Lexer.x 57.7 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 22 23
--
-- 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:
--    - Unicode
--    - 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

{
24
{-# OPTIONS -w #-}
25 26 27
-- 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
28
--     http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
29
-- for details
twanvl's avatar
twanvl committed
30 31 32
--
-- Note that Alex itself generates code with with some unused bindings and
-- without type signatures, so removing the flag might not be possible.
33

34 35
{-# OPTIONS_GHC -funbox-strict-fields #-}

36
module Lexer (
37
   Token(..), lexer, pragState, mkPState, PState(..),
38
   P(..), ParseResult(..), getSrcLoc, 
39
   failLocMsgP, failSpanMsgP, srcParseFail,
40
   getMessages,
41
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
42
   getLexState, popLexState, pushLexState,
43 44
   extension, standaloneDerivingEnabled, bangPatEnabled,
   addWarning
45 46
  ) where

47 48
import Bag
import ErrUtils
49 50 51 52 53 54
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
55
import DynFlags
56
import Ctype
57
import Util		( maybePrefixMatch, readRational )
58

59
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
60
import Data.Bits
61
import Data.Char 	( chr, ord, isSpace )
Simon Marlow's avatar
Simon Marlow committed
62 63
import Data.Ratio
import Debug.Trace
64

Ian Lynagh's avatar
Ian Lynagh committed
65
import Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper )
66 67
}

68
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
69
$whitechar   = [\ \n\r\f\v $unispace]
70
$white_no_nl = $whitechar # \n
71
$tab         = \t
72 73

$ascdigit  = 0-9
74
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
75
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
76 77 78
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
79
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
80
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
81 82
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

83
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
84
$asclarge  = [A-Z]
85 86
$large     = [$asclarge $unilarge]

87
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
88
$ascsmall  = [a-z]
89 90
$small     = [$ascsmall $unismall \_]

91
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
92
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
93 94

$octit	   = 0-7
95
$hexit     = [$decdigit A-F a-f]
96 97 98 99
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

100 101
$docsym    = [\| \^ \* \$]

102 103 104 105 106 107
@varid     = $small $idchar*
@conid     = $large $idchar*

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

108
@decimal     = $decdigit+
109 110 111 112 113 114 115 116 117
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

118 119 120 121 122
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

123 124 125 126
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
127
$tab+         { warn Opt_WarnTabs (text "Tab character") }
128 129 130 131 132

-- 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
133 134 135 136
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
137 138 139 140

-- 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.
141 142 143 144 145 146 147 148 149 150

-- 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.

David Waern's avatar
David Waern committed
151
"-- " ~[$docsym \#] .* ;
152 153 154 155
"--" [^$symbol : \ ] .* ;

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

156
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } ;
157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174

-- 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.

"---"\-* [^$symbol :] .* ;

-- 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.

"--"\-* / { atEOL } ;

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

"-- " / { atEOL } ;
175 176 177 178 179 180 181 182 183 184 185 186 187

-- '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 }
188
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
189
  ^\# \! .* \n				; -- #!, for scripts
190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
  ()					{ 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 }

212
<0,option_prags> \n				{ begin bol }
213 214 215 216 217

"{-#" $whitechar* (line|LINE) 		{ begin line_prag2 }

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
218
<line_prag1> $decdigit+			{ setLine line_prag1a }
219
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
220 221 222 223
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
224
<line_prag2> $decdigit+			{ setLine line_prag2a }
225
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
226 227 228
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
229

230 231 232
-- We only want RULES pragmas to be picked up when explicit forall
-- syntax is enabled is on, because the contents of the pragma always
-- uses it. If it's not on then we're sure to get a parse error.
233
-- (ToDo: we should really emit a warning when ignoring pragmas)
234 235
-- XXX Now that we can enable this without the -fglasgow-exts hammer,
-- is it better just to let the parse error happen?
236
<0>
237
  "{-#" $whitechar* (RULES|rules) / { ifExtension explicitForallEnabled } { token ITrules_prag }
238

239
<0,option_prags> {
240 241 242 243 244
  "{-#" $whitechar* (INLINE|inline)	{ token (ITinline_prag True) }
  "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
  					{ token (ITinline_prag False) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
  					{ token ITspec_prag }
245
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
246 247 248 249
	$whitechar* (INLINE|inline)	{ token (ITspec_inline_prag True) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
	$whitechar* (NO(T?)INLINE|no(t?)inline)
					{ token (ITspec_inline_prag False) }
250
  "{-#" $whitechar* (SOURCE|source)	{ token ITsource_prag }
Ian Lynagh's avatar
Ian Lynagh committed
251 252
  "{-#" $whitechar* (WARNING|warning)
  					{ token ITwarning_prag }
253 254 255
  "{-#" $whitechar* (DEPRECATED|deprecated)
  					{ token ITdeprecated_prag }
  "{-#" $whitechar* (SCC|scc)		{ token ITscc_prag }
andy@galois.com's avatar
andy@galois.com committed
256 257
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
258
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
259
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
260

