Lexer.x 54.3 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 24
--
-- 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

{
module Lexer (
25
   Token(..), lexer, pragState, mkPState, PState(..),
26
   P(..), ParseResult(..), getSrcLoc, 
27
   failLocMsgP, failSpanMsgP, srcParseFail,
28
   getMessages,
29
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
30
   getLexState, popLexState, pushLexState,
31 32
   extension, standaloneDerivingEnabled, bangPatEnabled,
   addWarning
33 34 35 36
  ) where

#include "HsVersions.h"

37 38
import Bag
import ErrUtils
39 40 41 42 43 44
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
45
import DynFlags
46
import Ctype
47
import Util		( maybePrefixMatch, readRational )
48

49
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
50
import Data.Bits
51
import Data.Char 	( chr, isSpace )
Simon Marlow's avatar
Simon Marlow committed
52 53
import Data.Ratio
import Debug.Trace
54 55

#if __GLASGOW_HASKELL__ >= 605
56
import Data.Char 	( GeneralCategory(..), generalCategory, isPrint, isUpper )
57
#else
58
import Compat.Unicode	( GeneralCategory(..), generalCategory, isPrint, isUpper )
59
#endif
60 61
}

62
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
63
$whitechar   = [\ \n\r\f\v\xa0 $unispace]
64
$white_no_nl = $whitechar # \n
65
$tab         = \t
66 67

$ascdigit  = 0-9
68
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
69
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
70 71 72
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
73
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~ \xa1-\xbf \xd7 \xf7]
74
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
75 76
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

77
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
78 79 80
$asclarge  = [A-Z \xc0-\xd6 \xd8-\xde]
$large     = [$asclarge $unilarge]

81
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
82 83 84
$ascsmall  = [a-z \xdf-\xf6 \xf8-\xff]
$small     = [$ascsmall $unismall \_]

85
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
86
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
87 88

$octit	   = 0-7
89
$hexit     = [$decdigit A-F a-f]
90 91 92 93
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

94 95
$docsym    = [\| \^ \* \$]

96 97 98 99 100 101
@varid     = $small $idchar*
@conid     = $large $idchar*

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

102
@decimal     = $decdigit+
103 104 105 106 107 108 109 110 111
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

112 113 114 115 116
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

117 118 119 120
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
121
$tab+         { warn Opt_WarnTabs (text "Tab character") }
122 123 124 125 126

-- 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
127 128 129 130
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
131 132 133 134

-- 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.
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168

-- 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.

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

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

"-- " $docsym .* / { ifExtension (not . haddockEnabled) } ;

-- 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 } ;
169 170 171 172 173 174 175 176 177 178 179 180 181

-- '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 }
182
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
183
  ^\# \! .* \n				; -- #!, for scripts
184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205
  ()					{ 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 }

206
<0,option_prags> \n				{ begin bol }
207 208 209 210 211

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

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
212
<line_prag1> $decdigit+			{ setLine line_prag1a }
213
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
214 215 216 217
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
218
<line_prag2> $decdigit+			{ setLine line_prag2a }
219
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
220 221 222
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
223

224 225 226
-- 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.
227
-- (ToDo: we should really emit a warning when ignoring pragmas)
228 229
-- XXX Now that we can enable this without the -fglasgow-exts hammer,
-- is it better just to let the parse error happen?
230
<0>
231
  "{-#" $whitechar* (RULES|rules) / { ifExtension explicitForallEnabled } { token ITrules_prag }
232

233
<0,option_prags> {
234 235 236 237 238
  "{-#" $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 }
239
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
240 241 242 243
	$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) }
244 245 246 247
  "{-#" $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
248 249
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
250
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
251
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
252

253 254 255 256
  "{-#" $whitechar* (DOCOPTIONS|docoptions)
  / { ifExtension haddockEnabled }     { lex_string_prag ITdocOptions }

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

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

262 263 264 265 266 267 268
<option_prags> {
  "{-#" $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#" $whitechar* (OPTIONS_GHC|options_ghc)
                                        { lex_string_prag IToptions_prag }
  "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
  "{-#" $whitechar* (INCLUDE|include)   { lex_string_prag ITinclude_prag }
}
269

