Lexer.x 80.9 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9
--
-- 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
Ian Lynagh's avatar
Ian Lynagh committed
10 11
-- file is maintained.  Every token has a start and end RealSrcLoc
-- attached to it.
12 13 14 15 16 17 18 19 20 21 22
--
-----------------------------------------------------------------------------

--   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

23 24 25 26 27 28 29 30 31 32 33
--   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.

34
{
35
-- XXX The above flags turn off warnings in the generated code:
36
{-# LANGUAGE BangPatterns #-}
37 38 39 40 41 42 43 44
{-# OPTIONS_GHC -fno-warn-unused-matches #-}
{-# OPTIONS_GHC -fno-warn-unused-binds #-}
{-# OPTIONS_GHC -fno-warn-unused-imports #-}
{-# OPTIONS_GHC -fno-warn-missing-signatures #-}
-- But alex still generates some code that causes the "lazy unlifted bindings"
-- warning, and old compilers don't know about it so we can't easily turn
-- it off, so for now we use the sledge hammer:
{-# OPTIONS_GHC -w #-}
45

46 47
{-# OPTIONS_GHC -funbox-strict-fields #-}

48
module Lexer (
49
   Token(..), lexer, pragState, mkPState, PState(..),
50
   P(..), ParseResult(..), getSrcLoc, 
51
   getPState, getDynFlags, withThisPackage,
52
   failLocMsgP, failSpanMsgP, srcParseFail,
53
   getMessages, 
54
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
vivian's avatar
vivian committed
55
   activeContext, nextIsEOF,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
56
   getLexState, popLexState, pushLexState,
57
   extension, bangPatEnabled, datatypeContextsEnabled,
Jedai's avatar
Jedai committed
58 59
   addWarning,
   lexTokenStream
60 61
  ) where

62 63
import Bag
import ErrUtils
64 65 66 67 68
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
69
import DynFlags
70
import Module
71
import Ctype
72
import BasicTypes	( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) )
73
import Util		( readRational )
74

75
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
76
import Data.Bits
77
import Data.Char
78
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
79
import Data.Maybe
80 81
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
82
import Data.Ratio
83 84
}

85
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
86
$whitechar   = [\ \n\r\f\v $unispace]
87
$white_no_nl = $whitechar # \n
88
$tab         = \t
89 90

$ascdigit  = 0-9
91
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
92
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
93 94 95
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
96
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
97
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
98 99
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

100
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
101
$asclarge  = [A-Z]
102 103
$large     = [$asclarge $unilarge]

104
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
105
$ascsmall  = [a-z]
106 107
$small     = [$ascsmall $unismall \_]

108
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
109
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
110 111

$octit	   = 0-7
112
$hexit     = [$decdigit A-F a-f]
113 114 115 116
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

117 118
$pragmachar = [$small $large $digit]

119 120
$docsym    = [\| \^ \* \$]

121 122 123 124 125 126
@varid     = $small $idchar*
@conid     = $large $idchar*

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

127
@decimal     = $decdigit+
128 129 130 131 132 133 134 135 136
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

137 138 139 140 141
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

142 143 144 145
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
146
$tab+         { warn Opt_WarnTabs (text "Warning: Tab character") }
147 148 149 150 151

-- 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
152 153 154 155
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
156 157 158 159

-- 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.
160 161 162 163 164 165 166 167 168 169

-- 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
170 171
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
172 173 174

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

Jedai's avatar
Jedai committed
175
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
176 177 178 179 180 181 182

-- 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
183
"---"\-* [^$symbol :] .* { lineCommentToken }
184 185 186 187

-- 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
188
"--"\-* / { atEOL } { lineCommentToken }
189 190 191 192

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

Jedai's avatar
Jedai committed
193
"-- " / { atEOL } { lineCommentToken }
194 195 196 197 198 199 200 201 202 203 204 205 206

