Lexer.x 79.7 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
-- XXX The above flags turn off warnings in the generated code:
35
{-# LANGUAGE BangPatterns #-}
36 37 38 39 40 41 42 43
{-# 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 #-}
44

45 46
{-# OPTIONS_GHC -funbox-strict-fields #-}

47
module Lexer (
48
   Token(..), lexer, pragState, mkPState, PState(..),
49
   P(..), ParseResult(..), getSrcLoc, 
50
   getPState, getDynFlags, withThisPackage,
51
   failLocMsgP, failSpanMsgP, srcParseFail,
52
   getMessages, 
53
   popContext, pushCurrentContext, setLastToken, setSrcLoc,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
54
   getLexState, popLexState, pushLexState,
55
   extension, bangPatEnabled, datatypeContextsEnabled,
Jedai's avatar
Jedai committed
56 57
   addWarning,
   lexTokenStream
58 59
  ) where

60 61
import Bag
import ErrUtils
62 63 64 65 66
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
67
import DynFlags
68
import Module
69
import Ctype
70
import BasicTypes	( InlineSpec(..), RuleMatchInfo(..) )
71
import Util		( readRational )
72

73
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
74
import Data.Bits
75
import Data.Char
76
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
77
import Data.Maybe
78 79
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
80
import Data.Ratio
81 82
}

83
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
84
$whitechar   = [\ \n\r\f\v $unispace]
85
$white_no_nl = $whitechar # \n
86
$tab         = \t
87 88

$ascdigit  = 0-9
89
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
90
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
91 92 93
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
94
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
95
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
96 97
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

98
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
99
$asclarge  = [A-Z]
100 101
$large     = [$asclarge $unilarge]

102
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
103
$ascsmall  = [a-z]
104 105
$small     = [$ascsmall $unismall \_]

106
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
107
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
108 109

$octit	   = 0-7
110
$hexit     = [$decdigit A-F a-f]
111 112 113 114
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

115 116
$pragmachar = [$small $large $digit]

117 118
$docsym    = [\| \^ \* \$]

119 120 121 122 123 124
@varid     = $small $idchar*
@conid     = $large $idchar*

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

125
@decimal     = $decdigit+
126 127 128 129 130 131 132 133 134
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

135 136 137 138 139
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

140 141 142 143
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
144
$tab+         { warn Opt_WarnTabs (text "Warning: Tab character") }
145 146 147 148 149

-- 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
150 151 152 153
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
154 155 156 157

-- 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.
158 159 160 161 162 163 164 165 166 167

-- 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
168 169
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
170 171 172

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

Jedai's avatar
Jedai committed
173
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
174 175 176 177 178 179 180

-- 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
181
"---"\-* [^$symbol :] .* { lineCommentToken }
182 183 184 185

-- 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
186
"--"\-* / { atEOL } { lineCommentToken }
187 188 189 190

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

Jedai's avatar
Jedai committed
191
"-- " / { atEOL } { lineCommentToken }
192 193 194 195 196 197 198 199 200 201 202 203 204

-- '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 }
205
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
206
  ^\# \! .* \n				; -- #!, for scripts
207 208 209 210 211 212 213
  ()					{ 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> {
214
  \{ / { notFollowedBy '-' }		{ hopefully_open_brace }
215 216 217 218 219 220 221 222 223 224 225 226 227 228
	-- 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 }

229
<0,option_prags> \n				{ begin bol }
230

231 232
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
233 234 235

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
236
<line_prag1> $decdigit+			{ setLine line_prag1a }
237
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
238 239 240 241
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
242
<line_prag2> $decdigit+			{ setLine line_prag2a }
243
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
244 245 246
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
247

248
<0,option_prags> {
249 250 251 252 253 254
  "{-#" $whitechar* $pragmachar+ 
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
255

256
  -- We ignore all these pragmas, but don't generate a warning for them
257 258
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
259 260

  -- ToDo: should only be valid inside a pragma:
261
  "#-}" 				{ endPrag }
262 263
}