270
<0,option_prags> {
271
	-- This is to catch things like {-# OPTIONS OPTIONS_HUGS ... 
272
  "{-#" $whitechar* $idchar+		{ nested_comment lexToken }
273 274
}

275 276
-- '0' state: ordinary lexemes

277 278
-- Haddock comments

279
<0> {
280 281
  "-- " $docsym    / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
282 283
}

284 285
-- "special" symbols

286
<0> {
287 288 289 290
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
291
<0> {
292 293 294 295 296 297 298 299 300 301
  "[|"	    / { 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 }
}

302
<0> {
ross's avatar
ross committed
303 304
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
305 306 307
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

308
<0> {
309 310 311
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

312
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
313 314 315 316 317 318
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

319
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
320 321
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
322 323
}

324
<0,option_prags> {
325 326 327 328 329 330 331 332 333 334 335 336
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

337
<0,option_prags> {
338
  @qual @varid			{ idtoken qvarid }
339 340 341 342 343
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

344
<0> {
345 346 347 348
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
349 350 351 352
}

-- ToDo: M.(,,,)

353
<0> {
354 355 356 357 358 359
  @qual @varsym			{ idtoken qvarsym }
  @qual @consym			{ idtoken qconsym }
  @varsym			{ varsym }
  @consym			{ consym }
}

360 361
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
362
<0> {
363 364 365 366 367 368 369
  -- Normal integral literals (:: Num a => a, from Integer)
  @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 }

  -- Normal rational literals (:: Fractional a => a, from Rational)
  @floating_point		{ strtoken tok_float }
370 371
}

372
<0> {
373 374 375
  -- Unboxed ints (:: Int#)
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
376 377 378 379 380 381
  @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 }
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
382 383 384

  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
385 386
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
387 388 389 390 391 392
}

-- 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.
393
<0> {
394 395 396 397 398 399 400 401
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

402
data Token
403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440
  = 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
441
  | ITfamily
442

443 444 445 446
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
447 448 449 450 451
  | ITsource_prag
  | ITrules_prag
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
452
  | ITgenerated_prag
453
  | ITcore_prag                 -- hdaume: core annotations
454
  | ITunpack_prag
455
  | ITclose_prag
456 457 458
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521

  | 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
  | ITopabrack			-- [:, for parallel arrays with -fparr
  | ITcpabrack			-- :], for parallel arrays with -fparr
  | 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
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
522 523 524 525 526 527 528 529 530
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
531 532 533 534

  -- Arrow notation extension
  | ITproc
  | ITrec
535 536 537 538 539 540
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
541 542 543

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
544 545 546 547 548 549 550 551

  -- 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)

552 553 554 555
#ifdef DEBUG
  deriving Show -- debugging
#endif

556
isSpecial :: Token -> Bool
557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572
-- 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
573
isSpecial ITfamily	= True
574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611
isSpecial _             = False

-- 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

Ian Lynagh's avatar
Ian Lynagh committed
612
      	( "forall",	ITforall,	 bit explicitForallBit),
Ian Lynagh's avatar
Ian Lynagh committed
613
	( "mdo",	ITmdo,		 bit recursiveDoBit),
614
	( "family",	ITfamily,	 bit tyFamBit),
615 616 617 618 619 620 621 622 623 624 625 626 627

	( "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),
628
	( "proc",	ITproc,		 bit arrowsBit)
629 630
     ]

Ian Lynagh's avatar
Ian Lynagh committed
631
reservedSymsFM :: UniqFM (Token, Int -> Bool)
632
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649
    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
650
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
651
        -- For 'forall a . t'
Ian Lynagh's avatar
Ian Lynagh committed
652
       ,(".", ITdot, explicitForallEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
653 654 655 656 657

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
658 659

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
660 661
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
662 663
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
664 665 666
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
667 668 669
        -- 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).
670
#endif
671 672 673 674 675
       ]

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

676
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
677

678 679
special :: Token -> Action
special tok span _buf len = return (L span tok)
680

681 682 683
token, layout_token :: Token -> Action
token t span buf len = return (L span t)
layout_token t span buf len = pushLexState layout >> return (L span t)
684