-- '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 }
207
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
208
  ^\# \! .* \n				; -- #!, for scripts
209 210 211 212 213 214 215
  ()					{ 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> {
216
  \{ / { notFollowedBy '-' }		{ hopefully_open_brace }
217 218 219 220 221 222 223 224 225 226 227 228 229 230
	-- 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 }

231
<0,option_prags> \n				{ begin bol }
232

233 234
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
235 236 237

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
238
<line_prag1> $decdigit+			{ setLine line_prag1a }
239
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
240 241 242 243
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
244
<line_prag2> $decdigit+			{ setLine line_prag2a }
245
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
246 247 248
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
249

250
<0,option_prags> {
251 252 253 254 255 256
  "{-#" $whitechar* $pragmachar+ 
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
257

258
  -- We ignore all these pragmas, but don't generate a warning for them
259 260
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
261 262

  -- ToDo: should only be valid inside a pragma:
263
  "#-}" 				{ endPrag }
264 265
}

266
<option_prags> {
267 268 269
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

David Waern's avatar
David Waern committed
270 271 272
  "-- #"                                 { multiline_doc_comment }
}

273 274
<0> {
  -- In the "0" mode we ignore these pragmas
275
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
276 277 278
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
279
<0> {
Jedai's avatar
Jedai committed
280
  "-- #" .* { lineCommentToken }
281
}
282

283
<0,option_prags> {
284 285
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
286 287
}

288 289
-- '0' state: ordinary lexemes

290 291
-- Haddock comments

292
<0,option_prags> {
David Waern's avatar
David Waern committed
293 294
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
295 296
}

297 298
-- "special" symbols

299
<0> {
300 301 302 303
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
304
<0> {
305 306 307 308 309 310 311 312
  "[|"	    / { ifExtension thEnabled }	{ token ITopenExpQuote }
  "[e|"	    / { ifExtension thEnabled }	{ token ITopenExpQuote }
  "[p|"	    / { ifExtension thEnabled }	{ token ITopenPatQuote }
  "[d|"	    / { ifExtension thEnabled }	{ layout_token ITopenDecQuote }
  "[t|"	    / { ifExtension thEnabled }	{ token ITopenTypQuote }
  "|]"	    / { ifExtension thEnabled }	{ token ITcloseQuote }
  \$ @varid / { ifExtension thEnabled }	{ skip_one_varid ITidEscape }
  "$("	    / { ifExtension thEnabled }	{ token ITparenEscape }
313

314 315 316 317
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

318
  "[" @varid "|"  / { ifExtension qqEnabled }
319
                     { lex_quasiquote_tok }
320 321
}

322
<0> {
ross's avatar
ross committed
323 324
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
325 326 327
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

328
<0> {
329 330 331
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

332
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
333 334 335 336 337 338
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

339
<0,option_prags> {
340 341 342 343 344 345 346 347 348 349 350 351
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

352
<0,option_prags> {
353
  @qual @varid			{ idtoken qvarid }
354 355 356 357 358
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

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

366 367
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
368
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
369 370
  @qual @varsym                                    { idtoken qvarsym }
  @qual @consym                                    { idtoken qconsym }
371 372
  @varsym                                          { varsym }
  @consym                                          { consym }
373 374
}

375 376
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
377
<0> {
378
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
379 380 381
  @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 }
382 383

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
384
  @floating_point    { strtoken tok_float }
385 386
}

387
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
388
  -- Unboxed ints (:: Int#) and words (:: Word#)
389 390
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
391 392 393 394 395
  @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 }
396
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
397

Ian Lynagh's avatar
Ian Lynagh committed
398 399 400 401
  @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 }

402 403
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
404 405
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
406 407 408 409 410 411
}

-- 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.
412
<0> {
413 414 415 416 417 418 419 420
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

421
data Token
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
455
  | ITinterruptible
456 457 458
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
459
  | ITprimcallconv
460
  | ITmdo
461
  | ITfamily
462 463 464
  | ITgroup
  | ITby
  | ITusing
465