261 262 263 264
  -- We ignore all these pragmas, but don't generate a warning for them
  -- CFILES is a hugs-only thing.
  "{-#" $whitechar* (OPTIONS_HUGS|options_hugs|OPTIONS_NHC98|options_nhc98|OPTIONS_JHC|options_jhc|CFILES|cfiles)
                    { nested_comment lexToken }
265 266 267 268 269

  -- ToDo: should only be valid inside a pragma:
  "#-}" 				{ token ITclose_prag}
}

270
<option_prags> {
David Waern's avatar
David Waern committed
271 272
  "{-#"  $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc)
273
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
274 275 276 277 278 279 280
  "{-#"  $whitechar* (OPTIONS_HADDOCK|options_haddock)
                                         { lex_string_prag ITdocOptions }
  "-- #"                                 { multiline_doc_comment }
  "{-#"  $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
  "{-#"  $whitechar* (INCLUDE|include)   { lex_string_prag ITinclude_prag }
}

281 282 283 284 285 286
<0> {
  -- In the "0" mode we ignore these pragmas
  "{-#"  $whitechar* (OPTIONS|options|OPTIONS_GHC|options_ghc|OPTIONS_HADDOCK|options_haddock|LANGUAGE|language|INCLUDE|include)
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
287 288
<0> {
  "-- #" .* ;
289
}
290

291
<0,option_prags> {
292 293
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
294 295
}

296 297
-- '0' state: ordinary lexemes

298 299
-- Haddock comments

300
<0> {
David Waern's avatar
David Waern committed
301 302
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
303 304
}

305 306
-- "special" symbols

307
<0> {
308 309 310 311
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
312
<0> {
313 314 315 316 317 318 319 320
  "[|"	    / { 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 }
321 322 323

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
324 325
}

326
<0> {
ross's avatar
ross committed
327 328
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
329 330 331
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

332
<0> {
333 334 335
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

336
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
337 338 339 340 341 342
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

343
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
344 345
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
346 347
}

348
<0,option_prags> {
349 350 351 352 353 354 355 356 357 358 359 360
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

361
<0,option_prags> {
362
  @qual @varid			{ idtoken qvarid }
363 364 365 366 367
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

368
<0> {
369 370 371 372
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
373 374 375 376
}

-- ToDo: M.(,,,)

377
<0> {
378 379 380 381 382 383
  @qual @varsym			{ idtoken qvarsym }
  @qual @consym			{ idtoken qconsym }
  @varsym			{ varsym }
  @consym			{ consym }
}

384 385
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
386
<0> {
387
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
388 389 390
  @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 }
391 392

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
393
  @floating_point    { strtoken tok_float }
394 395
}

396
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
397
  -- Unboxed ints (:: Int#) and words (:: Word#)
398 399
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
400 401 402 403 404
  @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 }
405
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
406

Ian Lynagh's avatar
Ian Lynagh committed
407 408 409 410
  @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 }

411 412
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
413 414
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
415 416 417 418 419 420
}

-- 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.
421
<0> {
422 423 424 425 426 427 428 429
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

430
data Token
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
  = 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
469
  | ITfamily
470 471 472
  | ITgroup
  | ITby
  | ITusing
473

474 475 476 477
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
478 479
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
480
  | ITwarning_prag
481 482 483
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
484
  | ITgenerated_prag
485
  | ITcore_prag                 -- hdaume: core annotations
486
  | ITunpack_prag
487
  | ITclose_prag
488 489 490
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516

  | 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
517 518
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549
  | 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)

  | 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
550
  | ITprimword   Integer
551 552 553 554
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
555 556 557 558 559 560 561 562 563
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
564
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
565 566 567 568

  -- Arrow notation extension
  | ITproc
  | ITrec
569 570 571 572 573 574
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
575 576 577

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
578 579 580 581 582 583 584

  -- 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
585
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
586

587 588 589 590
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
591
{-
592
isSpecial :: Token -> Bool
593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608
-- 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
609
isSpecial ITfamily	= True
610 611 612
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
613
isSpecial _             = False
twanvl's avatar
twanvl committed
614
-}
615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651

-- 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)
--
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

652
    ( "forall",	ITforall,	 bit explicitForallBit),
Ian Lynagh's avatar
Ian Lynagh committed
653
	( "mdo",	ITmdo,		 bit recursiveDoBit),
654
	( "family",	ITfamily,	 bit tyFamBit),
655 656 657
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
658 659 660 661 662 663 664 665 666 667 668 669 670