685 686
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
687

688 689 690
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
691

692 693 694
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
695

696
init_strtoken :: Int -> (String -> Token) -> Action
697
-- like strtoken, but drops the last N character(s)
698 699
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
700 701

begin :: Int -> Action
702
begin code _span _str _len = do pushLexState code; lexToken
703 704

pop :: Action
705
pop _span _buf _len = do popLexState; lexToken
706 707

pop_and :: Action -> Action
708
pop_and act span buf len = do popLexState; act span buf len
709

710 711 712 713 714
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
715

716
notFollowedBySymbol _ _ _ (AI _ _ buf)
717
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
718

719 720 721 722 723
-- 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.
724
isNormalComment bits _ _ (AI _ _ buf)
725 726 727 728 729 730 731
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
732 733 734 735

haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)

736
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
737

738 739
ifExtension pred bits _ _ _ = pred bits

740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767
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

768 769 770 771
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
772 773
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
774
  input <- getInput
775
  go (1::Int) input
776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809
  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
        Just (c,_)          -> go n input
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> go (n+1) input
        Just (c,_)       -> go n input
      Just (c,input) -> go n input

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
        Just ('\125',input@(AI end _ buf2)) ->
          docCommentEnd input commentAcc docType buf span
        Just (c,_) -> go ('-':commentAcc) input docType False
      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
        Just (c,_) -> go ('\123':commentAcc) input docType False
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
810 811 812 813 814 815
  input@(AI _ _ buf) <- getInput
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
    '*' -> lexDocSection 1 input 
816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
      Just (c, _) -> lexDocComment input (ITdocSection n) True
      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 `{-'"
 
856
open_brace, close_brace :: Action
857
open_brace span _str _len = do 
858 859
  ctx <- getContext
  setContext (NoLayout:ctx)
860 861
  return (L span ITocurly)
close_brace span _str _len = do 
862
  popContext
863
  return (L span ITccurly)
864 865 866 867 868 869 870 871

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.
872
splitQualName orig_buf len = split orig_buf orig_buf
873
  where
874 875 876 877 878 879
    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
880 881 882 883
  
    -- 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.
884 885 886 887 888 889 890 891 892 893 894
    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
895

896
varid span buf len = 
897 898 899
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
900
		return (L span keyword)
901 902 903
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
904 905 906
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
907 908 909 910 911 912 913 914 915 916 917 918
  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

919
sym con span buf len = 
920 921
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
922
		b <- extension exts
923 924 925
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
926 927 928
  where
	fs = lexemeToFastString buf len

929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948
-- 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
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
949
tok_float        str = ITrational   $! readRational str
950 951
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
952 953 954 955 956 957

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
958
do_bol span _str _len = do
959
	pos <- getOffside
960 961 962 963 964
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
965
		return (L span ITvccurly)
966 967 968
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
969
		return (L span ITsemi)
970 971 972 973 974 975 976
	    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
977
maybe_layout ITmdo	= pushLexState layout_do
978 979 980
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
981
maybe_layout ITrec	= pushLexState layout
982 983 984 985 986 987 988 989 990 991 992
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.
--
993
new_layout_context strict span _buf _len = do
994
    popLexState
995
    (AI _ offset _) <- getInput
996 997 998 999 1000 1001 1002 1003
    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
1004
		return (L span ITvocurly)
1005 1006
	other -> do
		setContext (Layout offset : ctx)
1007
		return (L span ITvocurly)
1008

1009
do_layout_left span _buf _len = do
1010 1011
    popLexState
    pushLexState bol  -- we must be at the start of a line
1012
    return (L span ITvccurly)
1013 1014 1015 1016

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

1017 1018
setLine :: Int -> Action
setLine code span buf len = do
1019
  let line = parseUnsignedInteger buf len 10 octDecDigit
1020
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1021 1022 1023 1024 1025
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1026 1027
setFile :: Int -> Action
setFile code span buf len = do
1028
  let file = lexemeToFastString (stepOn buf) (len-2)
1029
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1030 1031 1032 1033
  popLexState
  pushLexState code
  lexToken

1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059

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

lex_string_prag :: (String -> Token) -> Action
lex_string_prag mkTok span buf len
    = 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
          isString i [] = True
          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"


1060 1061 1062 1063 1064 1065
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
1066
lex_string_tok span buf len = do
1067 1068
  tok <- lex_string ""
  end <- getSrcLoc 
1069
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1070