466
	-- Pragmas
467
  | ITinline_prag InlineSpec RuleMatchInfo
468 469
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
470 471
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
472
  | ITwarning_prag
473 474 475
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
476
  | ITgenerated_prag
477
  | ITcore_prag                 -- hdaume: core annotations
478
  | ITunpack_prag
479
  | ITann_prag
480
  | ITclose_prag
481 482 483
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
484 485
  | ITvect_prag
  | ITvect_scalar_prag
486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511

  | 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
benl's avatar
benl committed
512 513
  | ITopabrack			-- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack			-- :], for parallel arrays with -XParallelArrays
514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
  | 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)
532 533
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
534 535 536 537 538 539

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

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
540
  | ITrational   FractionalLit
541 542 543 544

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
545
  | ITprimword   Integer
546 547
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
548

549
  -- Template Haskell extension tokens
550 551 552 553 554 555 556 557 558
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
Ian Lynagh's avatar
Ian Lynagh committed
559
  | ITquasiQuote (FastString,FastString,RealSrcSpan) --  [:...|...|]
560 561 562 563

  -- Arrow notation extension
  | ITproc
  | ITrec
564 565 566 567 568 569
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
570 571 572

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
573 574 575 576 577 578 579

  -- 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
580
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
581 582
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
583

584 585 586 587
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
588
{-
589
isSpecial :: Token -> Bool
590 591 592 593 594 595 596 597 598 599 600 601
-- 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
602
isSpecial ITinterruptible = True
603 604 605
isSpecial ITunsafe    	= True
isSpecial ITccallconv   = True
isSpecial ITstdcallconv = True
606
isSpecial ITprimcallconv = True
607
isSpecial ITmdo		= True
608
isSpecial ITfamily	= True
609 610 611
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
612
isSpecial _             = False
twanvl's avatar
twanvl committed
613
-}
614 615 616 617 618 619 620 621

-- 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
622
reservedWordsFM :: UniqFM (Token, Int)
623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651
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

652
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
653
	( "mdo",	ITmdo,		 bit recursiveDoBit),
654
	( "family",	ITfamily,	 bit tyFamBit),
655 656 657
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
658 659 660 661 662 663

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
664
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),  -- ToDo: remove
665
	( "interruptible", ITinterruptible, bit ffiBit),
666 667 668
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
669
	( "prim",       ITprimcallconv,	 bit ffiBit),
670

671
	( "rec",	ITrec,		 bit recBit),
672
	( "proc",	ITproc,		 bit arrowsBit)
673 674
     ]

Ian Lynagh's avatar
Ian Lynagh committed
675
reservedSymsFM :: UniqFM (Token, Int -> Bool)
676
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693
    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
694
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
695
        -- For 'forall a . t'
696
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
697 698 699 700 701

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
702

Ian Lynagh's avatar
Ian Lynagh committed
703 704
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
705 706
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
707 708
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
709 710 711 712 713 714 715 716

       ,("⤙",   ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
       ,("⤚",   ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
       ,("⤛",   ITLarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
       ,("⤜",   ITRarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)

       ,("★", ITstar, unicodeSyntaxEnabled)

Simon Marlow's avatar
Simon Marlow committed
717 718 719
        -- 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).
720 721 722 723 724
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
725
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
726

727
special :: Token -> Action
twanvl's avatar
twanvl committed
728
special tok span _buf _len = return (L span tok)
729

730
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
731 732
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
733

734 735
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
736

737 738 739
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
740

741 742 743
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
744

745
init_strtoken :: Int -> (String -> Token) -> Action
746
-- like strtoken, but drops the last N character(s)
747 748
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
749 750

begin :: Int -> Action
751
begin code _span _str _len = do pushLexState code; lexToken
752 753

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
754 755
pop _span _buf _len = do _ <- popLexState
                         lexToken
756

757 758 759 760 761 762 763 764 765 766 767
hopefully_open_brace :: Action
hopefully_open_brace span buf len
 = do relaxed <- extension relaxedLayout
      ctx <- getContext
      (AI l _) <- getInput
      let offset = srcLocCol l
          isOK = relaxed ||
                 case ctx of
                 Layout prev_off : _ -> prev_off < offset
                 _                   -> True
      if isOK then pop_and open_brace span buf len
Ian Lynagh's avatar
Ian Lynagh committed
768
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
769

