >-
  | ITLarrowtail                --  -<<
  | ITRarrowtail                --  >>-
630

Ian Lynagh's avatar
Ian Lynagh committed
631 632
  | ITunknown String            -- Used when the lexer can't make sense of it
  | ITeof                       -- end of file token
633 634 635 636 637 638 639

  -- 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
640
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
641 642
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
643

644
  deriving Show
645 646 647 648 649 650 651 652

-- 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)
--
653
reservedWordsFM :: UniqFM (Token, ExtsBitmap)
654
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681
    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 ),

682 683 684
         ( "forall",         ITforall,        xbit ExplicitForallBit .|.
                                              xbit InRulePragBit),
         ( "mdo",            ITmdo,           xbit RecursiveDoBit),
685 686 687
             -- See Note [Lexing type pseudo-keywords]
         ( "family",         ITfamily,        0 ),
         ( "role",           ITrole,          0 ),
688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709
         ( "pattern",        ITpattern,       xbit PatternSynonymsBit),
         ( "group",          ITgroup,         xbit TransformComprehensionsBit),
         ( "by",             ITby,            xbit TransformComprehensionsBit),
         ( "using",          ITusing,         xbit TransformComprehensionsBit),

         ( "foreign",        ITforeign,       xbit FfiBit),
         ( "export",         ITexport,        xbit FfiBit),
         ( "label",          ITlabel,         xbit FfiBit),
         ( "dynamic",        ITdynamic,       xbit FfiBit),
         ( "safe",           ITsafe,          xbit FfiBit .|.
                                              xbit SafeHaskellBit),
         ( "interruptible",  ITinterruptible, xbit InterruptibleFfiBit),
         ( "unsafe",         ITunsafe,        xbit FfiBit),
         ( "stdcall",        ITstdcallconv,   xbit FfiBit),
         ( "ccall",          ITccallconv,     xbit FfiBit),
         ( "capi",           ITcapiconv,      xbit CApiFfiBit),
         ( "prim",           ITprimcallconv,  xbit FfiBit),
         ( "javascript",     ITjavascriptcallconv, xbit FfiBit),

         ( "rec",            ITrec,           xbit ArrowsBit .|.
                                              xbit RecursiveDoBit),
         ( "proc",           ITproc,          xbit ArrowsBit)
710 711
     ]

712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727
{-----------------------------------
Note [Lexing type pseudo-keywords]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

One might think that we wish to treat 'family' and 'role' as regular old
varids whenever -XTypeFamilies and -XRoleAnnotations are off, respectively.
But, there is no need to do so. These pseudo-keywords are not stolen syntax:
they are only used after the keyword 'type' at the top-level, where varids are
not allowed. Furthermore, checks further downstream (TcTyClsDecls) ensure that
type families and role annotations are never declared without their extensions
on. In fact, by unconditionally lexing these pseudo-keywords as special, we
can get better error messages.

Also, note that these are included in the `varid` production in the parser --
a key detail to make all this work.
-------------------------------------}
728

729
reservedSymsFM :: UniqFM (Token, ExtsBitmap -> Bool)
730
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
731 732 733 734 735 736 737 738 739 740 741 742
    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)
743
       ,("~#",  ITtildehsh, magicHashEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
744 745 746 747 748
       ,("=>",  ITdarrow,   always)
       ,("-",   ITminus,    always)
       ,("!",   ITbang,     always)

        -- For data T (a::*) = MkT
749
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
750
        -- For 'forall a . t'
751
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
752 753 754 755 756

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
757

Ian Lynagh's avatar
Ian Lynagh committed
758 759
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
760
       ,("∀",   ITforall, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
761 762
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
763 764 765 766 767 768 769 770

       ,("⤙",   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
771 772 773
        -- 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).
774 775 776 777 778
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
779
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
780

781
special :: Token -> Action
twanvl's avatar
twanvl committed
782
special tok span _buf _len = return (L span tok)
783

784
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
785 786
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
787

788 789
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
790

791
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
792
skip_one_varid f span buf len
793
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
794

795 796 797 798
skip_two_varid :: (FastString -> Token) -> Action
skip_two_varid f span buf len
  = return (L span $! f (lexemeToFastString (stepOn (stepOn buf)) (len-2)))

799
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
800
strtoken f span buf len =
801
  return (L span $! (f $! lexemeToString buf len))
802

803
init_strtoken :: Int -> (String -> Token) -> Action
804
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
805
init_strtoken drop f span buf len =
806
  return (L span $! (f $! lexemeToString buf (len-drop)))
807 808

begin :: Int -> Action
809
begin code _span _str _len = do pushLexState code; lexToken
810 811

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
812 813
pop _span _buf _len = do _ <- popLexState
                         lexToken
814

815 816 817 818 819 820 821 822 823 824 825
hopefully_open_brace :: Action
hopefully_open_brace span buf len
 = do relaxed <- extension relaxedLayout
      ctx <- getContext
      (AI l _) <- getInput
      let offset = srcLocCol l
          isOK = relaxed ||
                 case ctx of
                 Layout prev_off : _ -> prev_off < offset
                 _                   -> True
      if isOK then pop_and open_brace span buf len
Ian Lynagh's avatar
Ian Lynagh committed
826
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
827