1071
lex_string :: String -> P Token
1072 1073
lex_string s = do
  i <- getInput
1074
  case alexGetChar' i of
1075 1076 1077 1078
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1079 1080
	magicHash <- extension magicHashEnabled
	if magicHash
1081 1082
	  then do
	    i <- getInput
1083
	    case alexGetChar' i of
1084 1085 1086 1087
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1088
                    else let s' = mkZFastString (reverse s) in
1089
			 return (ITprimstring s')
1090 1091
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1092 1093 1094 1095 1096 1097 1098 1099 1100 1101
	      _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
1102
	where next = alexGetChar' i
1103

1104 1105 1106
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116

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
1117 1118 1119 1120 1121 1122
-- 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
1123
lex_char_tok span buf len = do	-- We've seen '
1124
   i1 <- getInput	-- Look ahead to first character
1125
   let loc = srcSpanStart span
1126
   case alexGetChar' i1 of
1127 1128
	Nothing -> lit_error 

1129
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1130 1131 1132
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1133
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1134 1135
		   else lit_error

1136
	Just ('\\', i2@(AI end2 _ _)) -> do 	-- We've seen 'backslash 
1137 1138 1139 1140
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1141
			        else do setInput i2; lit_error 
1142

1143 1144 1145
        Just (c, i2@(AI end2 _ _)) 
		| not (isAny c) -> lit_error
		| otherwise ->
1146 1147 1148

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
1149
	   case alexGetChar' i2 of	-- Look ahead one more character
1150 1151 1152 1153 1154 1155 1156
		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	
1157 1158
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1159
				   else do setInput i2; lit_error
1160

1161
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1162 1163
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
1164
  = do	magicHash <- extension magicHashEnabled
1165
	i@(AI end _ _) <- getInput
1166
	if magicHash then do
1167
		case alexGetChar' i of
1168
			Just ('#',i@(AI end _ _)) -> do
1169
				setInput i
1170
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1171
			_other ->
1172 1173 1174 1175
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1176 1177 1178 1179 1180
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
1181
      _other -> lit_error
1182

1183 1184 1185
isAny c | c > '\xff' = isPrint c
	| otherwise  = is_any c

1186 1187 1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
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

1205 1206 1207
	'x'   -> readNum is_hexdigit 16 hexDigit
	'o'   -> readNum is_octdigit  8 octDecDigit
	x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x)
1208 1209 1210

	c1 ->  do
	   i <- getInput
1211
	   case alexGetChar' i of
1212 1213
	    Nothing -> lit_error
	    Just (c2,i2) -> 
1214 1215
              case alexGetChar' i2 of
		Nothing	-> do setInput i2; lit_error
1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227 1228 1229
		Just (c3,i3) -> 
		   let str = [c1,c2,c3] in
		   case [ (c,rest) | (p,c) <- silly_escape_chars,
			      	     Just rest <- [maybePrefixMatch p str] ] of
			  (escape_char,[]):_ -> do
				setInput i3
				return escape_char
			  (escape_char,_:_):_ -> do
				setInput i2
				return escape_char
			  [] -> lit_error

readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
readNum is_digit base conv = do
1230
  i <- getInput
1231 1232 1233
  c <- getCharOrFail
  if is_digit c 
	then readNum2 is_digit base conv (conv c)
1234
	else do setInput i; lit_error
1235 1236 1237 1238 1239

readNum2 is_digit base conv i = do
  input <- getInput
  read i input
  where read i input = do
1240
	  case alexGetChar' input of
1241 1242 1243 1244
	    Just (c,input') | is_digit c -> do
		read (i*base + conv c) input'
	    _other -> do
		if i >= 0 && i <= 0x10FFFF
1245
		   then do setInput input; return (chr i)
1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262 1263 1264 1265 1266 1267 1268 1269 1270 1271 1272 1273 1274 1275 1276 1277 1278 1279 1280 1281 1282 1283 1284
		   else lit_error

