Lexer.x 57 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
module Lexer (
35
   Token(..), lexer, pragState, mkPState, PState(..),
36
   P(..), ParseResult(..), getSrcLoc, 
37
   failLocMsgP, failSpanMsgP, srcParseFail,
38
   getMessages,
39
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
40
   getLexState, popLexState, pushLexState,
41 42
   extension, standaloneDerivingEnabled, bangPatEnabled,
   addWarning
43 44
  ) where

45 46
import Bag
import ErrUtils
47 48 49 50 51 52
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
53
import DynFlags
54
import Ctype
55
import Util		( maybePrefixMatch, readRational )
56

57
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
58
import Data.Bits
59
import Data.Char 	( chr, ord, isSpace )
Simon Marlow's avatar
Simon Marlow committed
60 61
import Data.Ratio
import Debug.Trace
62

Ian Lynagh's avatar
Ian Lynagh committed
63
import Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper )
64 65
}

66
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
67
$whitechar   = [\ \n\r\f\v $unispace]
68
$white_no_nl = $whitechar # \n
69
$tab         = \t
70 71

$ascdigit  = 0-9
72
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
73
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
74 75 76
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
77
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
78
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
79 80
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

81
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
82
$asclarge  = [A-Z]
83 84
$large     = [$asclarge $unilarge]

85
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
86
$ascsmall  = [a-z]
87 88
$small     = [$ascsmall $unismall \_]

89
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
90
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
91 92

$octit	   = 0-7
93
$hexit     = [$decdigit A-F a-f]
94 95 96 97
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

98 99
$docsym    = [\| \^ \* \$]

100 101 102 103 104 105
@varid     = $small $idchar*
@conid     = $large $idchar*

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

106
@decimal     = $decdigit+
107 108 109 110 111 112 113 114 115
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

116 117 118 119 120
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

121 122 123 124
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
125
$tab+         { warn Opt_WarnTabs (text "Tab character") }
126 127 128 129 130

-- 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
131 132 133 134
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
135 136 137 138

-- 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.
139 140 141 142 143 144 145 146 147 148

-- 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
149
"-- " ~[$docsym \#] .* ;
150 151 152 153
"--" [^$symbol : \ ] .* ;

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

154
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } ;
155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172

-- 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 } ;
173 174 175 176 177 178 179 180 181 182 183 184 185

-- '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 }
186
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
187
  ^\# \! .* \n				; -- #!, for scripts
188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
  ()					{ 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 }

210
<0,option_prags> \n				{ begin bol }
211 212 213 214 215

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

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
216
<line_prag1> $decdigit+			{ setLine line_prag1a }
217
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
218 219 220 221
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
222
<line_prag2> $decdigit+			{ setLine line_prag2a }
223
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
224 225 226
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
227

228 229 230
-- 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.
231
-- (ToDo: we should really emit a warning when ignoring pragmas)
232 233
-- XXX Now that we can enable this without the -fglasgow-exts hammer,
-- is it better just to let the parse error happen?
234
<0>
235
  "{-#" $whitechar* (RULES|rules) / { ifExtension explicitForallEnabled } { token ITrules_prag }
236

237
<0,option_prags> {
238 239 240 241 242
  "{-#" $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 }
243
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
244 245 246 247
	$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) }
248 249 250 251
  "{-#" $whitechar* (SOURCE|source)	{ token ITsource_prag }
  "{-#" $whitechar* (DEPRECATED|deprecated)
  					{ token ITdeprecated_prag }
  "{-#" $whitechar* (SCC|scc)		{ token ITscc_prag }
andy@galois.com's avatar
andy@galois.com committed
252 253
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
254
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
255
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
256

257
 "{-#"                                 { nested_comment lexToken }
258 259 260 261 262

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

263
<option_prags> {
David Waern's avatar
David Waern committed
264 265
  "{-#"  $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc)
266
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
267 268 269 270 271 272 273 274 275
  "{-#"  $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 }
}