	( "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),
671
	( "proc",	ITproc,		 bit arrowsBit)
672 673
     ]

Ian Lynagh's avatar
Ian Lynagh committed
674
reservedSymsFM :: UniqFM (Token, Int -> Bool)
675
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692
    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
693
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
694
        -- For 'forall a . t'
Ian Lynagh's avatar
Ian Lynagh committed
695
       ,(".", ITdot, explicitForallEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
696 697 698 699 700

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
701 702

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
703 704
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
705 706
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
707 708 709
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
710 711 712
        -- 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).
713
#endif
714 715 716 717 718
       ]

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

719
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
720

721
special :: Token -> Action
twanvl's avatar
twanvl committed
722
special tok span _buf _len = return (L span tok)
723

724
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
725 726
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
727

728 729
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
730

731 732 733
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
734

735 736 737
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
738

739
init_strtoken :: Int -> (String -> Token) -> Action
740
-- like strtoken, but drops the last N character(s)
741 742
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
743 744

begin :: Int -> Action
745
begin code _span _str _len = do pushLexState code; lexToken
746 747

pop :: Action
748
pop _span _buf _len = do popLexState; lexToken
749 750

pop_and :: Action -> Action
751
pop_and act span buf len = do popLexState; act span buf len
752

753 754 755 756 757
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
758

759
notFollowedBySymbol _ _ _ (AI _ _ buf)
760
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
761

762 763 764 765 766
-- 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.
767
isNormalComment bits _ _ (AI _ _ buf)
768 769 770 771 772 773 774
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
775

twanvl's avatar
twanvl committed
776
{-
777 778
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
779
-}
780

781
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
782

783 784
ifExtension pred bits _ _ _ = pred bits

785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812
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

813 814 815 816
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
817 818
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
819
  input <- getInput
820
  go (1::Int) input
821 822 823 824 825 826 827
  where
    go 0 input = do setInput input; cont
    go n input = case alexGetChar input of
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('\125',input) -> go (n-1) input
twanvl's avatar
twanvl committed
828
        Just (_,_)          -> go n input
829 830 831
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> go (n+1) input
twanvl's avatar
twanvl committed
832 833
        Just (_,_)       -> go n input
      Just (_,input) -> go n input
834 835 836 837 838 839 840 841

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
842
        Just ('\125',input) ->
843
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
844
        Just (_,_) -> go ('-':commentAcc) input docType False
845 846 847 848 849 850
      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
851
        Just (_,_) -> go ('\123':commentAcc) input docType False
852 853 854
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
855 856 857 858 859
  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
860 861
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
862 863 864
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
865
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

-- 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))
 
errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
 
902
open_brace, close_brace :: Action
903
open_brace span _str _len = do 
904 905
  ctx <- getContext
  setContext (NoLayout:ctx)
906 907
  return (L span ITocurly)
close_brace span _str _len = do 
908
  popContext
909
  return (L span ITccurly)
910 911 912 913 914 915 916 917

qvarid buf len = ITqvarid $! splitQualName buf len
qconid buf len = ITqconid $! splitQualName buf len

splitQualName :: StringBuffer -> Int -> (FastString,FastString)
-- 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.
918
splitQualName orig_buf len = split orig_buf orig_buf
919
  where
920 921 922 923 924 925
    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
926 927 928 929
  
    -- 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.
930 931 932 933 934 935 936 937 938 939 940
    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),
	 lexemeToFastString dot_buf (len - qual_size))
      where
	qual_size = orig_buf `byteDiff` dot_buf
941

942
varid span buf len = 
Ian Lynagh's avatar
Ian Lynagh committed
943
  fs `seq`
944 945 946
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
947
		return (L span keyword)
948 949 950
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
951 952 953
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
954 955 956 957 958 959 960 961 962 963 964 965
  where
	fs = lexemeToFastString buf len

conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

qvarsym buf len = ITqvarsym $! splitQualName buf len
qconsym buf len = ITqconsym $! splitQualName buf len

varsym = sym ITvarsym
consym = sym ITconsym

966
sym con span buf len = 
967 968
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
969
		b <- extension exts
970 971 972
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
973 974 975
  where
	fs = lexemeToFastString buf len

976 977 978 979 980 981 982 983 984 985 986 987 988
-- 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
tok_num = tok_integral ITinteger
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
989
tok_primword = tok_integral ITprimword positive
990 991 992 993 994 995 996
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
997
tok_float        str = ITrational   $! readRational str
998 999
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
1000 1001 1002 1003 1004 1005

