Lexer.x 42.8 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 24
--
-- GHC's lexer.
--
-- This is a combination of an Alex-generated lexer from a regex
-- definition, with some hand-coded bits.
--
-- Completely accurate information about token-spans within the source
-- file is maintained.  Every token has a start and end SrcLoc attached to it.
--
-----------------------------------------------------------------------------

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

{
module Lexer (
25
   Token(..), lexer, pragState, mkPState, PState(..),
26
   P(..), ParseResult(..), getSrcLoc, 
27 28
   failLocMsgP, failSpanMsgP, srcParseFail,
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
29 30
   getLexState, popLexState, pushLexState,
   extension, bangPatEnabled
31 32 33 34 35 36 37 38 39 40 41
  ) where

#include "HsVersions.h"

import ErrUtils		( Message )
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
42
import DynFlags
43
import Ctype
44
import Util		( maybePrefixMatch, readRational )
45 46

import DATA_BITS
47
import Data.Char 	( chr )
48
import Ratio
49
--import TRACE
50 51

#if __GLASGOW_HASKELL__ >= 605
52
import Data.Char 	( GeneralCategory(..), generalCategory, isPrint, isUpper )
53
#else
54
import Compat.Unicode	( GeneralCategory(..), generalCategory, isPrint, isUpper )
55
#endif
56 57
}

58 59
$unispace    = \x05
$whitechar   = [\ \t\n\r\f\v\xa0 $unispace]
60 61 62
$white_no_nl = $whitechar # \n

$ascdigit  = 0-9
63 64
$unidigit  = \x03
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
65 66 67
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
68
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~ \xa1-\xbf \xd7 \xf7]
69
$unisymbol = \x04
70 71
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

72
$unilarge  = \x01
73 74 75
$asclarge  = [A-Z \xc0-\xd6 \xd8-\xde]
$large     = [$asclarge $unilarge]

76
$unismall  = \x02
77 78 79
$ascsmall  = [a-z \xdf-\xf6 \xf8-\xff]
$small     = [$ascsmall $unismall \_]

80 81
$unigraphic = \x06
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
82 83

$octit	   = 0-7
84
$hexit     = [$decdigit A-F a-f]
85 86 87 88 89 90 91 92 93 94
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

@varid     = $small $idchar*
@conid     = $large $idchar*

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

95
@decimal     = $decdigit+
96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;

-- 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
-- are).
"{-" / { notFollowedBy '#' }		{ nested_comment }

-- 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.
-- The regex says: "munch all the characters after the dashes, as long as
-- the first one is not a symbol".
122
"--"\-* [^$symbol :] .*			;
123
"--"\-* / { atEOL }			;
124 125 126 127 128 129 130 131 132 133 134 135 136

-- '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 }
137
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
138
  ^\# \! .* \n				; -- #!, for scripts
139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160
  ()					{ 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 }

161
<0,option_prags,glaexts> \n				{ begin bol }
162 163 164 165 166

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

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
167
<line_prag1> $decdigit+			{ setLine line_prag1a }
168
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
169 170 171 172
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
173
<line_prag2> $decdigit+			{ setLine line_prag2a }
174
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
175 176 177
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
178

179 180 181 182 183 184 185 186
-- We only want RULES pragmas to be picked up when -fglasgow-exts
-- is on, because the contents of the pragma is always written using
-- glasgow-exts syntax (using forall etc.), so if glasgow exts are not
-- enabled, we're sure to get a parse error.
-- (ToDo: we should really emit a warning when ignoring pragmas)
<glaexts>
  "{-#" $whitechar* (RULES|rules)	{ token ITrules_prag }

187
<0,option_prags,glaexts> {
188 189 190 191 192
  "{-#" $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 }
193
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
194 195 196 197
	$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) }
198 199 200 201 202
  "{-#" $whitechar* (SOURCE|source)	{ token ITsource_prag }
  "{-#" $whitechar* (DEPRECATED|deprecated)
  					{ token ITdeprecated_prag }
  "{-#" $whitechar* (SCC|scc)		{ token ITscc_prag }
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
203
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
204

205 206 207 208 209 210
  "{-#" 				{ nested_comment }

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

211 212 213 214 215 216 217
<option_prags> {
  "{-#" $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#" $whitechar* (OPTIONS_GHC|options_ghc)
                                        { lex_string_prag IToptions_prag }
  "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
  "{-#" $whitechar* (INCLUDE|include)   { lex_string_prag ITinclude_prag }
}
218 219 220 221 222 223

-- '0' state: ordinary lexemes
-- 'glaexts' state: glasgow extensions (postfix '#', etc.)

