Lexer.x 65.3 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
--
-- 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:
--    - 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

22 23 24 25 26 27 28 29 30 31 32
--   qualified operator NOTES.
--   
--   - If M.(+) is a single lexeme, then..
--     - Probably (+) should be a single lexeme too, for consistency.
--       Otherwise ( + ) would be a prefix operator, but M.( + ) would not be.
--     - But we have to rule out reserved operators, otherwise (..) becomes
--       a different lexeme.
--     - Should we therefore also rule out reserved operators in the qualified
--       form?  This is quite difficult to achieve.  We don't do it for
--       qualified varids.

33
{
34
{-# OPTIONS -Wwarn -w #-}
35 36 37
-- 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
38
--     http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
39
-- for details
twanvl's avatar
twanvl committed
40 41 42
--
-- Note that Alex itself generates code with with some unused bindings and
-- without type signatures, so removing the flag might not be possible.
43

44 45
{-# OPTIONS_GHC -funbox-strict-fields #-}

46
module Lexer (
47
   Token(..), lexer, pragState, mkPState, PState(..),
48
   P(..), ParseResult(..), getSrcLoc, 
49
   failLocMsgP, failSpanMsgP, srcParseFail,
50
   getMessages,
51
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
52
   getLexState, popLexState, pushLexState,
53
   extension, standaloneDerivingEnabled, bangPatEnabled,
Jedai's avatar
Jedai committed
54 55
   addWarning,
   lexTokenStream
56 57
  ) where

58 59
import Bag
import ErrUtils
60
import Maybe
61 62 63 64 65
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
66
import DynFlags
67
import Ctype
68
import Util		( readRational )
69

70
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
71
import Data.Bits
72
import Data.Char
73 74 75
import Data.List
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
76
import Data.Ratio
77 78
}

79
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
80
$whitechar   = [\ \n\r\f\v $unispace]
81
$white_no_nl = $whitechar # \n
82
$tab         = \t
83 84

$ascdigit  = 0-9
85
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
86
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
87 88 89
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
90
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
91
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
92 93
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

94
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
95
$asclarge  = [A-Z]
96 97
$large     = [$asclarge $unilarge]

98
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
99
$ascsmall  = [a-z]
100 101
$small     = [$ascsmall $unismall \_]

102
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
103
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
104 105

$octit	   = 0-7
106
$hexit     = [$decdigit A-F a-f]
107 108 109 110
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

111 112
$pragmachar = [$small $large $digit]

113 114
$docsym    = [\| \^ \* \$]

115 116 117 118 119 120
@varid     = $small $idchar*
@conid     = $large $idchar*

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

121
@decimal     = $decdigit+
122 123 124 125 126 127 128 129 130
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

131 132 133 134 135
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

136 137 138 139
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
140
$tab+         { warn Opt_WarnTabs (text "Tab character") }
141 142 143 144 145

-- 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
146 147 148 149
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
150 151 152 153

-- 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.
154 155 156 157 158 159 160 161 162 163

-- 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.

Jedai's avatar
Jedai committed
164 165
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
166 167 168

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

Jedai's avatar
Jedai committed
169
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
170 171 172 173 174 175 176

-- 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.

Jedai's avatar
Jedai committed
177
"---"\-* [^$symbol :] .* { lineCommentToken }
178 179 180 181

-- 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.

Jedai's avatar
Jedai committed
182
"--"\-* / { atEOL } { lineCommentToken }
183 184 185 186

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

Jedai's avatar
Jedai committed
187
"-- " / { atEOL } { lineCommentToken }
188 189 190 191 192 193 194 195 196 197 198 199 200

-- '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 }
201
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
202
  ^\# \! .* \n				; -- #!, for scripts
203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224
  ()					{ 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 }

225
<0,option_prags> \n				{ begin bol }
226

227 228
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
229 230 231

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
232
<line_prag1> $decdigit+			{ setLine line_prag1a }
233
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
234 235 236 237
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
238
<line_prag2> $decdigit+			{ setLine line_prag2a }
239
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
240 241 242
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
243

244
<0,option_prags> {
245 246 247 248 249 250
  "{-#" $whitechar* $pragmachar+ 
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
251

252
  -- We ignore all these pragmas, but don't generate a warning for them
253 254
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
255 256

  -- ToDo: should only be valid inside a pragma:
257
  "#-}" 				{ endPrag }
258 259
}

260
<option_prags> {
261 262 263
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

David Waern's avatar
David Waern committed
264 265 266
  "-- #"                                 { multiline_doc_comment }
}

