Lexer.x 65.1 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
-- The above -Wwarn supression flag is a temporary kludge.
36 37
-- 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
   getPState,
50
   failLocMsgP, failSpanMsgP, srcParseFail,
51
   getMessages,
52
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
53
   getLexState, popLexState, pushLexState,
54
   extension, standaloneDerivingEnabled, bangPatEnabled,
Jedai's avatar
Jedai committed
55 56
   addWarning,
   lexTokenStream
57 58
  ) where

59 60
import Bag
import ErrUtils
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
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
74
import Data.Maybe
75 76
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
77
import Data.Ratio
78 79
}

80
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
81
$whitechar   = [\ \n\r\f\v $unispace]
82
$white_no_nl = $whitechar # \n
83
$tab         = \t
84 85

$ascdigit  = 0-9
86
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
87
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
88 89 90
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
91
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
92
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
93 94
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

95
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
96
$asclarge  = [A-Z]
97 98
$large     = [$asclarge $unilarge]

99
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
100
$ascsmall  = [a-z]
101 102
$small     = [$ascsmall $unismall \_]

103
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
104
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
105 106

$octit	   = 0-7
107
$hexit     = [$decdigit A-F a-f]
108 109 110 111
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

112 113
$pragmachar = [$small $large $digit]

114 115
$docsym    = [\| \^ \* \$]

116 117 118 119 120 121
@varid     = $small $idchar*
@conid     = $large $idchar*

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

122
@decimal     = $decdigit+
123 124 125 126 127 128 129 130 131
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

132 133 134 135 136
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

137 138 139 140
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
141
$tab+         { warn Opt_WarnTabs (text "Tab character") }
142 143 144 145 146

-- 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
147 148 149 150
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
151 152 153 154

-- 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.
155 156 157 158 159 160 161 162 163 164

-- 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
165 166
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
167 168 169

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

Jedai's avatar
Jedai committed
170
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
171 172 173 174 175 176 177

-- 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
178
"---"\-* [^$symbol :] .* { lineCommentToken }
179 180 181 182

-- 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
183
"--"\-* / { atEOL } { lineCommentToken }
184 185 186 187

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

Jedai's avatar
Jedai committed
188
"-- " / { atEOL } { lineCommentToken }
189 190 191 192 193 194 195 196 197 198 199 200 201

-- '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 }
202
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
203
  ^\# \! .* \n				; -- #!, for scripts
204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225
  ()					{ 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 }

226
<0,option_prags> \n				{ begin bol }
227

228 229
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
230 231 232

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
233
<line_prag1> $decdigit+			{ setLine line_prag1a }
234
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
235 236 237 238
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
239
<line_prag2> $decdigit+			{ setLine line_prag2a }
240
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
241 242 243
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
244

245
<0,option_prags> {
246 247 248 249 250 251
  "{-#" $whitechar* $pragmachar+ 
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
252

253
  -- We ignore all these pragmas, but don't generate a warning for them
254 255
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
256 257

  -- ToDo: should only be valid inside a pragma:
258
  "#-}" 				{ endPrag }
259 260
}

261
<option_prags> {
262 263 264
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

David Waern's avatar
David Waern committed
265 266 267
  "-- #"                                 { multiline_doc_comment }
}

268 269
<0> {
  -- In the "0" mode we ignore these pragmas
270
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
271 272 273
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
274
<0> {
Jedai's avatar
Jedai committed
275
  "-- #" .* { lineCommentToken }
276
}
277

278
<0,option_prags> {
279 280
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
281 282
}

283 284
-- '0' state: ordinary lexemes

285 286
-- Haddock comments

287
<0> {
David Waern's avatar
David Waern committed
288 289
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
290 291
}

292 293
-- "special" symbols

294
<0> {
295 296 297 298
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
299
<0> {
300 301 302 303 304 305 306 307
  "[|"	    / { 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 }
308 309 310

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
311 312
}

313
<0> {
ross's avatar
ross committed
314 315
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
316 317 318
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

319
<0> {
320 321 322
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

323
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
324 325 326 327 328 329
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

330
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
331 332
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
333 334
}

335
<0,option_prags> {
336 337 338 339 340 341 342 343 344 345 346 347
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

348
<0,option_prags> {
349
  @qual @varid			{ idtoken qvarid }
350 351 352 353 354
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

355
<0> {
356 357 358 359
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
360 361
}

362 363
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
364
<0> {
365 366 367 368 369 370
  @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 }
371 372
}

373 374
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
375
<0> {
376
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
377 378 379
  @decimal           { tok_num positive 0 0 decimal }
  0[oO] @octal       { tok_num positive 2 2 octal }
  0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal }
380 381

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

385
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
386
  -- Unboxed ints (:: Int#) and words (:: Word#)
387 388
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
389 390 391 392 393
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
  0[oO] @octal                 \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
  0[xX] @hexadecimal           \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
  @negative @decimal           \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
394
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
395