-- "special" symbols

224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
<0,glaexts> {
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
<0,glaexts> {
  "[|"	    / { 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 }
}

<0,glaexts> {
ross's avatar
ross committed
241 242
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
243 244 245 246 247 248 249 250
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

<0,glaexts> {
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
  \% @varid / { ifExtension ipEnabled } { skip_one_varid ITsplitipvarid }
}

251
<glaexts> {
ross's avatar
ross committed
252
  "(#" / { notFollowedBySymbol }	{ token IToubxparen }
253 254 255 256 257
  "#)"					{ token ITcubxparen }
  "{|"					{ token ITocurlybar }
  "|}"					{ token ITccurlybar }
}

258
<0,option_prags,glaexts> {
259 260 261 262 263 264 265 266 267 268 269 270
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

271
<0,option_prags,glaexts> {
272 273 274 275 276 277 278 279 280 281 282 283 284 285 286
  @qual @varid			{ check_qvarid }
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

-- after an illegal qvarid, such as 'M.let', 
-- we back up and try again in the bad_qvarid state:
<bad_qvarid> {
  @conid			{ pop_and (idtoken conid) }
  @qual @conid			{ pop_and (idtoken qconid) }
}

<glaexts> {
  @qual @varid "#"+		{ idtoken qvarid }
287
  @qual @conid "#"+		{ idtoken qconid }
288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334
  @varid "#"+			{ varid }
  @conid "#"+			{ idtoken conid }
}

-- ToDo: M.(,,,)

<0,glaexts> {
  @qual @varsym			{ idtoken qvarsym }
  @qual @consym			{ idtoken qconsym }
  @varsym			{ varsym }
  @consym			{ consym }
}

<0,glaexts> {
  @decimal			{ tok_decimal }
  0[oO] @octal			{ tok_octal }
  0[xX] @hexadecimal		{ tok_hexadecimal }
}

<glaexts> {
  @decimal \#			{ prim_decimal }
  0[oO] @octal \#		{ prim_octal }
  0[xX] @hexadecimal \#		{ prim_hexadecimal }
}

<0,glaexts> @floating_point		{ strtoken tok_float }
<glaexts>   @floating_point \#		{ init_strtoken 1 prim_float }
<glaexts>   @floating_point \# \#	{ init_strtoken 2 prim_double }

-- 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.
<0,glaexts> {
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

{
-- work around bug in Alex 2.0
#if __GLASGOW_HASKELL__ < 503
unsafeAt arr i = arr ! i
#endif

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

335
data Token
336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374
  = 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

375 376 377 378
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
379 380 381 382 383 384
  | ITsource_prag
  | ITrules_prag
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
  | ITcore_prag                 -- hdaume: core annotations
385
  | ITunpack_prag
386
  | ITclose_prag
387 388 389
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453

  | 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
  | ITsplitipvarid 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
454 455 456 457 458 459 460 461 462
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
463 464 465 466

  -- Arrow notation extension
  | ITproc
  | ITrec
467 468 469 470 471 472
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
473 474 475 476 477 478 479

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
#ifdef DEBUG
  deriving Show -- debugging
#endif

480
isSpecial :: Token -> Bool
481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
-- 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
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

535
      	( "forall",	ITforall,	 bit tvBit),
536 537 538 539 540 541 542 543 544 545 546 547 548 549
	( "mdo",	ITmdo,		 bit glaExtsBit),