267 268
<0> {
  -- In the "0" mode we ignore these pragmas
269
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
270 271 272
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
273
<0> {
Jedai's avatar
Jedai committed
274
  "-- #" .* { lineCommentToken }
275
}
276

277
<0,option_prags> {
278 279
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
280 281
}

282 283
-- '0' state: ordinary lexemes

284 285
-- Haddock comments

286
<0> {
David Waern's avatar
David Waern committed
287 288
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
289 290
}

291 292
-- "special" symbols

293
<0> {
294 295 296 297
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
298
<0> {
299 300 301 302 303 304 305 306
  "[|"	    / { 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 }
307 308 309

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
310 311
}

312
<0> {
ross's avatar
ross committed
313 314
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
315 316 317
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

318
<0> {
319 320 321
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

322
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
323 324 325 326 327 328
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

329
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
330 331
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
332 333
}

334
<0,option_prags> {
335 336 337 338 339 340 341 342 343 344 345 346
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

347
<0,option_prags> {
348
  @qual @varid			{ idtoken qvarid }
349 350 351 352 353
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

354
<0> {
355 356 357 358
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
359 360
}

361 362
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
363
<0> {
364 365 366 367 368 369
  @qual @varsym       / { ifExtension oldQualOps } { idtoken qvarsym }
  @qual @consym       / { ifExtension oldQualOps } { idtoken qconsym }
  @qual \( @varsym \) / { ifExtension newQualOps } { idtoken prefixqvarsym }
  @qual \( @consym \) / { ifExtension newQualOps } { idtoken prefixqconsym }
  @varsym                                          { varsym }
  @consym                                          { consym }
370 371
}

372 373
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
374
<0> {
375
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
376 377 378
  @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 }
379 380

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
381
  @floating_point    { strtoken tok_float }
382 383
}

384
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
385
  -- Unboxed ints (:: Int#) and words (:: Word#)
386 387
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
388 389 390 391 392
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
  0[oO] @octal                 \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
  0[xX] @hexadecimal           \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
  @negative @decimal           \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
393
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
394

Ian Lynagh's avatar
Ian Lynagh committed
395 396 397 398
  @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 }

399 400
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
401 402
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
403 404 405 406 407 408
}

-- 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.
409
<0> {
410 411 412 413 414 415 416 417
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

418
data Token
419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454
  = 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
455
  | ITprimcallconv
456 457
  | ITdotnet
  | ITmdo
458
  | ITfamily
459 460 461
  | ITgroup
  | ITby
  | ITusing
462

463 464
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
465
  | ITinline_conlike_prag Bool  -- same
466 467
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
468 469
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
470
  | ITwarning_prag
471 472 473
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
474
  | ITgenerated_prag
475
  | ITcore_prag                 -- hdaume: core annotations
476
  | ITunpack_prag
477
  | ITann_prag
478
  | ITclose_prag
479 480 481
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
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

  | 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
508 509
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527
  | 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)
528 529
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
530 531 532 533 534 535 536 537 538 539 540 541 542

  | 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
543
  | ITprimword   Integer
544 545 546 547
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
548 549 550 551 552 553 554 555 556
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
557
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
558 559 560 561

  -- Arrow notation extension
  | ITproc
  | ITrec
562 563 564 565 566 567
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
568 569 570

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
571 572 573 574 575 576 577

  -- 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
578
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
579 580
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
581

582 583 584 585
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
586
{-
587
isSpecial :: Token -> Bool
588 589 590 591 592 593 594 595 596 597 598 599 600 601 602
-- 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
603
isSpecial ITprimcallconv = True
604
isSpecial ITmdo		= True
605
isSpecial ITfamily	= True
606 607 608
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
609
isSpecial _             = False
twanvl's avatar
twanvl committed
610
-}
611 612 613 614 615 616 617 618

-- 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)
--
Ian Lynagh's avatar
Ian Lynagh committed
619
reservedWordsFM :: UniqFM (Token, Int)
620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648
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

649
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
650
	( "mdo",	ITmdo,		 bit recursiveDoBit),
651
	( "family",	ITfamily,	 bit tyFamBit),
652 653 654
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
655 656 657 658 659 660

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
661
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),  -- ToDo: remove
662 663 664
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
665
	( "prim",       ITprimcallconv,	 bit ffiBit),
666 667 668
	( "dotnet",     ITdotnet,	 bit ffiBit),

	( "rec",	ITrec,		 bit arrowsBit),
669
	( "proc",	ITproc,		 bit arrowsBit)
670 671
     ]

