Lexer.x 57 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23
--
-- GHC's lexer.
--
-- This is a combination of an Alex-generated lexer from a regex
-- definition, with some hand-coded bits.
--
-- Completely accurate information about token-spans within the source
-- file is maintained.  Every token has a start and end SrcLoc attached to it.
--
-----------------------------------------------------------------------------

--   ToDo / known bugs:
--    - Unicode
--    - parsing integers is a bit slow
--    - readRational is a bit slow
--
--   Known bugs, that were also in the previous version:
--    - M... should be 3 tokens, not 1.
--    - pragma-end should be only valid in a pragma

{
24
{-# OPTIONS -w #-}
25 26 27
-- The above warning supression flag is a temporary kludge.
-- While working on this module you are encouraged to remove it and fix
-- any warnings in the module. See
Ian Lynagh's avatar
Ian Lynagh committed
28
--     http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
29
-- for details
twanvl's avatar
twanvl committed
30 31 32
--
-- Note that Alex itself generates code with with some unused bindings and
-- without type signatures, so removing the flag might not be possible.
33

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

123 124 125 126
haskell :-

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

-- Everywhere: deal with nested comments.  We explicitly rule out
-- pragmas, "{-#", so that we don't accidentally treat them as comments.
-- (this can happen even though pragmas will normally take precedence due to
-- longest-match, because pragmas aren't valid in every state, but comments
133 134 135 136
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

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

-- Single-line comments are a bit tricky.  Haskell 98 says that two or
-- more dashes followed by a symbol should be parsed as a varsym, so we
-- have to exclude those.
141 142 143 144 145 146 147 148 149 150

-- Since Haddock comments aren't valid in every state, we need to rule them
-- out here.  

-- The following two rules match comments that begin with two dashes, but
-- continue with a different character. The rules test that this character
-- is not a symbol (in which case we'd have a varsym), and that it's not a
-- space followed by a Haddock comment symbol (docsym) (in which case we'd
-- have a Haddock comment). The rules then munch the rest of the line.

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

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

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

-- Now, when we've matched comments that begin with 2 dashes and continue
-- with a different character, we need to match comments that begin with three
-- or more dashes (which clearly can't be Haddock comments). We only need to
-- make sure that the first non-dash character isn't a symbol, and munch the
-- rest of the line.

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

-- Since the previous rules all match dashes followed by at least one
-- character, we also need to match a whole line filled with just dashes.

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

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

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

-- 'bol' state: beginning of a line.  Slurp up all the whitespace (including
-- blank lines) until we find a non-whitespace character, then do layout
-- processing.
--
-- One slight wibble here: what if the line begins with {-#? In
-- theory, we have to lex the pragma to see if it's one we recognise,
-- and if it is, then we backtrack and do_bol, otherwise we treat it
-- as a nested comment.  We don't bother with this: if the line begins
-- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
<bol> {
  \n					;
  ^\# (line)?				{ begin line_prag1 }
188
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
189
  ^\# \! .* \n				; -- #!, for scripts
190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211
  ()					{ do_bol }
}

-- after a layout keyword (let, where, do, of), we begin a new layout
-- context if the curly brace is missing.
-- Careful! This stuff is quite delicate.
<layout, layout_do> {
  \{ / { notFollowedBy '-' }		{ pop_and open_brace }
	-- we might encounter {-# here, but {- has been handled already
  \n					;
  ^\# (line)?				{ begin line_prag1 }
}

-- do is treated in a subtly different way, see new_layout_context
<layout>    ()				{ new_layout_context True }
<layout_do> ()				{ new_layout_context False }

-- after a new layout context which was found to be to the left of the
-- previous context, we have generated a '{' token, and we now need to
-- generate a matching '}' token.
<layout_left>  ()			{ do_layout_left }

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

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

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

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

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

239
<0,option_prags> {
240 241 242 243 244
  "{-#" $whitechar* (INLINE|inline)	{ token (ITinline_prag True) }
  "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
  					{ token (ITinline_prag False) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
  					{ token ITspec_prag }
245
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
246 247 248 249
	$whitechar* (INLINE|inline)	{ token (ITspec_inline_prag True) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
	$whitechar* (NO(T?)INLINE|no(t?)inline)
					{ token (ITspec_inline_prag False) }
250 251 252 253
  "{-#" $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
254 255
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
256
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
257
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
258

259
 "{-#"                                 { nested_comment lexToken }
260 261 262 263 264