	( "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),
550
	( "proc",	ITproc,		 bit arrowsBit)
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
     ]

reservedSymsFM = listToUFM $
	map (\ (x,y,z) -> (mkFastString x,(y,z)))
      [ ("..",	ITdotdot,	0)
       ,(":",	ITcolon,	0)	-- (:) is a reserved op, 
						-- meaning only list cons
       ,("::",	ITdcolon, 	0)
       ,("=",	ITequal, 	0)
       ,("\\",	ITlam, 		0)
       ,("|",	ITvbar, 	0)
       ,("<-",	ITlarrow, 	0)
       ,("->",	ITrarrow, 	0)
       ,("@",	ITat, 		0)
       ,("~",	ITtilde, 	0)
       ,("=>",	ITdarrow, 	0)
       ,("-",	ITminus, 	0)
       ,("!",	ITbang, 	0)

       ,("*",	ITstar,		bit glaExtsBit)	-- For data T (a::*) = MkT
571
       ,(".",	ITdot,		bit tvBit)	-- For 'forall a . t'
572 573 574 575 576

       ,("-<",	ITlarrowtail,	bit arrowsBit)
       ,(">-",	ITrarrowtail,	bit arrowsBit)
       ,("-<<",	ITLarrowtail,	bit arrowsBit)
       ,(">>-",	ITRarrowtail,	bit arrowsBit)
577 578

#if __GLASGOW_HASKELL__ >= 605
579
       ,("λ",	ITlam,          bit glaExtsBit)
580 581
       ,("∷",   ITdcolon,       bit glaExtsBit)
       ,("⇒",   ITdarrow,	bit glaExtsBit)
582 583 584 585
       ,("∀",	ITforall,	bit glaExtsBit)
       ,("→",   ITrarrow,	bit glaExtsBit)
       ,("←",   ITlarrow,	bit glaExtsBit)
       ,("⋯", 	ITdotdot,	bit glaExtsBit)
586
#endif
587 588 589 590 591
       ]

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

592
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
593

594 595
special :: Token -> Action
special tok span _buf len = return (L span tok)
596

597 598 599
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)
600

601 602
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
603

604 605 606
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
607

608 609 610
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
611

612
init_strtoken :: Int -> (String -> Token) -> Action
613
-- like strtoken, but drops the last N character(s)
614 615
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
616 617

begin :: Int -> Action
618
begin code _span _str _len = do pushLexState code; lexToken
619 620

pop :: Action
621
pop _span _buf _len = do popLexState; lexToken
622 623

pop_and :: Action -> Action
624
pop_and act span buf len = do popLexState; act span buf len
625

626
notFollowedBy char _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf /= char
627

628
notFollowedBySymbol _ _ _ (AI _ _ buf)
ross's avatar
ross committed
629 630
  = atEnd buf || currentChar buf `notElem` "!#$%&*+./<=>?@\\^|-~"

631
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
632

633 634
ifExtension pred bits _ _ _ = pred bits

635 636 637 638 639
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
nested_comment :: Action
640
nested_comment span _str _len = do
641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660
  input <- getInput
  go 1 input
  where go 0 input = do setInput input; lexToken
	go n input = do
	  case alexGetChar input of
	    Nothing  -> err input
	    Just (c,input) -> do
	      case c of
	    	'-' -> do
		  case alexGetChar input of
		    Nothing  -> err input
		    Just ('\125',input) -> go (n-1) input
		    Just (c,_)          -> go n input
	     	'\123' -> do
		  case alexGetChar input of
		    Nothing  -> err input
		    Just ('-',input') -> go (n+1) input'
		    Just (c,input)    -> go n input
	    	c -> go n input

661
        err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
662 663

open_brace, close_brace :: Action
664
open_brace span _str _len = do 
665 666
  ctx <- getContext
  setContext (NoLayout:ctx)
667 668
  return (L span ITocurly)
close_brace span _str _len = do 
669
  popContext
670
  return (L span ITccurly)
671 672 673 674