Ian Lynagh's avatar
Ian Lynagh committed
672
reservedSymsFM :: UniqFM (Token, Int -> Bool)
673
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690
    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
691
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
692
        -- For 'forall a . t'
693
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
694 695 696 697 698

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
699

Ian Lynagh's avatar
Ian Lynagh committed
700 701
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
702 703
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
704 705 706
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
707 708 709
        -- 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).
710 711 712 713 714
       ]

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

715
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
716

717
special :: Token -> Action
twanvl's avatar
twanvl committed
718
special tok span _buf _len = return (L span tok)
719

720
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
721 722
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
723

724 725
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
726

727 728 729
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
730

731 732 733
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
734

735
init_strtoken :: Int -> (String -> Token) -> Action
736
-- like strtoken, but drops the last N character(s)
737 738
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
739 740

begin :: Int -> Action
741
begin code _span _str _len = do pushLexState code; lexToken
742 743

pop :: Action
744
pop _span _buf _len = do popLexState; lexToken
745 746

pop_and :: Action -> Action
747
pop_and act span buf len = do popLexState; act span buf len
748

749
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
750
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
751 752
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
753
notFollowedBy :: Char -> AlexAccPred Int
754 755
notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
756

Ian Lynagh's avatar
Ian Lynagh committed
757
notFollowedBySymbol :: AlexAccPred Int
758
notFollowedBySymbol _ _ _ (AI _ _ buf)
759
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
760

761 762 763 764 765
-- 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.
Ian Lynagh's avatar
Ian Lynagh committed
766
isNormalComment :: AlexAccPred Int
767
isNormalComment bits _ _ (AI _ _ buf)
768 769 770 771 772 773
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
774
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
775
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
776

twanvl's avatar
twanvl committed
777
{-
778 779
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
780
-}
781

Ian Lynagh's avatar
Ian Lynagh committed
782
atEOL :: AlexAccPred Int
783
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
784

Ian Lynagh's avatar
Ian Lynagh committed
785
ifExtension :: (Int -> Bool) -> AlexAccPred Int
786 787
ifExtension pred bits _ _ _ = pred bits

788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815
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

Jedai's avatar
Jedai committed
816 817 818 819 820
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

821 822 823 824
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
825 826
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
827
  input <- getInput
Jedai's avatar
Jedai committed
828
  go "" (1::Int) input
829
  where
Jedai's avatar
Jedai committed
830 831 832 833 834 835
    go commentAcc 0 input = do setInput input
                               b <- extension rawTokenStreamEnabled
                               if b
                                 then docCommentEnd input commentAcc ITblockComment _str span
                                 else cont
    go commentAcc n input = case alexGetChar input of
836 837 838
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
839 840
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
841 842
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
843 844 845
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
846 847 848 849 850 851 852 853

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
854
        Just ('\125',input) ->
855
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
856
        Just (_,_) -> go ('-':commentAcc) input docType False
857 858 859 860 861 862
      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
863
        Just (_,_) -> go ('\123':commentAcc) input docType False
864 865
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
866 867
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
                 -> P (Located Token)
868
withLexedDocType lexDocComment = do
869 870 871 872 873
  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
874 875
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
876
    _ -> panic "withLexedDocType: Bad doc type"
877 878 879
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
880
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
881 882
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

883 884 885
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
Ian Lynagh's avatar
Ian Lynagh committed
886
rulePrag span _ _ = do
887
  setExts (.|. bit inRulePragBit)
888 889 890
  return (L span ITrules_prag)

endPrag :: Action
Ian Lynagh's avatar
Ian Lynagh committed
891
endPrag span _ _ = do
892 893 894
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926
-- 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))
 
Ian Lynagh's avatar
Ian Lynagh committed
927
errBrace :: AlexInput -> SrcSpan -> P a
928
errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
929

930
open_brace, close_brace :: Action
931
open_brace span _str _len = do 
932 933
  ctx <- getContext
  setContext (NoLayout:ctx)
934 935
  return (L span ITocurly)
close_brace span _str _len = do 
936
  popContext
937
  return (L span ITccurly)
938

Ian Lynagh's avatar
Ian Lynagh committed
939
qvarid, qconid :: StringBuffer -> Int -> Token
940 941
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
942

943
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
944 945 946
-- 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.
947
splitQualName orig_buf len parens = split orig_buf orig_buf
948
  where
949 950 951 952 953 954
    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
955 956 957 958
  