silly_escape_chars = [
	("NUL", '\NUL'),
	("SOH", '\SOH'),
	("STX", '\STX'),
	("ETX", '\ETX'),
	("EOT", '\EOT'),
	("ENQ", '\ENQ'),
	("ACK", '\ACK'),
	("BEL", '\BEL'),
	("BS", '\BS'),
	("HT", '\HT'),
	("LF", '\LF'),
	("VT", '\VT'),
	("FF", '\FF'),
	("CR", '\CR'),
	("SO", '\SO'),
	("SI", '\SI'),
	("DLE", '\DLE'),
	("DC1", '\DC1'),
	("DC2", '\DC2'),
	("DC3", '\DC3'),
	("DC4", '\DC4'),
	("NAK", '\NAK'),
	("SYN", '\SYN'),
	("ETB", '\ETB'),
	("CAN", '\CAN'),
	("EM", '\EM'),
	("SUB", '\SUB'),
	("ESC", '\ESC'),
	("FS", '\FS'),
	("GS", '\GS'),
	("RS", '\RS'),
	("US", '\US'),
	("SP", '\SP'),
	("DEL", '\DEL')
	]

1285 1286 1287 1288
-- before calling lit_error, ensure that the current input is pointing to
-- the position of the error in the buffer.  This is so that we can report
-- a correct location to the user, but also so we can detect UTF-8 decoding
-- errors if they occur.
1289 1290 1291 1292 1293
lit_error = lexError "lexical error in string/character literal"

getCharOrFail :: P Char
getCharOrFail =  do
  i <- getInput
1294
  case alexGetChar' i of
1295 1296 1297
	Nothing -> lexError "unexpected end-of-file in string/character literal"
	Just (c,i)  -> do setInput i; return c

1298 1299 1300 1301
-- -----------------------------------------------------------------------------
-- Warnings

warn :: DynFlag -> SDoc -> Action
1302 1303
warn option warning srcspan _buf _len = do
    addWarning option srcspan warning
1304 1305
    lexToken

1306 1307 1308 1309 1310 1311
-- -----------------------------------------------------------------------------
-- The Parse Monad

data LayoutContext
  = NoLayout
  | Layout !Int
1312
  deriving Show
1313 1314 1315 1316

data ParseResult a
  = POk PState a
  | PFailed 
1317
	SrcSpan		-- The start and end of the text span related to
1318 1319
			-- the error.  Might be used in environments which can 
			-- show this span, e.g. by highlighting it.
1320 1321 1322 1323
	Message		-- The error message

data PState = PState { 
	buffer	   :: StringBuffer,
1324 1325
    dflags     :: DynFlags,
    messages   :: Messages,
1326
        last_loc   :: SrcSpan,	-- pos of previous token
1327 1328 1329
        last_offs  :: !Int, 	-- offset of the previous token from the
				-- beginning of  the current line.
				-- \t is equal to 8 spaces.
1330
	last_len   :: !Int,	-- len of previous token
1331
  last_line_len :: !Int,
1332 1333 1334 1335 1336
        loc        :: SrcLoc,   -- current loc (end of prev token + 1)
	extsBitmap :: !Int,	-- bitmap that determines permitted extensions
	context	   :: [LayoutContext],
	lex_state  :: [Int]
     }
1337
	-- last_loc and last_len are used when generating error messages,
1338 1339 1340 1341
	-- and in pushCurrentContext only.  Sigh, if only Happy passed the
	-- current token to happyError, we could at least get rid of last_len.
	-- Getting rid of last_loc would require finding another way to 
	-- implement pushCurrentContext (which is only called from one place).
1342 1343 1344 1345 1346 1347 1348 1349 1350 1351 1352 1353 1354 1355

newtype P a = P { unP :: PState -> ParseResult a }

instance Monad P where
  return = returnP
  (>>=) = thenP
  fail = failP

returnP :: a -> P a
returnP a = P $ \s -> POk s a

thenP :: P a -> (a -> P b) -> P b
(P m) `thenP` k = P $ \ s ->
	case m s of
1356 1357
		POk s1 a         -> (unP (k a)) s1
		PFailed span err -> PFailed span err
1358 1359

failP :: String -> P a
1360
failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1361 1362

failMsgP :: String -> P a
1363
failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1364 1365

failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1366 1367 1368 1369
failLocMsgP loc1 loc2 str = P $ \s -> PFailed (mkSrcSpan loc1 loc2) (text str)

failSpanMsgP :: SrcSpan -> String -> P a
failSpanMsgP span msg = P $ \s -> PFailed span (text msg)