-- We have to be careful not to count M.<varid> as a qualified name
-- when <varid> is a keyword.  We hack around this by catching 
-- the offending tokens afterward, and re-lexing in a different state.
675
check_qvarid span buf len = do
676 677 678 679 680 681 682 683 684 685 686 687
  case lookupUFM reservedWordsFM var of
	Just (keyword,exts)
	  | not (isSpecial keyword) ->
	  if exts == 0 
	     then try_again
	     else do
		b <- extension (\i -> exts .&. i /= 0)
		if b then try_again
		     else return token
	_other -> return token
  where
	(mod,var) = splitQualName buf len
688
	token     = L span (ITqvarid (mod,var))
689 690

	try_again = do
691 692
		(AI _ offs _) <- getInput	
		setInput (AI (srcSpanStart span) (offs-len) buf)
693 694 695 696 697 698 699 700 701 702
		pushLexState bad_qvarid
		lexToken

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.
703
splitQualName orig_buf len = split orig_buf orig_buf
704
  where
705 706 707 708 709 710
    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
711 712 713 714
  
    -- 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.
715 716 717 718 719 720 721 722 723 724 725
    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
726

727
varid span buf len = 
728 729 730
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
731
		return (L span keyword)
732 733 734
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
735 736 737
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
738 739 740 741 742 743 744 745 746 747 748 749
  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

750
sym con span buf len = 
751
  case lookupUFM reservedSymsFM fs of
752
	Just (keyword,0)    -> return (L span keyword)
753 754
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
755 756 757
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
758 759 760
  where
	fs = lexemeToFastString buf len

761
tok_decimal span buf len 
762
  = return (L span (ITinteger  $! parseInteger buf len 10 octDecDigit))
763

764
tok_octal span buf len 
765
  = return (L span (ITinteger  $! parseInteger (offsetBytes 2 buf) (len-2) 8 octDecDigit))
766

767
tok_hexadecimal span buf len 
768
  = return (L span (ITinteger  $! parseInteger (offsetBytes 2 buf) (len-2) 16 hexDigit))
769

770
prim_decimal span buf len 
771
  = return (L span (ITprimint  $! parseInteger buf (len-1) 10 octDecDigit))
772

773
prim_octal span buf len 
774
  = return (L span (ITprimint  $! parseInteger (offsetBytes 2 buf) (len-3) 8 octDecDigit))
775

776
prim_hexadecimal span buf len 
777
  = return (L span (ITprimint  $! parseInteger (offsetBytes 2 buf) (len-3) 16 hexDigit))
778

779 780 781
tok_float        str = ITrational   $! readRational str
prim_float       str = ITprimfloat  $! readRational str
prim_double      str = ITprimdouble $! readRational str
782 783 784 785 786 787

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
788
do_bol span _str _len = do
789
	pos <- getOffside
790 791 792 793 794
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
795
		return (L span ITvccurly)
796 797 798
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
799
		return (L span ITsemi)
800 801 802 803 804 805 806
	    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
807
maybe_layout ITmdo	= pushLexState layout_do
808 809 810
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
811
maybe_layout ITrec	= pushLexState layout
812 813 814 815 816 817 818 819 820 821 822
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.
--
823
new_layout_context strict span _buf _len = do
824
    popLexState
825
    (AI _ offset _) <- getInput
826 827 828 829 830 831 832 833
    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
834
		return (L span ITvocurly)
835 836
	other -> do
		setContext (Layout offset : ctx)
837
		return (L span ITvocurly)
838

839
do_layout_left span _buf _len = do
840 841
    popLexState
    pushLexState bol  -- we must be at the start of a line
842
    return (L span ITvccurly)
843 844 845 846

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

847 848 849
setLine :: Int -> Action
setLine code span buf len = do
  let line = parseInteger buf len 10 octDecDigit
850
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
851 852 853 854 855
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

856 857
setFile :: Int -> Action
setFile code span buf len = do
858
  let file = lexemeToFastString (stepOn buf) (len-2)
859
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
860 861 862 863
  popLexState
  pushLexState code
  lexToken