    -- 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.
959 960 961 962 963 964 965 966
    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),
967 968 969
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
970 971
      where
	qual_size = orig_buf `byteDiff` dot_buf
972

Ian Lynagh's avatar
Ian Lynagh committed
973 974
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
975
  fs `seq`
976 977 978
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
979
		return (L span keyword)
980 981 982
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
983 984 985
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
986 987 988
  where
	fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
989
conid :: StringBuffer -> Int -> Token
990 991 992
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
993
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
994 995 996 997
qvarsym buf len = ITqvarsym $! splitQualName buf len False
qconsym buf len = ITqconsym $! splitQualName buf len False
prefixqvarsym buf len = ITprefixqvarsym $! splitQualName buf len True
prefixqconsym buf len = ITprefixqconsym $! splitQualName buf len True
998

Ian Lynagh's avatar
Ian Lynagh committed
999
varsym, consym :: Action
1000 1001 1002
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
1003 1004
sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
    -> P (Located Token)
1005
sym con span buf len = 
1006 1007
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
1008
		b <- extension exts
1009 1010 1011
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
1012 1013 1014
  where
	fs = lexemeToFastString buf len

1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025
-- 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
Ian Lynagh's avatar
Ian Lynagh committed
1026 1027 1028
tok_num :: (Integer -> Integer)
        -> Int -> Int
        -> (Integer, (Char->Int)) -> Action
1029
tok_num = tok_integral ITinteger
Ian Lynagh's avatar
Ian Lynagh committed
1030 1031 1032
tok_primint :: (Integer -> Integer)
            -> Int -> Int
            -> (Integer, (Char->Int)) -> Action
1033
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1034 1035
tok_primword :: Int -> Int
             -> (Integer, (Char->Int)) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
1036
tok_primword = tok_integral ITprimword positive
Ian Lynagh's avatar
Ian Lynagh committed
1037
positive, negative :: (Integer -> Integer)
1038 1039
positive = id
negative = negate
Ian Lynagh's avatar
Ian Lynagh committed
1040
decimal, octal, hexadecimal :: (Integer, Char -> Int)
1041 1042 1043 1044 1045
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
Ian Lynagh's avatar
Ian Lynagh committed
1046
tok_float, tok_primfloat, tok_primdouble :: String -> Token
1047
tok_float        str = ITrational   $! readRational str
1048 1049
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
1050 1051 1052 1053 1054 1055

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1056
do_bol span _str _len = do
1057
	pos <- getOffside
1058 1059 1060 1061 1062
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1063
		return (L span ITvccurly)
1064 1065 1066
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1067
		return (L span ITsemi)
1068 1069 1070 1071 1072 1073
	    GT -> do
		popLexState
		lexToken

-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
Ian Lynagh's avatar
Ian Lynagh committed
1074
maybe_layout :: Token -> P ()
1075
maybe_layout ITdo	= pushLexState layout_do
1076
maybe_layout ITmdo	= pushLexState layout_do
1077 1078 1079
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1080
maybe_layout ITrec	= pushLexState layout
1081 1082 1083 1084 1085 1086 1087 1088 1089 1090 1091
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.
--
Ian Lynagh's avatar
Ian Lynagh committed
1092
new_layout_context :: Bool -> Action
1093
new_layout_context strict span _buf _len = do
1094
    popLexState
1095
    (AI _ offset _) <- getInput
1096 1097 1098 1099 1100 1101 1102 1103
    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
1104
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1105
	_ -> do
1106
		setContext (Layout offset : ctx)
1107
		return (L span ITvocurly)
1108

Ian Lynagh's avatar
Ian Lynagh committed
1109
do_layout_left :: Action
1110
do_layout_left span _buf _len = do
1111 1112
    popLexState
    pushLexState bol  -- we must be at the start of a line
1113
    return (L span ITvccurly)
1114 1115 1116 1117

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

1118 1119
setLine :: Int -> Action
setLine code span buf len = do
1120
  let line = parseUnsignedInteger buf len 10 octDecDigit
1121
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1122 1123 1124 1125 1126
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1127 1128
setFile :: Int -> Action
setFile code span buf len = do
1129
  let file = lexemeToFastString (stepOn buf) (len-2)
1130
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1131 1132 1133 1134
  popLexState
  pushLexState code
  lexToken

1135 1136 1137 1138 1139

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1140
lex_string_prag mkTok span _buf _len
1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152
    = 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
1153
          isString _ [] = True
1154 1155 1156 1157 1158 1159 1160
          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"


1161 1162 1163 1164 1165 1166
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1167
lex_string_tok span _buf _len = do