770
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
771 772
pop_and act span buf len = do _ <- popLexState
                              act span buf len
773

774
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
775
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
776 777
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
778
notFollowedBy :: Char -> AlexAccPred Int
779
notFollowedBy char _ _ _ (AI _ buf) 
780
  = nextCharIs buf (/=char)
781

Ian Lynagh's avatar
Ian Lynagh committed
782
notFollowedBySymbol :: AlexAccPred Int
783
notFollowedBySymbol _ _ _ (AI _ buf)
784
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
785

786 787 788 789 790
-- 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
791
isNormalComment :: AlexAccPred Int
792
isNormalComment bits _ _ (AI _ buf)
793 794 795 796 797 798
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
799
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
800
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
801

twanvl's avatar
twanvl committed
802
{-
803
haddockDisabledAnd p bits _ _ (AI _ buf)
804
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
805
-}
806

Ian Lynagh's avatar
Ian Lynagh committed
807
atEOL :: AlexAccPred Int
808
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
809

Ian Lynagh's avatar
Ian Lynagh committed
810
ifExtension :: (Int -> Bool) -> AlexAccPred Int
811 812
ifExtension pred bits _ _ _ = pred bits

813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840
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
841 842 843 844 845
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

846 847 848 849
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
850
nested_comment :: P (RealLocated Token) -> Action
851
nested_comment cont span _str _len = do
852
  input <- getInput
Jedai's avatar
Jedai committed
853
  go "" (1::Int) input
854
  where
Jedai's avatar
Jedai committed
855 856 857 858 859 860
    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
861 862 863
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
864 865
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
866 867
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
868 869 870
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
871 872 873 874 875 876 877 878

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
879
        Just ('\125',input) ->
880
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
881
        Just (_,_) -> go ('-':commentAcc) input docType False
882 883 884 885 886 887
      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
888
        Just (_,_) -> go ('\123':commentAcc) input docType False
889 890
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
891 892
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
893
withLexedDocType lexDocComment = do
894
  input@(AI _ buf) <- getInput
895 896 897 898
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
899 900
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
901
    _ -> panic "withLexedDocType: Bad doc type"
902 903 904
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
905
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
906 907
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

908 909 910
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
911
rulePrag span _buf _len = do
912
  setExts (.|. bit inRulePragBit)
913 914 915
  return (L span ITrules_prag)

endPrag :: Action
916
endPrag span _buf _len = do
917 918 919
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

920 921 922 923 924 925 926 927 928
-- 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. 

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
929
                 RealSrcSpan -> P (RealLocated Token) 
930 931
docCommentEnd input commentAcc docType buf span = do
  setInput input
932
  let (AI loc nextBuf) = input
933
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
934
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
935 936
      last_len = byteDiff buf nextBuf
      
937
  span `seq` setLastToken span' last_len
938 939
  return (L span' (docType comment))
 
Ian Lynagh's avatar
Ian Lynagh committed
940 941
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
942

943
open_brace, close_brace :: Action
944
open_brace span _str _len = do 
945 946
  ctx <- getContext
  setContext (NoLayout:ctx)
947 948
  return (L span ITocurly)
close_brace span _str _len = do 
949
  popContext
950
  return (L span ITccurly)
951

Ian Lynagh's avatar
Ian Lynagh committed
952
qvarid, qconid :: StringBuffer -> Int -> Token
953 954
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
955

956
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
957 958 959
-- 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.
960
splitQualName orig_buf len parens = split orig_buf orig_buf
961
  where
962 963 964 965 966 967
    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
968 969 970 971
  
    -- 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.
972 973 974 975 976 977 978 979
    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),
980 981 982
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
983 984
      where
	qual_size = orig_buf `byteDiff` dot_buf
985