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

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


890 891 892 893 894 895
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
896
lex_string_tok span buf len = do
897 898
  tok <- lex_string ""
  end <- getSrcLoc 
899
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
900

901
lex_string :: String -> P Token
902 903
lex_string s = do
  i <- getInput
904
  case alexGetChar' i of
905 906 907 908 909 910 911 912
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
	glaexts <- extension glaExtsEnabled
	if glaexts
	  then do
	    i <- getInput
913
	    case alexGetChar' i of
914 915 916 917
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
918
                    else let s' = mkZFastString (reverse s) in
919
			 return (ITprimstring s')
920 921
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
922 923 924 925 926 927 928 929 930 931
	      _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
932
	where next = alexGetChar' i
933

934 935 936
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
937 938 939 940 941 942 943 944 945 946

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
947 948 949 950 951 952
-- 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
953
lex_char_tok span buf len = do	-- We've seen '
954
   i1 <- getInput	-- Look ahead to first character
955
   let loc = srcSpanStart span
956
   case alexGetChar' i1 of
957 958
	Nothing -> lit_error 

959
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
960 961 962
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
963
			return (L (mkSrcSpan loc end2)  ITtyQuote)
964 965
		   else lit_error

966
	Just ('\\', i2@(AI end2 _ _)) -> do 	-- We've seen 'backslash 
967 968 969 970
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
971
			        else do setInput i2; lit_error 
972

973 974 975
        Just (c, i2@(AI end2 _ _)) 
		| not (isAny c) -> lit_error
		| otherwise ->
976 977 978

		-- We've seen 'x, where x is a valid character
		--  (i.e. not newline etc) but not a quote or backslash
979
	   case alexGetChar' i2 of	-- Look ahead one more character
980 981 982 983 984 985 986
		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	
987 988
			let (AI end _ _) = i1
			if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
989
				   else do setInput i2; lit_error
990

991
finish_char_tok :: SrcLoc -> Char -> P (Located Token)
992 993 994
finish_char_tok loc ch	-- We've already seen the closing quote
			-- Just need to check for trailing #
  = do	glaexts <- extension glaExtsEnabled
995
	i@(AI end _ _) <- getInput
996
	if glaexts then do
997
		case alexGetChar' i of
998
			Just ('#',i@(AI end _ _)) -> do
999
				setInput i
1000
				return (L (mkSrcSpan loc end) (ITprimchar ch))
1001
			_other ->
1002 1003 1004 1005
				return (L (mkSrcSpan loc end) (ITchar ch))
	        else do
		   return (L (mkSrcSpan loc end) (ITchar ch))

1006 1007 1008 1009 1010
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
1011
      _other -> lit_error
1012

1013 1014 1015
isAny c | c > '\xff' = isPrint c
	| otherwise  = is_any c

1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034
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

1035 1036 1037
	'x'   -> readNum is_hexdigit 16 hexDigit
	'o'   -> readNum is_octdigit  8 octDecDigit
	x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x)
1038 1039 1040

	c1 ->  do
	   i <- getInput
1041
	   case alexGetChar' i of
1042 1043
	    Nothing -> lit_error
	    Just (c2,i2) -> 
1044 1045
              case alexGetChar' i2 of
		Nothing	-> do setInput i2; lit_error
1046 1047 1048 1049 1050 1051 1052 1053 1054 1055 1056 1057 1058 1059
		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
1060
  i <- getInput
1061 1062 1063
  c <- getCharOrFail
  if is_digit c 
	then readNum2 is_digit base conv (conv c)
1064
	else do setInput i; lit_error
1065 1066 1067 1068 1069

readNum2 is_digit base conv i = do
  input <- getInput
  read i input
  where read i input = do
1070
	  case alexGetChar' input of
1071 1072 1073 1074
	    Just (c,input') | is_digit c -> do
		read (i*base + conv c) input'
	    _other -> do
		if i >= 0 && i <= 0x10FFFF
1075
		   then do setInput input; return (chr i)
1076 1077 1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110 1111 1112 1113 1114
		   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')
	]

