Lexer.x 54.8 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 30
-- for details

31
module Lexer (
32
   Token(..), lexer, pragState, mkPState, PState(..),
33
   P(..), ParseResult(..), getSrcLoc, 
34
   failLocMsgP, failSpanMsgP, srcParseFail,
35
   getMessages,
36
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
37
   getLexState, popLexState, pushLexState,
38 39
   extension, standaloneDerivingEnabled, bangPatEnabled,
   addWarning
40 41 42 43
  ) where

#include "HsVersions.h"

44 45
import Bag
import ErrUtils
46 47 48 49 50 51
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
52
import DynFlags
53
import Ctype
54
import Util		( maybePrefixMatch, readRational )
55

56
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
57
import Data.Bits
58
import Data.Char 	( chr, isSpace )
Simon Marlow's avatar
Simon Marlow committed
59 60
import Data.Ratio
import Debug.Trace
61 62

#if __GLASGOW_HASKELL__ >= 605
63
import Data.Char 	( GeneralCategory(..), generalCategory, isPrint, isUpper )
64
#else
65
import Compat.Unicode	( GeneralCategory(..), generalCategory, isPrint, isUpper )
66
#endif
67 68
}

69
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
70
$whitechar   = [\ \n\r\f\v\xa0 $unispace]
71
$white_no_nl = $whitechar # \n
72
$tab         = \t
73 74

$ascdigit  = 0-9
75
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
76
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
77 78 79
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
80
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~ \xa1-\xbf \xd7 \xf7]
81
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
82 83
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

84
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
85 86 87
$asclarge  = [A-Z \xc0-\xd6 \xd8-\xde]
$large     = [$asclarge $unilarge]

88
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
89 90 91
$ascsmall  = [a-z \xdf-\xf6 \xf8-\xff]
$small     = [$ascsmall $unismall \_]

92
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
93
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
94 95

$octit	   = 0-7
96
$hexit     = [$decdigit A-F a-f]
97 98 99 100
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

101 102
$docsym    = [\| \^ \* \$]

103 104 105 106 107 108
@varid     = $small $idchar*
@conid     = $large $idchar*

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

109
@decimal     = $decdigit+
110 111 112 113 114 115 116 117 118
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

119 120 121 122 123
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

124 125 126 127
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
128
$tab+         { warn Opt_WarnTabs (text "Tab character") }
129 130 131 132 133

-- 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
134 135 136 137
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
138 139 140 141

-- 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.
142 143 144 145 146 147 148 149 150 151

-- 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
152
"-- " ~[$docsym \#] .* ;
153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175
"--" [^$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 } ;
176 177 178 179 180 181 182 183 184 185 186 187 188

-- '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 }
189
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
190
  ^\# \! .* \n				; -- #!, for scripts
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212
  ()					{ 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 }

213
<0,option_prags> \n				{ begin bol }
214 215 216 217 218

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

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
219
<line_prag1> $decdigit+			{ setLine line_prag1a }
220
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
221 222 223 224
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
225
<line_prag2> $decdigit+			{ setLine line_prag2a }
226
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
227 228 229
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
230

231 232 233
-- 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.
234
-- (ToDo: we should really emit a warning when ignoring pragmas)
235 236
-- XXX Now that we can enable this without the -fglasgow-exts hammer,
-- is it better just to let the parse error happen?
237
<0>
238
  "{-#" $whitechar* (RULES|rules) / { ifExtension explicitForallEnabled } { token ITrules_prag }
239

240
<0,option_prags> {
241 242 243 244 245
  "{-#" $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 }
246
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
247 248 249 250
	$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) }
251 252 253 254
  "{-#" $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
255 256
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
257
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
258
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
259

260
 "{-#"                                 { nested_comment lexToken }
261 262 263 264 265

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

266
<option_prags> {
David Waern's avatar
David Waern committed
267 268
  "{-#"  $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc)
269
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
270 271 272 273 274 275 276 277 278
  "{-#"  $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> {
  "-- #" .* ;
279
}
280