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

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

-- Strings and chars are lexed by hand-written code.  The reason is
-- that even if we recognise the string or char here in the regex
-- lexer, we would still have to parse the string afterward in order
-- to convert it to a String.
410
<0> {
411 412 413 414 415 416 417 418
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

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

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

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

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
541
  | ITprimword   Integer
542 543 544
  | ITprimfloat  Rational
  | ITprimdouble Rational

545
  -- Template Haskell extension tokens
546 547 548 549 550 551 552 553 554
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
555
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
556 557 558 559

  -- Arrow notation extension
  | ITproc
  | ITrec
560 561 562 563 564 565
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
566 567 568

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
569 570 571 572 573 574 575

  -- 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
576
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
577 578
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
579

580 581 582 583
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
584
{-
585
isSpecial :: Token -> Bool
586 587 588 589 590 591 592 593 594 595 596 597 598 599 600
-- 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
601
isSpecial ITprimcallconv = True
602
isSpecial ITmdo		= True
603
isSpecial ITfamily	= True
604 605 606
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
607
isSpecial _             = False
twanvl's avatar
twanvl committed
608
-}
609 610 611 612 613 614 615 616

-- 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
617
reservedWordsFM :: UniqFM (Token, Int)
618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646
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

647
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
648
	( "mdo",	ITmdo,		 bit recursiveDoBit),
649
	( "family",	ITfamily,	 bit tyFamBit),
650 651 652
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
653 654 655 656 657 658

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
659
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),  -- ToDo: remove
660 661 662
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
663
	( "prim",       ITprimcallconv,	 bit ffiBit),
664

665
	( "rec",	ITrec,		 bit recBit),
666
	( "proc",	ITproc,		 bit arrowsBit)
667 668
     ]

Ian Lynagh's avatar
Ian Lynagh committed
669
reservedSymsFM :: UniqFM (Token, Int -> Bool)
670
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687
    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
688
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
689
        -- For 'forall a . t'
690
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
691 692 693 694 695

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
696

Ian Lynagh's avatar
Ian Lynagh committed
697 698
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
699 700
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
701 702 703
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
704 705 706
        -- 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).
707 708 709 710 711
       ]

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

712
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
713

714
special :: Token -> Action
twanvl's avatar
twanvl committed
715
special tok span _buf _len = return (L span tok)
716

717
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
718 719
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
720

721 722
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
723

724 725 726
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
727

728 729 730
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
731

732
init_strtoken :: Int -> (String -> Token) -> Action
733
-- like strtoken, but drops the last N character(s)
734 735
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
736 737

begin :: Int -> Action
738
begin code _span _str _len = do pushLexState code; lexToken
739 740

pop :: Action
741
pop _span _buf _len = do popLexState; lexToken
742 743

pop_and :: Action -> Action
744
pop_and act span buf len = do popLexState; act span buf len
745

746
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
747
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
748 749
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
750
notFollowedBy :: Char -> AlexAccPred Int
751 752
notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
753

Ian Lynagh's avatar
Ian Lynagh committed
754
notFollowedBySymbol :: AlexAccPred Int
755
notFollowedBySymbol _ _ _ (AI _ _ buf)
756
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
757

758 759 760 761 762
-- 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
763
isNormalComment :: AlexAccPred Int
764
isNormalComment bits _ _ (AI _ _ buf)
765 766 767 768 769 770
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
771
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
772
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
773

twanvl's avatar
twanvl committed
774
{-
775 776
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
777
-}
778

Ian Lynagh's avatar
Ian Lynagh committed
779
atEOL :: AlexAccPred Int
780
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
781

Ian Lynagh's avatar
Ian Lynagh committed
782
ifExtension :: (Int -> Bool) -> AlexAccPred Int
783 784
ifExtension pred bits _ _ _ = pred bits

785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812
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
813 814 815 816 817
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

818 819 820 821
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
822 823
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
824
  input <- getInput
Jedai's avatar
Jedai committed
825
  go "" (1::Int) input
826
  where
Jedai's avatar
Jedai committed
827 828 829 830 831 832
    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
833 834 835
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
836 837
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
838 839
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
840 841 842
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
843 844 845 846 847 848 849 850

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
851
        Just ('\125',input) ->
852
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
853
        Just (_,_) -> go ('-':commentAcc) input docType False
854 855 856 857 858 859
      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
860
        Just (_,_) -> go ('\123':commentAcc) input docType False
861 862
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
863 864
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
                 -> P (Located Token)
865
withLexedDocType lexDocComment = do
866 867 868 869 870
  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
871 872
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
873
    _ -> panic "withLexedDocType: Bad doc type"
874 875 876
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
877
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
878 879
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