1115 1116 1117 1118
-- 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.
1119 1120 1121 1122 1123
lit_error = lexError "lexical error in string/character literal"

getCharOrFail :: P Char
getCharOrFail =  do
  i <- getInput
1124
  case alexGetChar' i of
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137
	Nothing -> lexError "unexpected end-of-file in string/character literal"
	Just (c,i)  -> do setInput i; return c

-- -----------------------------------------------------------------------------
-- The Parse Monad

data LayoutContext
  = NoLayout
  | Layout !Int

data ParseResult a
  = POk PState a
  | PFailed 
1138
	SrcSpan		-- The start and end of the text span related to
1139 1140
			-- the error.  Might be used in environments which can 
			-- show this span, e.g. by highlighting it.
1141 1142 1143 1144
	Message		-- The error message

data PState = PState { 
	buffer	   :: StringBuffer,
1145
        last_loc   :: SrcSpan,	-- pos of previous token
1146 1147 1148
        last_offs  :: !Int, 	-- offset of the previous token from the
				-- beginning of  the current line.
				-- \t is equal to 8 spaces.
1149
	last_len   :: !Int,	-- len of previous token
1150 1151 1152 1153 1154
        loc        :: SrcLoc,   -- current loc (end of prev token + 1)
	extsBitmap :: !Int,	-- bitmap that determines permitted extensions
	context	   :: [LayoutContext],
	lex_state  :: [Int]
     }
1155
	-- last_loc and last_len are used when generating error messages,
1156 1157 1158 1159
	-- and in pushCurrentContext only.  Sigh, if only Happy passed the
	-- current token to happyError, we could at least get rid of last_len.
	-- Getting rid of last_loc would require finding another way to 
	-- implement pushCurrentContext (which is only called from one place).
1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173

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

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

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

thenP :: P a -> (a -> P b) -> P b
(P m) `thenP` k = P $ \ s ->
	case m s of
1174 1175
		POk s1 a         -> (unP (k a)) s1
		PFailed span err -> PFailed span err
1176 1177

failP :: String -> P a
1178
failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1179 1180

failMsgP :: String -> P a
1181
failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1182 1183

failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1184 1185 1186 1187
failLocMsgP loc1 loc2 str = P $ \s -> PFailed (mkSrcSpan loc1 loc2) (text str)

failSpanMsgP :: SrcSpan -> String -> P a
failSpanMsgP span msg = P $ \s -> PFailed span (text msg)
1188 1189 1190 1191 1192 1193 1194 1195 1196 1197 1198 1199 1200

extension :: (Int -> Bool) -> P Bool
extension p = P $ \s -> POk s (p $! extsBitmap s)

getExts :: P Int
getExts = P $ \s -> POk s (extsBitmap s)

setSrcLoc :: SrcLoc -> P ()
setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()

getSrcLoc :: P SrcLoc
getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc

1201
setLastToken :: SrcSpan -> Int -> P ()
1202
setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } ()
1203

1204
data AlexInput = AI SrcLoc {-#UNPACK#-}!Int StringBuffer
1205 1206

alexInputPrevChar :: AlexInput -> Char
1207
alexInputPrevChar (AI _ _ buf) = prevChar buf '\n'
1208 1209

alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1210
alexGetChar (AI loc ofs s) 
1211
  | atEnd s   = Nothing
1212
  | otherwise = adj_c `seq` loc' `seq` ofs' `seq` s' `seq` 
1213
		--trace (show (ord c)) $
1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225 1226 1227
		Just (adj_c, (AI loc' ofs' s'))
  where (c,s') = nextChar s
        loc'   = advanceSrcLoc loc c
        ofs'   = advanceOffs c ofs

	non_graphic     = '\x0'
	upper 		= '\x1'
	lower		= '\x2'
	digit		= '\x3'
	symbol		= '\x4'
	space		= '\x5'
	other_graphic	= '\x6'