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1006
do_bol span _str _len = do
1007
	pos <- getOffside
1008 1009 1010 1011 1012
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1013
		return (L span ITvccurly)
1014 1015 1016
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1017
		return (L span ITsemi)
1018 1019 1020 1021 1022 1023 1024
	    GT -> do
		popLexState
		lexToken

-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
maybe_layout ITdo	= pushLexState layout_do
1025
maybe_layout ITmdo	= pushLexState layout_do
1026 1027 1028
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1029
maybe_layout ITrec	= pushLexState layout
1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040
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.
--
1041
new_layout_context strict span _buf _len = do
1042
    popLexState
1043
    (AI _ offset _) <- getInput
1044 1045 1046 1047 1048 1049 1050 1051
    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
1052
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1053
	_ -> do
1054
		setContext (Layout offset : ctx)
1055
		return (L span ITvocurly)
1056

1057
do_layout_left span _buf _len = do
1058 1059
    popLexState
    pushLexState bol  -- we must be at the start of a line
1060
    return (L span ITvccurly)
1061 1062 1063 1064

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

1065 1066
setLine :: Int -> Action
setLine code span buf len = do
1067
  let line = parseUnsignedInteger buf len 10 octDecDigit
1068
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1069 1070 1071 1072 1073
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1074 1075
setFile :: Int -> Action
setFile code span buf len = do
1076
  let file = lexemeToFastString (stepOn buf) (len-2)
1077
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1078 1079 1080 1081
  popLexState
  pushLexState code
  lexToken

1082 1083 1084 1085 1086

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1087
lex_string_prag mkTok span _buf _len
1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099
    = 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
1100
          isString _ [] = True
1101 1102 1103 1104 1105 1106 1107
          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"


1108 1109 1110 1111 1112 1113
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1114
lex_string_tok span _buf _len = do
1115 1116
  tok <- lex_string ""
  end <- getSrcLoc 
1117
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1118

1119
lex_string :: String -> P Token
1120 1121
lex_string s = do
  i <- getInput
1122
  case alexGetChar' i of
1123 1124 1125 1126
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1127 1128
	magicHash <- extension magicHashEnabled
	if magicHash
1129 1130
	  then do
	    i <- getInput
1131
	    case alexGetChar' i of
1132 1133 1134 1135
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1136
                    else let s' = mkZFastString (reverse s) in
1137
			 return (ITprimstring s')
1138 1139
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1140 1141 1142 1143 1144 1145 1146 1147 1148 1149
	      _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
1150
	where next = alexGetChar' i
1151

1152 1153 1154
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1155 1156 1157 1158 1159 1160 1161 1162 1163 1164

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
1165 1166 1167 1168 1169 1170
-- 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
1171
lex_char_tok span _buf _len = do	-- We've seen '
1172
   i1 <- getInput	-- Look ahead to first character
1173
   let loc = srcSpanStart span
1174
   case alexGetChar' i1 of
1175 1176
	Nothing -> lit_error 

1177
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1178 1179 1180
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1181
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1182 1183
		   else lit_error

twanvl's avatar
twanvl committed
1184
	Just ('\\', i2@(AI _end2 _ _)) -> do 	-- We've seen 'backslash
1185 1186 1187 1188
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1189
			        else do setInput i2; lit_error 
1190

twanvl's avatar
twanvl committed
1191
        Just (c, i2@(AI _end2 _ _))
1192 1193
		| not (isAny c) -> lit_error
		| otherwise ->
1194 1195 1196

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
1197
	   case alexGetChar' i2 of	-- Look ahead one more character
1198 1199 1200 1201 1202 1203 1204
		Nothing -> lit_error
		Just ('\'', i3) -> do 	-- We've seen 'x'
			setInput i3 
			finish_char_tok loc c
		_other -> do 		-- We've seen 'x not followed by quote
					-- If TH is on, just parse the quote only
			th_exts <- extension thEnabled	
1205 1206
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1207
				   else do setInput i2; lit_error
1208

1209
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1210 1211
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
1212
  = do	magicHash <- extension magicHashEnabled
1213
	i@(AI end _ _) <- getInput
1214
	if magicHash then do
1215
		case alexGetChar' i of
1216
			Just ('#',i@(AI end _ _)) -> do
1217
				setInput i
1218
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1219
			_other ->
1220 1221 1222 1223
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1224 1225 1226 1227 1228
lex_char :: Char -> AlexInput -> P Char
lex_char c inp = do
  case c of
      '\\' -> do setInput inp; lex_escape
      c | isAny c -> do setInput inp; return c
1229
      _other -> lit_error
1230

1231
isAny c | c > '\x7f' = isPrint c
1232 1233
	| otherwise  = is_any c

1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249