281
<0,option_prags> {
282
	-- This is to catch things like {-# OPTIONS OPTIONS_HUGS ... 
283
  "{-#" $whitechar* $idchar+		{ nested_comment lexToken }
284 285
}

286 287
-- '0' state: ordinary lexemes

288 289
-- Haddock comments

290
<0> {
David Waern's avatar
David Waern committed
291 292
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
293 294
}

295 296
-- "special" symbols

297
<0> {
298 299 300 301
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
302
<0> {
303 304 305 306 307 308 309 310 311 312
  "[|"	    / { 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 }
}

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> {
374 375 376 377 378 379 380
  -- 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 }
381 382
}

383
<0> {
384 385 386
  -- Unboxed ints (:: Int#)
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
387 388 389 390 391 392
  @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 }
393 394 395

  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
396 397
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
398 399 400 401 402 403
}

-- 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.
404
<0> {
405 406 407 408 409 410 411 412
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

413
data Token
414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
  = 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
452
  | ITfamily
453

454 455 456 457
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
458 459 460 461 462
  | ITsource_prag
  | ITrules_prag
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
463
  | ITgenerated_prag
464
  | ITcore_prag                 -- hdaume: core annotations
465
  | ITunpack_prag
466
  | ITclose_prag
467 468 469
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532

  | 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
533 534 535 536 537 538 539 540 541
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
542 543 544 545

  -- Arrow notation extension
  | ITproc
  | ITrec
546 547 548 549 550 551
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
552 553 554

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
555 556 557 558 559 560 561

  -- 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
562
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
563

564 565 566 567
#ifdef DEBUG
  deriving Show -- debugging
#endif

568
isSpecial :: Token -> Bool
569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584
-- 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
585
isSpecial ITfamily	= True
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623
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
624
      	( "forall",	ITforall,	 bit explicitForallBit),
Ian Lynagh's avatar
Ian Lynagh committed
625
	( "mdo",	ITmdo,		 bit recursiveDoBit),
626
	( "family",	ITfamily,	 bit tyFamBit),
627 628 629 630 631 632 633 634 635 636 637 638 639

	( "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),
640
	( "proc",	ITproc,		 bit arrowsBit)
641 642
     ]

Ian Lynagh's avatar
Ian Lynagh committed
643
reservedSymsFM :: UniqFM (Token, Int -> Bool)
644
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661
    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
662
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
663
        -- For 'forall a . t'
Ian Lynagh's avatar
Ian Lynagh committed
664
       ,(".", ITdot, explicitForallEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
665 666 667 668 669

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
670 671

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
672 673
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
674 675
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
676 677 678
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
679 680 681
        -- 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).
682
#endif
683 684 685 686 687
       ]

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

688
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
689

690 691
special :: Token -> Action
special tok span _buf len = return (L span tok)
692

693 694 695
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)
696

697 698
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
699

700 701 702
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
703

704 705 706
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
707

708
init_strtoken :: Int -> (String -> Token) -> Action
709
-- like strtoken, but drops the last N character(s)
710 711
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
712 713

begin :: Int -> Action
714
begin code _span _str _len = do pushLexState code; lexToken
715 716

pop :: Action
717
pop _span _buf _len = do popLexState; lexToken
718 719

pop_and :: Action -> Action
720
pop_and act span buf len = do popLexState; act span buf len
721

722 723 724 725 726
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
727

728
notFollowedBySymbol _ _ _ (AI _ _ buf)
729
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
730

731 732 733 734 735
-- 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.
736
isNormalComment bits _ _ (AI _ _ buf)
737 738 739 740 741 742 743
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
744 745 746 747

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

748
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
749

750 751
ifExtension pred bits _ _ _ = pred bits

752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779
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

780 781 782 783
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
784 785
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
786
  input <- getInput
787
  go (1::Int) input
788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821
  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
822 823 824 825 826
  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
827 828
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868
 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 `{-'"
 