<0> {
  "-- #" .* ;
276
}
277

278
<0,option_prags> {
279
	-- This is to catch things like {-# OPTIONS OPTIONS_HUGS ... 
280
  "{-#" $whitechar* $idchar+		{ nested_comment lexToken }
281 282
}

283 284
-- '0' state: ordinary lexemes

285 286
-- Haddock comments

287
<0> {
David Waern's avatar
David Waern committed
288 289
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
290 291
}

292 293
-- "special" symbols

294
<0> {
295 296 297 298
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
299
<0> {
300 301 302 303 304 305 306 307
  "[|"	    / { 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 }
308 309 310

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

313
<0> {
ross's avatar
ross committed
314 315
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
316 317 318
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

319
<0> {
320 321 322
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

323
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
324 325 326 327 328 329
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

330
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
331 332
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
333 334
}

335
<0,option_prags> {
336 337 338 339 340 341 342 343 344 345 346 347
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

348
<0,option_prags> {
349
  @qual @varid			{ idtoken qvarid }
350 351 352 353 354
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

355
<0> {
356 357 358 359
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
360 361 362 363
}

-- ToDo: M.(,,,)

364
<0> {
365 366 367 368 369 370
  @qual @varsym			{ idtoken qvarsym }
  @qual @consym			{ idtoken qconsym }
  @varsym			{ varsym }
  @consym			{ consym }
}

371 372
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
373
<0> {
374
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
375 376 377
  @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 }
378 379

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
380
  @floating_point    { strtoken tok_float }
381 382
}

383
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
384
  -- Unboxed ints (:: Int#) and words (:: Word#)
385 386
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
387 388 389 390 391
  @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 }
392
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
393

Ian Lynagh's avatar
Ian Lynagh committed
394 395 396 397
  @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 }

398 399
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
400 401
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
402 403 404 405 406 407
}

-- 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.
408
<0> {
409 410 411 412 413 414 415 416
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

417
data Token
418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455
  = 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
456
  | ITfamily
457 458 459
  | ITgroup
  | ITby
  | ITusing
460

461 462 463 464
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
465 466 467 468 469
  | ITsource_prag
  | ITrules_prag
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
470
  | ITgenerated_prag
471
  | ITcore_prag                 -- hdaume: core annotations
472
  | ITunpack_prag
473
  | ITclose_prag
474 475 476
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502

  | 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
503 504
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535
  | 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
536
  | ITprimword   Integer
537 538 539 540
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
541 542 543 544 545 546 547 548 549
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
550
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
551 552 553 554

  -- Arrow notation extension
  | ITproc
  | ITrec
555 556 557 558 559 560
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
561 562 563

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
564 565 566 567 568 569 570

  -- 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
571
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
572

573 574 575 576
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
577
{-
578
isSpecial :: Token -> Bool
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
-- 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
595
isSpecial ITfamily	= True
596 597 598
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
599
isSpecial _             = False
twanvl's avatar
twanvl committed
600
-}
601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637

-- 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

638
    ( "forall",	ITforall,	 bit explicitForallBit),
Ian Lynagh's avatar
Ian Lynagh committed
639
	( "mdo",	ITmdo,		 bit recursiveDoBit),
640
	( "family",	ITfamily,	 bit tyFamBit),
641 642 643
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
644 645 646 647 648 649 650 651 652 653 654 655 656

	( "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),
657
	( "proc",	ITproc,		 bit arrowsBit)
658 659
     ]

Ian Lynagh's avatar
Ian Lynagh committed
660
reservedSymsFM :: UniqFM (Token, Int -> Bool)
661
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678
    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
679
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
680
        -- For 'forall a . t'
Ian Lynagh's avatar
Ian Lynagh committed
681
       ,(".", ITdot, explicitForallEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
682 683 684 685 686

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
687 688

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
689 690
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
691 692
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
693 694 695
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
696 697 698
        -- 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).
699
#endif
700 701 702 703 704
       ]

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

705
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
706

707
special :: Token -> Action
twanvl's avatar
twanvl committed
708
special tok span _buf _len = return (L span tok)
709

710
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
711 712
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
713