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

265
<option_prags> {
David Waern's avatar
David Waern committed
266 267
  "{-#"  $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc)
268
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
269 270 271 272 273 274 275 276 277
  "{-#"  $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> {
  "-- #" .* ;
278
}
279

280
<0,option_prags> {
281
	-- This is to catch things like {-# OPTIONS OPTIONS_HUGS ... 
282
  "{-#" $whitechar* $idchar+		{ nested_comment lexToken }
283 284
}

285 286
-- '0' state: ordinary lexemes

287 288
-- Haddock comments

289
<0> {
David Waern's avatar
David Waern committed
290 291
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
292 293
}

294 295
-- "special" symbols

296
<0> {
297 298 299 300
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
301
<0> {
302 303 304 305 306 307 308 309
  "[|"	    / { 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 }
310 311 312

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
313 314
}

315
<0> {
ross's avatar
ross committed
316 317
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
318 319 320
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

321
<0> {
322 323 324
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

325
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
326 327 328 329 330 331
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

332
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
333 334
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
335 336
}

337
<0,option_prags> {
338 339 340 341 342 343 344 345 346 347 348 349
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

350
<0,option_prags> {
351
  @qual @varid			{ idtoken qvarid }
352 353 354 355 356
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

357
<0> {
358 359 360 361
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
362 363 364 365
}

-- ToDo: M.(,,,)

366
<0> {
367 368 369 370 371 372
  @qual @varsym			{ idtoken qvarsym }
  @qual @consym			{ idtoken qconsym }
  @varsym			{ varsym }
  @consym			{ consym }
}

373 374
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
375
<0> {
376
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
377 378 379
  @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 }
380 381

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
382
  @floating_point    { strtoken tok_float }
383 384
}

385
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
386
  -- Unboxed ints (:: Int#) and words (:: Word#)
387 388
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
389 390 391 392 393
  @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 }
394
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
395

Ian Lynagh's avatar
Ian Lynagh committed
396 397 398 399
  @decimal                     \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
  0[oO] @octal                 \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
  0[xX] @hexadecimal           \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }

400 401
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
402 403
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
404 405 406 407 408 409
}

-- 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.
410
<0> {
411 412 413 414 415 416 417 418
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

419
data Token
420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457
  = 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
458
  | ITfamily
459 460 461
  | ITgroup
  | ITby
  | ITusing
462

463 464 465 466
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
467 468 469 470 471
  | ITsource_prag
  | ITrules_prag
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
472
  | ITgenerated_prag
473
  | ITcore_prag                 -- hdaume: core annotations
474
  | ITunpack_prag
475
  | ITclose_prag
476 477 478
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
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

  | 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
505 506
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote

  | ITvarid   FastString	-- identifiers
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)

  | ITdupipvarid   FastString	-- GHC extension: implicit param: ?x

  | ITpragma StringBuffer

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
  | ITrational   Rational

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
538
  | ITprimword   Integer
539 540 541 542
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
543 544 545 546 547 548 549 550 551
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
552
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
553 554 555 556

  -- Arrow notation extension
  | ITproc
  | ITrec
557 558 559 560 561 562
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
563 564 565

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
566 567 568 569 570 571 572

  -- 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
573
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
574

575 576 577 578
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
579
{-
580
isSpecial :: Token -> Bool
581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596
-- 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
597
isSpecial ITfamily	= True
598 599 600
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
601
isSpecial _             = False
twanvl's avatar
twanvl committed
602
-}
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639

-- 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

640
    ( "forall",	ITforall,	 bit explicitForallBit),
Ian Lynagh's avatar
Ian Lynagh committed
641
	( "mdo",	ITmdo,		 bit recursiveDoBit),
642
	( "family",	ITfamily,	 bit tyFamBit),
643 644 645
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
646 647 648 649 650 651 652 653 654 655 656 657 658

	( "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),
659
	( "proc",	ITproc,		 bit arrowsBit)
660 661
     ]

Ian Lynagh's avatar
Ian Lynagh committed
662
reservedSymsFM :: UniqFM (Token, Int -> Bool)
663
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680
    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
681
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
682
        -- For 'forall a . t'
Ian Lynagh's avatar
Ian Lynagh committed
683
       ,(".", ITdot, explicitForallEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
684 685 686 687 688

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
689 690

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
691 692
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
693 694
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
695 696 697
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
698 699 700
        -- 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).
701
#endif
702 703 704 705 706
       ]

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

707
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
708

709
special :: Token -> Action
twanvl's avatar
twanvl committed
710
special tok span _buf _len = return (L span tok)
711

712
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
713 714
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
715