	adj_c 
1228
	  | c <= '\x06' = non_graphic
1229 1230 1231 1232 1233 1234 1235 1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255 1256 1257 1258 1259 1260 1261 1262
	  | c <= '\xff' = c
	  | otherwise = 
		case generalCategory c of
		  UppercaseLetter       -> upper
		  LowercaseLetter       -> lower
		  TitlecaseLetter       -> upper
		  ModifierLetter        -> other_graphic
		  OtherLetter           -> other_graphic
		  NonSpacingMark        -> other_graphic
		  SpacingCombiningMark  -> other_graphic
		  EnclosingMark         -> other_graphic
		  DecimalNumber         -> digit
		  LetterNumber          -> other_graphic
		  OtherNumber           -> other_graphic
		  ConnectorPunctuation  -> other_graphic
		  DashPunctuation       -> other_graphic
		  OpenPunctuation       -> other_graphic
		  ClosePunctuation      -> other_graphic
		  InitialQuote          -> other_graphic
		  FinalQuote            -> other_graphic
		  OtherPunctuation      -> other_graphic
		  MathSymbol            -> symbol
		  CurrencySymbol        -> symbol
		  ModifierSymbol        -> symbol
		  OtherSymbol           -> symbol
		  Space                 -> space
		  _other		-> non_graphic

-- This version does not squash unicode characters, it is used when
-- lexing strings.
alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
alexGetChar' (AI loc ofs s) 
  | atEnd s   = Nothing
  | otherwise = c `seq` loc' `seq` ofs' `seq` s' `seq` 
1263
		--trace (show (ord c)) $
1264 1265 1266 1267
		Just (c, (AI loc' ofs' s'))
  where (c,s') = nextChar s
        loc'   = advanceSrcLoc loc c
        ofs'   = advanceOffs c ofs
1268

1269 1270 1271 1272
advanceOffs :: Char -> Int -> Int
advanceOffs '\n' offs = 0
advanceOffs '\t' offs = (offs `quot` 8 + 1) * 8
advanceOffs _    offs = offs + 1
1273

1274
getInput :: P AlexInput
1275
getInput = P $ \s@PState{ loc=l, last_offs=o, buffer=b } -> POk s (AI l o b)
1276 1277

setInput :: AlexInput -> P ()
1278
setInput (AI l o b) = P $ \s -> POk s{ loc=l, last_offs=o, buffer=b } ()
1279 1280 1281 1282 1283 1284 1285 1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297

pushLexState :: Int -> P ()
pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()

popLexState :: P Int
popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls

getLexState :: P Int
getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls

-- for reasons of efficiency, flags indicating language extensions (eg,
-- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
-- integer

glaExtsBit, ffiBit, parrBit :: Int
glaExtsBit = 0
ffiBit	   = 1
parrBit	   = 2
arrowsBit  = 4
1298 1299
thBit	   = 5
ipBit      = 6
1300
tvBit	   = 7	-- Scoped type variables enables 'forall' keyword
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
1301 1302
bangPatBit = 8	-- Tells the parser to understand bang-patterns
		-- (doesn't affect the lexer)
1303 1304 1305 1306 1307 1308

glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
glaExtsEnabled flags = testBit flags glaExtsBit
ffiEnabled     flags = testBit flags ffiBit
parrEnabled    flags = testBit flags parrBit
arrowsEnabled  flags = testBit flags arrowsBit
1309 1310
thEnabled      flags = testBit flags thBit
ipEnabled      flags = testBit flags ipBit
1311
tvEnabled      flags = testBit flags tvBit
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
1312
bangPatEnabled flags = testBit flags bangPatBit
1313

1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329
-- PState for parsing options pragmas
--
pragState :: StringBuffer -> SrcLoc -> PState
pragState buf loc  = 
  PState {
      buffer	 = buf,
      last_loc   = mkSrcSpan loc loc,
      last_offs  = 0,
      last_len   = 0,
      loc        = loc,
      extsBitmap = 0,
      context    = [],
      lex_state  = [bol, option_prags, 0]
    }