828
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
829 830
pop_and act span buf len = do _ <- popLexState
                              act span buf len
831

832
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
833
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
834 835
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

836 837 838 839
{-# INLINE nextCharIsNot #-}
nextCharIsNot :: StringBuffer -> (Char -> Bool) -> Bool
nextCharIsNot buf p = not (nextCharIs buf p)

840
notFollowedBy :: Char -> AlexAccPred ExtsBitmap
Ian Lynagh's avatar
Ian Lynagh committed
841
notFollowedBy char _ _ _ (AI _ buf)
842
  = nextCharIsNot buf (== char)
843

844
notFollowedBySymbol :: AlexAccPred ExtsBitmap
845
notFollowedBySymbol _ _ _ (AI _ buf)
846
  = nextCharIsNot buf (`elem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
847

848 849 850 851 852
-- 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.
853
isNormalComment :: AlexAccPred ExtsBitmap
854
isNormalComment bits _ _ (AI _ buf)
855
  | haddockEnabled bits = notFollowedByDocOrPragma
856
  | otherwise           = nextCharIsNot buf (== '#')
857 858
  where
    notFollowedByDocOrPragma
859
       = afterOptionalSpace buf (\b -> nextCharIsNot b (`elem` "|^*$#"))
860

861 862 863 864 865
afterOptionalSpace :: StringBuffer -> (StringBuffer -> Bool) -> Bool
afterOptionalSpace buf p
    = if nextCharIs buf (== ' ')
      then p (snd (nextChar buf))
      else p buf
866

867
atEOL :: AlexAccPred ExtsBitmap
868
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
869

870
ifExtension :: (ExtsBitmap -> Bool) -> AlexAccPred ExtsBitmap
871 872
ifExtension pred bits _ _ _ = pred bits

873 874 875
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
876
    worker commentAcc input docType oneLine = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
877
      Just ('\n', input')
878 879 880 881 882 883
        | 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
Ian Lynagh's avatar
Ian Lynagh committed
884

885 886
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
887 888 889
        check input = case alexGetChar' input of
          Just ('-', input) -> case alexGetChar' input of
            Just ('-', input) -> case alexGetChar' input of
890 891 892 893 894
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

895
        dropNonNewlineSpace input = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
896
          Just (c, input')
897 898 899 900
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
901 902 903 904 905
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

906 907 908 909
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
910
nested_comment :: P (RealLocated Token) -> Action
911
nested_comment cont span _str _len = do
912
  input <- getInput
Jedai's avatar
Jedai committed
913
  go "" (1::Int) input
914
  where
Jedai's avatar
Jedai committed
915 916 917 918 919
    go commentAcc 0 input = do setInput input
                               b <- extension rawTokenStreamEnabled
                               if b
                                 then docCommentEnd input commentAcc ITblockComment _str span
                                 else cont
920
    go commentAcc n input = case alexGetChar' input of
921
      Nothing -> errBrace input span
922
      Just ('-',input) -> case alexGetChar' input of
923
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
924 925
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
926
      Just ('\123',input) -> case alexGetChar' input of
927
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
928 929 930
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
931 932 933 934

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
935
    go commentAcc input docType _ = case alexGetChar' input of
936
      Nothing -> errBrace input span
937
      Just ('-',input) -> case alexGetChar' input of
938
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
939
        Just ('\125',input) ->
940
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
941
        Just (_,_) -> go ('-':commentAcc) input docType False
942
      Just ('\123', input) -> case alexGetChar' input of
943 944 945 946 947
        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
948
        Just (_,_) -> go ('\123':commentAcc) input docType False
949 950
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
951 952
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
953
withLexedDocType lexDocComment = do
954
  input@(AI _ buf) <- getInput
955 956 957 958
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
959 960
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
961
    _ -> panic "withLexedDocType: Bad doc type"
Ian Lynagh's avatar
Ian Lynagh committed
962
 where
963
    lexDocSection n input = case alexGetChar' input of
964
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
965
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
966 967
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

968 969 970
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
971
rulePrag span _buf _len = do
972
  setExts (.|. xbit InRulePragBit)
973 974 975
  return (L span ITrules_prag)

endPrag :: Action
976
endPrag span _buf _len = do
977
  setExts (.&. complement (xbit InRulePragBit))
978 979
  return (L span ITclose_prag)

980 981 982 983
-- 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
Ian Lynagh's avatar
Ian Lynagh committed
984
-- than what was lexed by Alex, and the lexToken function doesn't know this, so
985
-- it writes the wrong token length to the parser state. This function is
Ian Lynagh's avatar
Ian Lynagh committed
986
-- called afterwards, so it can just update the state.
987 988

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
989
                 RealSrcSpan -> P (RealLocated Token)
990 991
docCommentEnd input commentAcc docType buf span = do
  setInput input
992
  let (AI loc nextBuf) = input
993
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
994
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
995
      last_len = byteDiff buf nextBuf
Ian Lynagh's avatar
Ian Lynagh committed
996

997
  span `seq` setLastToken span' last_len
998
  return (L span' (docType comment))
Ian Lynagh's avatar
Ian Lynagh committed
999

Ian Lynagh's avatar
Ian Lynagh committed
1000 1001
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
1002

1003
open_brace, close_brace :: Action
Ian Lynagh's avatar
Ian Lynagh committed