716 717
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
718

719 720 721
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
722

723 724 725
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
726

727
init_strtoken :: Int -> (String -> Token) -> Action
728
-- like strtoken, but drops the last N character(s)
729 730
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
731 732

begin :: Int -> Action
733
begin code _span _str _len = do pushLexState code; lexToken
734 735

pop :: Action
736
pop _span _buf _len = do popLexState; lexToken
737 738

pop_and :: Action -> Action
739
pop_and act span buf len = do popLexState; act span buf len
740

741 742 743 744 745
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
746

747
notFollowedBySymbol _ _ _ (AI _ _ buf)
748
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
749

750 751 752 753 754
-- 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.
755
isNormalComment bits _ _ (AI _ _ buf)
756 757 758 759 760 761 762
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
763

twanvl's avatar
twanvl committed
764
{-
765 766
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
767
-}
768

769
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
770

771 772
ifExtension pred bits _ _ _ = pred bits

773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800
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

801 802 803 804
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
805 806
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
807
  input <- getInput
808
  go (1::Int) input
809 810 811 812 813 814 815
  where
    go 0 input = do setInput input; cont
    go n input = case alexGetChar input of
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('\125',input) -> go (n-1) input
twanvl's avatar
twanvl committed
816
        Just (_,_)          -> go n input
817 818 819
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> go (n+1) input
twanvl's avatar
twanvl committed
820 821
        Just (_,_)       -> go n input
      Just (_,input) -> go n input
822 823 824 825 826 827 828 829

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
    go commentAcc input docType _ = case alexGetChar input of
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
830
        Just ('\125',input) ->
831
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
832
        Just (_,_) -> go ('-':commentAcc) input docType False
833 834 835 836 837 838
      Just ('\123', input) -> case alexGetChar input of
        Nothing  -> errBrace input span
        Just ('-',input) -> do
          setInput input
          let cont = do input <- getInput; go commentAcc input docType False
          nested_comment cont span buf _len
twanvl's avatar
twanvl committed
839
        Just (_,_) -> go ('\123':commentAcc) input docType False
840 841 842
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
843 844 845 846 847
  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
848 849
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
850 851 852
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
853
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889
      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 `{-'"
 
890
open_brace, close_brace :: Action
891
open_brace span _str _len = do 
892 893
  ctx <- getContext
  setContext (NoLayout:ctx)
894 895
  return (L span ITocurly)
close_brace span _str _len = do 
896
  popContext
897
  return (L span ITccurly)
898 899 900 901 902 903 904 905

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.
906
splitQualName orig_buf len = split orig_buf orig_buf
907
  where
908 909 910 911 912 913
    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
914 915 916 917
  
    -- 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.
918 919 920 921 922 923 924 925 926 927 928
    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
929

930
varid span buf len = 
Ian Lynagh's avatar
Ian Lynagh committed
931
  fs `seq`
932 933 934
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
935
		return (L span keyword)
936 937 938
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
939 940 941
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
942 943 944 945 946 947 948 949 950 951 952 953
  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

954
sym con span buf len = 
955 956
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
957
		b <- extension exts
958 959 960
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
961 962 963
  where
	fs = lexemeToFastString buf len

964 965 966 967 968 969 970 971 972 973 974 975 976
-- Variations on the integral numeric literal.
tok_integral :: (Integer -> Token)
     -> (Integer -> Integer)
 --    -> (StringBuffer -> StringBuffer) -> (Int -> Int)
     -> Int -> Int
     -> (Integer, (Char->Int)) -> Action
tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
  return $ L span $ itint $! transint $ parseUnsignedInteger
     (offsetBytes transbuf buf) (subtract translen len) radix char_to_int

-- some conveniences for use with tok_integral
tok_num = tok_integral ITinteger
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
977
tok_primword = tok_integral ITprimword positive
978 979 980 981 982 983 984
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
985
tok_float        str = ITrational   $! readRational str
986 987
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
988 989 990 991 992 993

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
994
do_bol span _str _len = do
995
	pos <- getOffside
996 997 998 999 1000
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1001
		return (L span ITvccurly)
1002 1003 1004
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1005
		return (L span ITsemi)
1006 1007 1008 1009 1010 1011 1012
	    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
1013
maybe_layout ITmdo	= pushLexState layout_do
1014 1015 1016
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1017
maybe_layout ITrec	= pushLexState layout
1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028
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.
--
1029
new_layout_context strict span _buf _len = do
1030
    popLexState
1031
    (AI _ offset _) <- getInput
1032 1033 1034 1035 1036 1037 1038 1039
    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
1040
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1041
	_ -> do
1042
		setContext (Layout offset : ctx)
1043
		return (L span ITvocurly)
1044