880 881 882
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
883
rulePrag span _buf _len = do
884
  setExts (.|. bit inRulePragBit)
885 886 887
  return (L span ITrules_prag)

endPrag :: Action
888
endPrag span _buf _len = do
889 890 891
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923
-- 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
924
errBrace :: AlexInput -> SrcSpan -> P a
925
errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
926

927
open_brace, close_brace :: Action
928
open_brace span _str _len = do 
929 930
  ctx <- getContext
  setContext (NoLayout:ctx)
931 932
  return (L span ITocurly)
close_brace span _str _len = do 
933
  popContext
934
  return (L span ITccurly)
935

Ian Lynagh's avatar
Ian Lynagh committed
936
qvarid, qconid :: StringBuffer -> Int -> Token
937 938
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
939

940
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
941 942 943
-- 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.
944
splitQualName orig_buf len parens = split orig_buf orig_buf
945
  where
946 947 948 949 950 951
    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
952 953 954 955
  
    -- 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.
956 957 958 959 960 961 962 963
    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),
964 965 966
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
967 968
      where
	qual_size = orig_buf `byteDiff` dot_buf
969

Ian Lynagh's avatar
Ian Lynagh committed
970 971
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
972
  fs `seq`
973 974 975
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
976
		return (L span keyword)
977 978 979
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
980 981 982
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
983 984 985
  where
	fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
986
conid :: StringBuffer -> Int -> Token
987 988 989
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
990
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
991 992 993 994
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
995

Ian Lynagh's avatar
Ian Lynagh committed
996
varsym, consym :: Action
997 998 999
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
1000 1001
sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
    -> P (Located Token)
1002
sym con span buf len = 
1003 1004
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
1005
		b <- extension exts
1006 1007 1008
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
1009 1010 1011
  where
	fs = lexemeToFastString buf len

1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022
-- 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
1023 1024 1025
tok_num :: (Integer -> Integer)
        -> Int -> Int
        -> (Integer, (Char->Int)) -> Action
1026
tok_num = tok_integral ITinteger
Ian Lynagh's avatar
Ian Lynagh committed
1027 1028 1029
tok_primint :: (Integer -> Integer)
            -> Int -> Int
            -> (Integer, (Char->Int)) -> Action
1030
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1031 1032
tok_primword :: Int -> Int
             -> (Integer, (Char->Int)) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
1033
tok_primword = tok_integral ITprimword positive
Ian Lynagh's avatar
Ian Lynagh committed
1034
positive, negative :: (Integer -> Integer)
1035 1036
positive = id
negative = negate
Ian Lynagh's avatar
Ian Lynagh committed
1037
decimal, octal, hexadecimal :: (Integer, Char -> Int)
1038 1039 1040 1041 1042
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
Ian Lynagh's avatar
Ian Lynagh committed
1043
tok_float, tok_primfloat, tok_primdouble :: String -> Token
1044
tok_float        str = ITrational   $! readRational str
1045 1046
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
1047 1048 1049 1050 1051 1052

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1053
do_bol span _str _len = do
1054
	pos <- getOffside
1055 1056 1057 1058 1059
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1060
		return (L span ITvccurly)
1061 1062 1063
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1064
		return (L span ITsemi)
1065 1066 1067 1068 1069 1070
	    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
1071
maybe_layout :: Token -> P ()
1072
maybe_layout ITdo	= pushLexState layout_do
1073
maybe_layout ITmdo	= pushLexState layout_do
1074 1075 1076
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1077
maybe_layout ITrec	= pushLexState layout
1078 1079 1080 1081 1082 1083 1084 1085 1086 1087 1088
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
1089
new_layout_context :: Bool -> Action
1090
new_layout_context strict span _buf _len = do
1091
    popLexState
1092
    (AI _ offset _) <- getInput
1093 1094 1095 1096 1097 1098 1099 1100
    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
1101
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1102
	_ -> do
1103
		setContext (Layout offset : ctx)
1104
		return (L span ITvocurly)
1105

Ian Lynagh's avatar
Ian Lynagh committed
1106
do_layout_left :: Action
1107
do_layout_left span _buf _len = do
1108 1109
    popLexState
    pushLexState bol  -- we must be at the start of a line
1110
    return (L span ITvccurly)
1111 1112 1113 1114

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

1115 1116
setLine :: Int -> Action
setLine code span buf len = do
1117
  let line = parseUnsignedInteger buf len 10 octDecDigit
1118
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1119 1120 1121 1122 1123
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1124 1125
setFile :: Int -> Action
setFile code span buf len = do
1126
  let file = lexemeToFastString (stepOn buf) (len-2)
1127
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1128 1129 1130 1131
  popLexState
  pushLexState code
  lexToken

1132 1133 1134 1135 1136

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1137
lex_string_prag mkTok span _buf _len
1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149
    = 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
1150
          isString _ [] = True
1151 1152 1153 1154 1155 1156 1157
          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"