714 715
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
716

717 718 719
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
720

721 722 723
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
724

725
init_strtoken :: Int -> (String -> Token) -> Action
726
-- like strtoken, but drops the last N character(s)
727 728
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
729 730

begin :: Int -> Action
731
begin code _span _str _len = do pushLexState code; lexToken
732 733

pop :: Action
734
pop _span _buf _len = do popLexState; lexToken
735 736

pop_and :: Action -> Action
737
pop_and act span buf len = do popLexState; act span buf len
738

739 740 741 742 743
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
744

745
notFollowedBySymbol _ _ _ (AI _ _ buf)
746
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
747

748 749 750 751 752
-- 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.
753
isNormalComment bits _ _ (AI _ _ buf)
754 755 756 757 758 759 760
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
761

twanvl's avatar
twanvl committed
762
{-
763 764
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
765
-}
766

767
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
768

769 770
ifExtension pred bits _ _ _ = pred bits

771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798
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

799 800 801 802
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
803 804
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
805
  input <- getInput
806
  go (1::Int) input
807 808 809 810 811 812 813
  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
814
        Just (_,_)          -> go n input
815 816 817
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> go (n+1) input
twanvl's avatar
twanvl committed
818 819
        Just (_,_)       -> go n input
      Just (_,input) -> go n input
820 821 822 823 824 825 826 827

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
828
        Just ('\125',input) ->
829
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
830
        Just (_,_) -> go ('-':commentAcc) input docType False
831 832 833 834 835 836
      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
837
        Just (_,_) -> go ('\123':commentAcc) input docType False
838 839 840
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
841 842 843 844 845
  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
846 847
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
848 849 850
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
851
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887
      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 `{-'"
 
888
open_brace, close_brace :: Action
889
open_brace span _str _len = do 
890 891
  ctx <- getContext
  setContext (NoLayout:ctx)
892 893
  return (L span ITocurly)
close_brace span _str _len = do 
894
  popContext
895
  return (L span ITccurly)
896 897 898 899 900 901 902 903

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.
904
splitQualName orig_buf len = split orig_buf orig_buf
905
  where
906 907 908 909 910 911
    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
912 913 914 915
  
    -- 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.
916 917 918 919 920 921 922 923 924 925 926
    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
927

928
varid span buf len = 
Ian Lynagh's avatar
Ian Lynagh committed
929
  fs `seq`
930 931 932
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
933
		return (L span keyword)
934 935 936
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
937 938 939
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
940 941 942 943 944 945 946 947 948 949 950 951
  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

952
sym con span buf len = 
953 954
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
955
		b <- extension exts
956 957 958
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
959 960 961
  where
	fs = lexemeToFastString buf len

962 963 964 965 966 967 968 969 970 971 972 973 974
-- 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
975
tok_primword = tok_integral ITprimword positive
976 977 978 979 980 981 982
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
983
tok_float        str = ITrational   $! readRational str
984 985
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
986 987 988 989 990 991

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
992
do_bol span _str _len = do
993
	pos <- getOffside
994 995 996 997 998
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
999
		return (L span ITvccurly)
1000 1001 1002
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1003
		return (L span ITsemi)
1004 1005 1006 1007 1008 1009 1010
	    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
1011
maybe_layout ITmdo	= pushLexState layout_do
1012 1013 1014
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1015
maybe_layout ITrec	= pushLexState layout
1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026
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.
--
1027
new_layout_context strict span _buf _len = do
1028
    popLexState
1029
    (AI _ offset _) <- getInput
1030 1031 1032 1033 1034 1035 1036 1037
    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
1038
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1039
	_ -> do
1040
		setContext (Layout offset : ctx)
1041
		return (L span ITvocurly)
1042

1043
do_layout_left span _buf _len = do
1044 1045
    popLexState
    pushLexState bol  -- we must be at the start of a line
1046
    return (L span ITvccurly)
1047 1048 1049 1050