264
<option_prags> {
265 266 267
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

David Waern's avatar
David Waern committed
268 269 270
  "-- #"                                 { multiline_doc_comment }
}

271 272
<0> {
  -- In the "0" mode we ignore these pragmas
273
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
274 275 276
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
277
<0> {
Jedai's avatar
Jedai committed
278
  "-- #" .* { lineCommentToken }
279
}
280

281
<0,option_prags> {
282 283
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
284 285
}

286 287
-- '0' state: ordinary lexemes

288 289
-- Haddock comments

290
<0,option_prags> {
David Waern's avatar
David Waern committed
291 292
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
293 294
}

295 296
-- "special" symbols

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

312 313 314 315
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

316
  "[" @varid "|"  / { ifExtension qqEnabled }
317
                     { lex_quasiquote_tok }
318 319
}

320
<0> {
ross's avatar
ross committed
321 322
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
323 324 325
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

326
<0> {
327 328 329
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

330
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
331 332 333 334 335 336
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

337
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
338 339
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
340 341
}

342
<0,option_prags> {
343 344 345 346 347 348 349 350 351 352 353 354
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

355
<0,option_prags> {
356
  @qual @varid			{ idtoken qvarid }
357 358 359 360 361
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

362
<0> {
363 364 365 366
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
367 368
}

369 370
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
371
<0> {
372 373 374 375 376 377
  @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 }
378 379
}

380 381
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
382
<0> {
383
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
384 385 386
  @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 }
387 388

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
389
  @floating_point    { strtoken tok_float }
390 391
}

392
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
393
  -- Unboxed ints (:: Int#) and words (:: Word#)
394 395
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
396 397 398 399 400
  @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 }
401
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
402

Ian Lynagh's avatar
Ian Lynagh committed
403 404 405 406
  @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 }

407 408
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
409 410
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
411 412 413 414 415 416
}

-- 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.
417
<0> {
418 419 420 421 422 423 424 425
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

426
data Token
427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459
  = 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
460
  | ITinterruptible
461 462 463
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
464
  | ITprimcallconv
465
  | ITmdo
466
  | ITfamily
467 468 469
  | ITgroup
  | ITby
  | ITusing
470

471
	-- Pragmas
472
  | ITinline_prag InlineSpec RuleMatchInfo
473 474
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
475 476
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
477
  | ITwarning_prag
478 479 480
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
481
  | ITgenerated_prag
482
  | ITcore_prag                 -- hdaume: core annotations
483
  | ITunpack_prag
484
  | ITann_prag
485
  | ITclose_prag
486 487 488
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514

  | 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
515 516
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534
  | 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)
535 536
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
537 538 539 540 541 542 543 544 545 546 547

  | 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
548
  | ITprimword   Integer
549 550 551
  | ITprimfloat  Rational
  | ITprimdouble Rational

552
  -- Template Haskell extension tokens
553 554 555 556 557 558 559 560 561
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
562
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
563 564 565 566

  -- Arrow notation extension
  | ITproc
  | ITrec
567 568 569 570 571 572
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
573 574 575

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
576 577 578 579 580 581 582

  -- 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
583
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
584 585
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
586

587 588 589 590
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
591
{-
592
isSpecial :: Token -> Bool
593 594 595 596 597 598 599 600 601 602 603 604
-- 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
605
isSpecial ITinterruptible = True
606 607 608
isSpecial ITunsafe    	= True
isSpecial ITccallconv   = True
isSpecial ITstdcallconv = True
609
isSpecial ITprimcallconv = True
610
isSpecial ITmdo		= True
611
isSpecial ITfamily	= True
612 613 614
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
615
isSpecial _             = False
twanvl's avatar
twanvl committed
616
-}
617 618 619 620 621 622 623 624

-- 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
625
reservedWordsFM :: UniqFM (Token, Int)
626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654
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