869
open_brace, close_brace :: Action
870
open_brace span _str _len = do 
871 872
  ctx <- getContext
  setContext (NoLayout:ctx)
873 874
  return (L span ITocurly)
close_brace span _str _len = do 
875
  popContext
876
  return (L span ITccurly)
877 878 879 880 881 882 883 884

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.
885
splitQualName orig_buf len = split orig_buf orig_buf
886
  where
887 888 889 890 891 892
    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
893 894 895 896
  
    -- 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.
897 898 899 900 901 902 903 904 905 906 907
    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
908

909
varid span buf len = 
910 911 912
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
913
		return (L span keyword)
914 915 916
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
917 918 919
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
920 921 922 923 924 925 926 927 928 929 930 931
  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

932
sym con span buf len = 
933 934
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
935
		b <- extension exts
936 937 938
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
939 940 941
  where
	fs = lexemeToFastString buf len

942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961
-- 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.
962
tok_float        str = ITrational   $! readRational str
963 964
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
965 966 967 968 969 970

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
971
do_bol span _str _len = do
972
	pos <- getOffside
973 974 975 976 977
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
978
		return (L span ITvccurly)
979 980 981
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
982
		return (L span ITsemi)
983 984 985 986 987 988 989
	    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
990
maybe_layout ITmdo	= pushLexState layout_do
991 992 993
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
994
maybe_layout ITrec	= pushLexState layout
995 996 997 998 999 1000 1001 1002 1003 1004 1005
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.
--
1006
new_layout_context strict span _buf _len = do
1007
    popLexState
1008
    (AI _ offset _) <- getInput
1009 1010 1011 1012 1013 1014 1015 1016
    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
1017
		return (L span ITvocurly)
1018 1019
	other -> do
		setContext (Layout offset : ctx)
1020
		return (L span ITvocurly)
1021

1022
do_layout_left span _buf _len = do
1023 1024
    popLexState
    pushLexState bol  -- we must be at the start of a line
1025
    return (L span ITvccurly)
1026 1027 1028 1029

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

1030 1031
setLine :: Int -> Action
setLine code span buf len = do
1032
  let line = parseUnsignedInteger buf len 10 octDecDigit
1033
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1034 1035 1036 1037 1038
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1039 1040
setFile :: Int -> Action
setFile code span buf len = do
1041
  let file = lexemeToFastString (stepOn buf) (len-2)
1042
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1043 1044 1045 1046
  popLexState
  pushLexState code
  lexToken

1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072

-- -----------------------------------------------------------------------------
-- 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"


1073 1074 1075 1076 1077 1078
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
1079
lex_string_tok span buf len = do
1080 1081
  tok <- lex_string ""
  end <- getSrcLoc 
1082
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1083

1084
lex_string :: String -> P Token
1085 1086
lex_string s = do
  i <- getInput
1087
  case alexGetChar' i of
1088 1089 1090 1091
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1092 1093
	magicHash <- extension magicHashEnabled
	if magicHash
1094 1095
	  then do
	    i <- getInput
1096
	    case alexGetChar' i of
1097 1098 1099 1100
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1101
                    else let s' = mkZFastString (reverse s) in
1102
			 return (ITprimstring s')
1103 1104
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
	      _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
1115
	where next = alexGetChar' i
1116

1117 1118 1119
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1120 1121 1122 1123 1124 1125 1126 1127 1128 1129

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
1130 1131 1132 1133 1134 1135
-- 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
1136
lex_char_tok span buf len = do	-- We've seen '
1137
   i1 <- getInput	-- Look ahead to first character
1138
   let loc = srcSpanStart span
1139
   case alexGetChar' i1 of
1140 1141
	Nothing -> lit_error 

1142
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1143 1144 1145
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1146
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1147 1148
		   else lit_error

1149
	Just ('\\', i2@(AI end2 _ _)) -> do 	-- We've seen 'backslash 
1150 1151 1152 1153
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1154
			        else do setInput i2; lit_error 
1155

1156 1157 1158
        Just (c, i2@(AI end2 _ _)) 
		| not (isAny c) -> lit_error
		| otherwise ->
1159 1160 1161