1045
do_layout_left span _buf _len = do
1046 1047
    popLexState
    pushLexState bol  -- we must be at the start of a line
1048
    return (L span ITvccurly)
1049 1050 1051 1052

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

1053 1054
setLine :: Int -> Action
setLine code span buf len = do
1055
  let line = parseUnsignedInteger buf len 10 octDecDigit
1056
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1057 1058 1059 1060 1061
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1062 1063
setFile :: Int -> Action
setFile code span buf len = do
1064
  let file = lexemeToFastString (stepOn buf) (len-2)
1065
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1066 1067 1068 1069
  popLexState
  pushLexState code
  lexToken

1070 1071 1072 1073 1074

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1075
lex_string_prag mkTok span _buf _len
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087
    = do input <- getInput
         start <- getSrcLoc
         tok <- go [] input
         end <- getSrcLoc
         return (L (mkSrcSpan start end) tok)
    where go acc input
              = if isString input "#-}"
                   then do setInput input
                           return (mkTok (reverse acc))
                   else case alexGetChar input of
                          Just (c,i) -> go (c:acc) i
                          Nothing -> err input
twanvl's avatar
twanvl committed
1088
          isString _ [] = True
1089 1090 1091 1092 1093 1094 1095
          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"


1096 1097 1098 1099 1100 1101
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1102
lex_string_tok span _buf _len = do
1103 1104
  tok <- lex_string ""
  end <- getSrcLoc 
1105
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1106

1107
lex_string :: String -> P Token
1108 1109
lex_string s = do
  i <- getInput
1110
  case alexGetChar' i of
1111 1112 1113 1114
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1115 1116
	magicHash <- extension magicHashEnabled
	if magicHash
1117 1118
	  then do
	    i <- getInput
1119
	    case alexGetChar' i of
1120 1121 1122 1123
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1124
                    else let s' = mkZFastString (reverse s) in
1125
			 return (ITprimstring s')
1126 1127
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
	      _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
1138
	where next = alexGetChar' i
1139

1140 1141 1142
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152

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
1153 1154 1155 1156 1157 1158
-- Here we are basically parsing character literals, such as 'x' or '\n'
-- but, when Template Haskell is on, we additionally spot
-- 'x and ''T, returning ITvarQuote and ITtyQuote respectively, 
-- but WIHTOUT CONSUMING the x or T part  (the parser does that).
-- So we have to do two characters of lookahead: when we see 'x we need to
-- see if there's a trailing quote
twanvl's avatar
twanvl committed
1159
lex_char_tok span _buf _len = do	-- We've seen '
1160
   i1 <- getInput	-- Look ahead to first character
1161
   let loc = srcSpanStart span
1162
   case alexGetChar' i1 of
1163 1164
	Nothing -> lit_error 

1165
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1166 1167 1168
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1169
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1170 1171
		   else lit_error

twanvl's avatar
twanvl committed
1172
	Just ('\\', i2@(AI _end2 _ _)) -> do 	-- We've seen 'backslash
1173 1174 1175 1176
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1177
			        else do setInput i2; lit_error 
1178

twanvl's avatar
twanvl committed
1179
        Just (c, i2@(AI _end2 _ _))
1180 1181
		| not (isAny c) -> lit_error
		| otherwise ->
1182 1183 1184

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
1185
	   case alexGetChar' i2 of	-- Look ahead one more character
1186 1187 1188 1189 1190 1191 1192
		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	
1193 1194
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1195
				   else do setInput i2; lit_error
1196

1197
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1198 1199
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
1200
  = do	magicHash <- extension magicHashEnabled
1201
	i@(AI end _ _) <- getInput
1202
	if magicHash then do
1203
		case alexGetChar' i of
1204
			Just ('#',i@(AI end _ _)) -> do
1205
				setInput i
1206
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1207
			_other ->
1208 1209 1210 1211
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1212 1213 1214 1215 1216
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
1217
      _other -> lit_error
1218

1219
isAny c | c > '\x7f' = isPrint c
1220 1221
	| otherwise  = is_any c

1222 1223 1224 1225 1226 1227 1228 1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240
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

1241 1242
	'x'   -> readNum is_hexdigit 16 hexDigit
	'o'   -> readNum is_octdigit  8 octDecDigit
1243
	x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x)
1244 1245 1246

	c1 ->  do
	   i <- getInput
1247
	   case alexGetChar' i of