655
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
656
	( "mdo",	ITmdo,		 bit recursiveDoBit),
657
	( "family",	ITfamily,	 bit tyFamBit),
658 659 660
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
661 662 663 664 665 666

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
667
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),  -- ToDo: remove
668
	( "interruptible", ITinterruptible, bit ffiBit),
669 670 671
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
672
	( "prim",       ITprimcallconv,	 bit ffiBit),
673

674
	( "rec",	ITrec,		 bit recBit),
675
	( "proc",	ITproc,		 bit arrowsBit)
676 677
     ]

Ian Lynagh's avatar
Ian Lynagh committed
678
reservedSymsFM :: UniqFM (Token, Int -> Bool)
679
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696
    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
697
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
698
        -- For 'forall a . t'
699
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
700 701 702 703 704

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
705

Ian Lynagh's avatar
Ian Lynagh committed
706 707
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
708 709
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
710 711
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
712 713 714 715 716 717 718 719

       ,("⤙",   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
720 721 722
        -- 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).
723 724 725 726 727
       ]

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

728
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
729

730
special :: Token -> Action
twanvl's avatar
twanvl committed
731
special tok span _buf _len = return (L span tok)
732

733
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
734 735
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
736

737 738
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
739

740 741 742
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
743

744 745 746
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
747

748
init_strtoken :: Int -> (String -> Token) -> Action
749
-- like strtoken, but drops the last N character(s)
750 751
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
752 753

begin :: Int -> Action
754
begin code _span _str _len = do pushLexState code; lexToken
755 756

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
757 758
pop _span _buf _len = do _ <- popLexState
                         lexToken
759

760 761 762 763 764 765 766 767 768 769 770 771 772
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
              else failSpanMsgP span (text "Missing block")

773
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
774 775
pop_and act span buf len = do _ <- popLexState
                              act span buf len
776

777
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
778
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
779 780
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
781
notFollowedBy :: Char -> AlexAccPred Int
782
notFollowedBy char _ _ _ (AI _ buf) 
783
  = nextCharIs buf (/=char)
784

Ian Lynagh's avatar
Ian Lynagh committed
785
notFollowedBySymbol :: AlexAccPred Int
786
notFollowedBySymbol _ _ _ (AI _ buf)
787
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
788

789 790 791 792 793
-- 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
794
isNormalComment :: AlexAccPred Int
795
isNormalComment bits _ _ (AI _ buf)
796 797 798 799 800 801
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
802
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
803
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
804

twanvl's avatar
twanvl committed
805
{-
806
haddockDisabledAnd p bits _ _ (AI _ buf)
807
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
808
-}
809

Ian Lynagh's avatar
Ian Lynagh committed
810
atEOL :: AlexAccPred Int
811
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
812

Ian Lynagh's avatar
Ian Lynagh committed
813
ifExtension :: (Int -> Bool) -> AlexAccPred Int
814 815
ifExtension pred bits _ _ _ = pred bits

816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843
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
844 845 846 847 848
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

849 850 851 852
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
853 854
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
855
  input <- getInput
Jedai's avatar
Jedai committed
856
  go "" (1::Int) input
857
  where
Jedai's avatar
Jedai committed
858 859 860 861 862 863
    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
864 865 866
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
867 868
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
869 870
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
871 872 873
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
874 875 876 877 878 879 880 881

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
882
        Just ('\125',input) ->
883
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
884
        Just (_,_) -> go ('-':commentAcc) input docType False
885 886 887 888 889 890
      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
891
        Just (_,_) -> go ('\123':commentAcc) input docType False
892 893
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
894 895
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
                 -> P (Located Token)
896
withLexedDocType lexDocComment = do
897
  input@(AI _ buf) <- getInput
898 899 900 901
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
902 903
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
904
    _ -> panic "withLexedDocType: Bad doc type"
905 906 907
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
908
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
909 910
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

911 912 913
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
914
rulePrag span _buf _len = do
915
  setExts (.|. bit inRulePragBit)
916 917 918
  return (L span ITrules_prag)

endPrag :: Action
919
endPrag span _buf _len = do
920 921 922
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