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
1162
	   case alexGetChar' i2 of	-- Look ahead one more character
1163 1164 1165 1166 1167 1168 1169
		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	
1170 1171
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1172
				   else do setInput i2; lit_error
1173

1174
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1175 1176
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
1177
  = do	magicHash <- extension magicHashEnabled
1178
	i@(AI end _ _) <- getInput
1179
	if magicHash then do
1180
		case alexGetChar' i of
1181
			Just ('#',i@(AI end _ _)) -> do
1182
				setInput i
1183
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1184
			_other ->
1185 1186 1187 1188
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1189 1190 1191 1192 1193
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
1194
      _other -> lit_error
1195

1196 1197 1198
isAny c | c > '\xff' = isPrint c
	| otherwise  = is_any c

1199 1200 1201 1202 1203 1204 1205 1206 1207 1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
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

1218 1219 1220
	'x'   -> readNum is_hexdigit 16 hexDigit
	'o'   -> readNum is_octdigit  8 octDecDigit
	x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x)
1221 1222 1223

	c1 ->  do
	   i <- getInput
1224
	   case alexGetChar' i of
1225 1226
	    Nothing -> lit_error
	    Just (c2,i2) -> 
1227 1228
              case alexGetChar' i2 of
		Nothing	-> do setInput i2; lit_error
1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242
		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
1243
  i <- getInput
1244 1245 1246
  c <- getCharOrFail
  if is_digit c 
	then readNum2 is_digit base conv (conv c)
1247
	else do setInput i; lit_error
1248 1249 1250 1251 1252

readNum2 is_digit base conv i = do
  input <- getInput
  read i input
  where read i input = do
1253
	  case alexGetChar' input of
1254 1255 1256 1257
	    Just (c,input') | is_digit c -> do
		read (i*base + conv c) input'
	    _other -> do
		if i >= 0 && i <= 0x10FFFF
1258
		   then do setInput input; return (chr i)
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 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297
		   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')
	]

1298 1299 1300 1301
-- 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.
1302 1303 1304 1305 1306
lit_error = lexError "lexical error in string/character literal"

getCharOrFail :: P Char
getCharOrFail =  do
  i <- getInput
1307
  case alexGetChar' i of
1308 1309 1310
	Nothing -> lexError "unexpected end-of-file in string/character literal"
	Just (c,i)  -> do setInput i; return c

1311 1312 1313 1314
-- -----------------------------------------------------------------------------
-- Warnings

warn :: DynFlag -> SDoc -> Action
1315 1316
warn option warning srcspan _buf _len = do
    addWarning option srcspan warning
1317 1318
    lexToken

1319 1320 1321 1322 1323 1324
-- -----------------------------------------------------------------------------
-- The Parse Monad

data LayoutContext
  = NoLayout
  | Layout !Int
1325
  deriving Show
1326 1327 1328 1329

data ParseResult a
  = POk PState a
  | PFailed 
1330
	SrcSpan		-- The start and end of the text span related to
1331 1332
			-- the error.  Might be used in environments which can 
			-- show this span, e.g. by highlighting it.
1333 1334 1335 1336
	Message		-- The error message

data PState = PState { 
	buffer	   :: StringBuffer,
1337 1338
    dflags     :: DynFlags,
    messages   :: Messages,
1339
        last_loc   :: SrcSpan,	-- pos of previous token
1340 1341 1342
        last_offs  :: !Int, 	-- offset of the previous token from the
				-- beginning of  the current line.
				-- \t is equal to 8 spaces.
1343
	last_len   :: !Int,	-- len of previous token
1344
  last_line_len :: !Int,
1345 1346 1347 1348 1349
        loc        :: SrcLoc,   -- current loc (end of prev token + 1)
	extsBitmap :: !Int,	-- bitmap that determines permitted extensions
	context	   :: [LayoutContext],
	lex_state  :: [Int]
     }
1350
	-- last_loc and last_len are used when generating error messages,