1330 1331
-- create a parse state
--
1332 1333
mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
mkPState buf loc flags  = 
1334 1335
  PState {
      buffer	 = buf,
1336
      last_loc   = mkSrcSpan loc loc,
1337
      last_offs  = 0,
1338 1339 1340 1341 1342 1343 1344 1345
      last_len   = 0,
      loc        = loc,
      extsBitmap = fromIntegral bitmap,
      context    = [],
      lex_state  = [bol, if glaExtsEnabled bitmap then glaexts else 0]
	-- we begin in the layout state if toplev_layout is set
    }
    where
1346 1347 1348 1349 1350 1351
      bitmap =     glaExtsBit `setBitIf` dopt Opt_GlasgowExts flags
	       .|. ffiBit     `setBitIf` dopt Opt_FFI         flags
	       .|. parrBit    `setBitIf` dopt Opt_PArr        flags
	       .|. arrowsBit  `setBitIf` dopt Opt_Arrows      flags
	       .|. thBit      `setBitIf` dopt Opt_TH          flags
	       .|. ipBit      `setBitIf` dopt Opt_ImplicitParams flags
1352
	       .|. tvBit      `setBitIf` dopt Opt_ScopedTypeVariables flags
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
1353
	       .|. bangPatBit `setBitIf` dopt Opt_BangPatterns flags
1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
      --
      setBitIf :: Int -> Bool -> Int
      b `setBitIf` cond | cond      = bit b
			| otherwise = 0

getContext :: P [LayoutContext]
getContext = P $ \s@PState{context=ctx} -> POk s ctx

setContext :: [LayoutContext] -> P ()
setContext ctx = P $ \s -> POk s{context=ctx} ()

popContext :: P ()
popContext = P $ \ s@(PState{ buffer = buf, context = ctx, 
1367
			   loc = loc, last_len = len, last_loc = last_loc }) ->
1368 1369
  case ctx of
	(_:tl) -> POk s{ context = tl } ()
1370
	[]     -> PFailed last_loc (srcParseErr buf len)
1371 1372 1373 1374 1375

-- Push a new layout context at the indentation of the last token read.
-- This is only used at the outer level of a module when the 'module'
-- keyword is missing.
pushCurrentContext :: P ()
1376 1377
pushCurrentContext = P $ \ s@PState{ last_offs=offs, last_len=len, context=ctx } ->
  POk s{context = Layout (offs-len) : ctx} ()
1378

1379 1380
getOffside :: P Ordering
getOffside = P $ \s@PState{last_offs=offs, context=stk} ->
1381
		let ord = case stk of
1382
			(Layout n:_) -> compare offs n
1383
			_            -> GT
1384
		in POk s ord
1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396 1397 1398

-- ---------------------------------------------------------------------------
-- Construct a parse error

srcParseErr
  :: StringBuffer	-- current buffer (placed just after the last token)
  -> Int		-- length of the previous token
  -> Message
srcParseErr buf len
  = hcat [ if null token 
	     then ptext SLIT("parse error (possibly incorrect indentation)")
	     else hcat [ptext SLIT("parse error on input "),
          	  	char '`', text token, char '\'']
    ]
1399
  where token = lexemeToString (offsetBytes (-len) buf) len
1400 1401 1402 1403 1404 1405

-- Report a parse failure, giving the span of the previous token as
-- the location of the error.  This is the entry point for errors
-- detected during parsing.
srcParseFail :: P a
srcParseFail = P $ \PState{ buffer = buf, last_len = len, 	
1406
			    last_loc = last_loc } ->
1407
    PFailed last_loc (srcParseErr buf len)
1408 1409

-- A lexical error is reported at a particular position in the source file,
1410
-- not over a token range.
1411 1412 1413
lexError :: String -> P a
lexError str = do
  loc <- getSrcLoc
1414
  i@(AI end _ buf) <- getInput
1415
  reportLexError loc end buf str
1416 1417 1418 1419 1420

-- -----------------------------------------------------------------------------
-- This is the top-level function: called from the parser each time a
-- new token is to be read from the input.

1421
lexer :: (Located Token -> P a) -> P a
1422
lexer cont = do<