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

1051 1052
setLine :: Int -> Action
setLine code span buf len = do
1053
  let line = parseUnsignedInteger buf len 10 octDecDigit
1054
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1055 1056 1057 1058 1059
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1060 1061
setFile :: Int -> Action
setFile code span buf len = do
1062
  let file = lexemeToFastString (stepOn buf) (len-2)
1063
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1064 1065 1066 1067
  popLexState
  pushLexState code
  lexToken

1068 1069 1070 1071 1072

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1073
lex_string_prag mkTok span _buf _len
1074 1075 1076 1077 1078 1079 1080 1081 1082 1083 1084 1085
    = 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
1086
          isString _ [] = True
1087 1088 1089 1090 1091 1092 1093
          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"


1094 1095 1096 1097 1098 1099
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1100
lex_string_tok span _buf _len = do
1101 1102
  tok <- lex_string ""
  end <- getSrcLoc 
1103
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1104

1105
lex_string :: String -> P Token
1106 1107
lex_string s = do
  i <- getInput
1108
  case alexGetChar' i of
1109 1110 1111 1112
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1113 1114
	magicHash <- extension magicHashEnabled
	if magicHash
1115 1116
	  then do
	    i <- getInput
1117
	    case alexGetChar' i of
1118 1119 1120 1121
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1122
                    else let s' = mkZFastString (reverse s) in
1123
			 return (ITprimstring s')
1124 1125
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1126 1127 1128 1129 1130 1131 1132 1133 1134 1135
	      _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
1136
	where next = alexGetChar' i
1137

1138 1139 1140
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150

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
1151 1152 1153 1154 1155 1156
-- 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
1157
lex_char_tok span _buf _len = do	-- We've seen '
1158
   i1 <- getInput	-- Look ahead to first character
1159
   let loc = srcSpanStart span
1160
   case alexGetChar' i1 of
1161 1162
	Nothing -> lit_error 

1163
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1164 1165 1166
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1167
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1168 1169
		   else lit_error

twanvl's avatar
twanvl committed
1170
	Just ('\\', i2@(AI _end2 _ _)) -> do 	-- We've seen 'backslash
1171 1172 1173 1174
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1175
			        else do setInput i2; lit_error 
1176

twanvl's avatar
twanvl committed
1177
        Just (c, i2@(AI _end2 _ _))
1178 1179
		| not (isAny c) -> lit_error
		| otherwise ->
1180 1181 1182

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
1183
	   case alexGetChar' i2 of	-- Look ahead one more character
1184 1185 1186 1187 1188 1189 1190
		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	
1191 1192
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1193
				   else do setInput i2; lit_error
1194

1195
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1196 1197
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
1198
  = do	magicHash <- extension magicHashEnabled
1199
	i@(AI end _ _) <- getInput
1200
	if magicHash then do
1201
		case alexGetChar' i of
1202
			Just ('#',i@(AI end _ _)) -> do
1203
				setInput i
1204
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1205
			_other ->
1206 1207 1208 1209
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1210 1211 1212 1213 1214
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
1215
      _other -> lit_error
1216

1217
isAny c | c > '\x7f' = isPrint c
1218 1219
	| otherwise  = is_any c

1220 1221 1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238
lex_escape :: P Char
lex_escape = do
  c <- getCharOrFail
  case c of
	'a'   -> return '\a'
	'b'   -> return '\b'
	'f'   -> return '\f'
	'n'   -> return '\n'
	'r'   -> return '\r'
	't'   -> return '\t'
	'v'   -> return '\v'
	'\\'  -> return '\\'
	'"'   -> return '\"'
	'\''  -> return '\''
	'^'   -> do c <- getCharOrFail
		    if c >= '@' && c <= '_'
			then return (chr (ord c - ord '@'))
			else lit_error

1239 1240
	'x'   -> readNum is_hexdigit 16 hexDigit
	'o'   -> readNum is_octdigit  8 octDecDigit
1241
	x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x)
1242 1243 1244

	c1 ->  do
	   i <- getInput
1245
	   case alexGetChar' i of
1246 1247
	    Nothing -> lit_error
	    Just (c2,i2) ->