Ian Lynagh's avatar
Ian Lynagh committed
986 987
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
988
  fs `seq`
989 990 991
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
992
		return (L span keyword)
993 994 995
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
996 997 998
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
999 1000 1001
  where
	fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
1002
conid :: StringBuffer -> Int -> Token
1003 1004 1005
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
1006
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
1007 1008 1009 1010
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
1011

Ian Lynagh's avatar
Ian Lynagh committed
1012
varsym, consym :: Action
1013 1014 1015
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
1016 1017
sym :: (FastString -> Token) -> RealSrcSpan -> StringBuffer -> Int
    -> P (RealLocated Token)
1018
sym con span buf len = 
1019 1020
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
1021
		b <- extension exts
1022 1023 1024
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
1025 1026 1027
  where
	fs = lexemeToFastString buf len

1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038
-- 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
1039 1040 1041
tok_num :: (Integer -> Integer)
        -> Int -> Int
        -> (Integer, (Char->Int)) -> Action
1042
tok_num = tok_integral ITinteger
Ian Lynagh's avatar
Ian Lynagh committed
1043 1044 1045
tok_primint :: (Integer -> Integer)
            -> Int -> Int
            -> (Integer, (Char->Int)) -> Action
1046
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1047 1048
tok_primword :: Int -> Int
             -> (Integer, (Char->Int)) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
1049
tok_primword = tok_integral ITprimword positive
Ian Lynagh's avatar
Ian Lynagh committed
1050
positive, negative :: (Integer -> Integer)
1051 1052
positive = id
negative = negate
Ian Lynagh's avatar
Ian Lynagh committed
1053
decimal, octal, hexadecimal :: (Integer, Char -> Int)
1054 1055 1056 1057 1058
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
Ian Lynagh's avatar
Ian Lynagh committed
1059
tok_float, tok_primfloat, tok_primdouble :: String -> Token
1060 1061 1062 1063 1064 1065
tok_float        str = ITrational   $! readFractionalLit str
tok_primfloat    str = ITprimfloat  $! readFractionalLit str
tok_primdouble   str = ITprimdouble $! readFractionalLit str

readFractionalLit :: String -> FractionalLit
readFractionalLit str = (FL $! str) $! readRational str
1066 1067 1068 1069 1070 1071

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1072
do_bol span _str _len = do
1073
	pos <- getOffside
1074 1075 1076 1077 1078
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1079
		return (L span ITvccurly)
1080 1081
	    EQ -> do
                --trace "layout: inserting ';'" $ do
Ian Lynagh's avatar
Ian Lynagh committed
1082
		_ <- popLexState
1083
		return (L span ITsemi)
1084
	    GT -> do
Ian Lynagh's avatar
Ian Lynagh committed
1085
		_ <- popLexState
1086 1087 1088 1089
		lexToken

-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
Ian Lynagh's avatar
Ian Lynagh committed
1090
maybe_layout :: Token -> P ()
1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105 1106
maybe_layout t = do -- If the alternative layout rule is enabled then
                    -- we never create an implicit layout context here.
                    -- Layout is handled XXX instead.
                    -- The code for closing implicit contexts, or
                    -- inserting implicit semi-colons, is therefore
                    -- irrelevant as it only applies in an implicit
                    -- context.
                    alr <- extension alternativeLayoutRule
                    unless alr $ f t
    where f ITdo    = pushLexState layout_do
          f ITmdo   = pushLexState layout_do
          f ITof    = pushLexState layout
          f ITlet   = pushLexState layout
          f ITwhere = pushLexState layout
          f ITrec   = pushLexState layout
          f _       = return ()
1107 1108 1109 1110 1111 1112 1113 1114 1115 1116

-- 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
1117
new_layout_context :: Bool -> Action
1118
new_layout_context strict span _buf _len = do
Ian Lynagh's avatar
Ian Lynagh committed
1119
    _ <- popLexState
1120 1121
    (AI l _) <- getInput
    let offset = srcLocCol l
1122
    ctx <- getContext