923 924 925 926 927 928 929 930 931 932 933 934
-- 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 ->
                 SrcSpan -> P (Located Token) 
docCommentEnd input commentAcc docType buf span = do
  setInput input
935
  let (AI loc nextBuf) = input
936 937 938 939
      comment = reverse commentAcc
      span' = mkSrcSpan (srcSpanStart span) loc
      last_len = byteDiff buf nextBuf
      
940
  span `seq` setLastToken span' last_len
941 942
  return (L span' (docType comment))
 
Ian Lynagh's avatar
Ian Lynagh committed
943
errBrace :: AlexInput -> SrcSpan -> P a
944
errBrace (AI end _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
945

946
open_brace, close_brace :: Action
947
open_brace span _str _len = do 
948 949
  ctx <- getContext
  setContext (NoLayout:ctx)
950 951
  return (L span ITocurly)
close_brace span _str _len = do 
952
  popContext
953
  return (L span ITccurly)
954

Ian Lynagh's avatar
Ian Lynagh committed
955
qvarid, qconid :: StringBuffer -> Int -> Token
956 957
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
958

959
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
960 961 962
-- 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.
963
splitQualName orig_buf len parens = split orig_buf orig_buf
964
  where
965 966 967 968 969 970
    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
971 972 973 974
  
    -- 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.
975 976 977 978 979 980 981 982
    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),
983 984 985
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
986 987
      where
	qual_size = orig_buf `byteDiff` dot_buf
988

Ian Lynagh's avatar
Ian Lynagh committed
989 990
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
991
  fs `seq`
992 993 994
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
995
		return (L span keyword)
996 997 998
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
999 1000 1001
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
1002 1003 1004
  where
	fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
1005
conid :: StringBuffer -> Int -> Token
1006 1007 1008
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
1009
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
1010 1011 1012 1013
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
1014

Ian Lynagh's avatar
Ian Lynagh committed
1015
varsym, consym :: Action
1016 1017 1018
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
1019 1020
sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
    -> P (Located Token)
1021
sym con span buf len = 
1022 1023
  case lookupUFM reservedSymsFM fs of
	Just (keyword,exts) -> do
Ian Lynagh's avatar
Ian Lynagh committed
1024
		b <- extension exts
1025 1026 1027
		if b then return (L span keyword)
		     else return (L span $! con fs)
	_other -> return (L span $! con fs)
1028 1029 1030
  where
	fs = lexemeToFastString buf len

1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041
-- 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
1042 1043 1044
tok_num :: (Integer -> Integer)
        -> Int -> Int
        -> (Integer, (Char->Int)) -> Action
1045
tok_num = tok_integral ITinteger
Ian Lynagh's avatar
Ian Lynagh committed
1046 1047 1048
tok_primint :: (Integer -> Integer)
            -> Int -> Int
            -> (Integer, (Char->Int)) -> Action
1049
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1050 1051
tok_primword :: Int -> Int
             -> (Integer, (Char->Int)) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
1052
tok_primword = tok_integral ITprimword positive
Ian Lynagh's avatar
Ian Lynagh committed
1053
positive, negative :: (Integer -> Integer)
1054 1055
positive = id
negative = negate
Ian Lynagh's avatar
Ian Lynagh committed
1056
decimal, octal, hexadecimal :: (Integer, Char -> Int)
1057 1058 1059 1060 1061
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
Ian Lynagh's avatar
Ian Lynagh committed
1062
tok_float, tok_primfloat, tok_primdouble :: String -> Token
1063
tok_float        str = ITrational   $! readRational str
1064 1065
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! 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
1123 1124
    nondecreasing <- extension nondecreasingIndentation
    let strict' = strict || not nondecreasing
1125 1126
    case ctx of
	Layout prev_off : _  | 
1127 1128
	   (strict'     && prev_off >= offset  ||
	    not strict' && prev_off > offset) -> do
1129 1130 1131
		-- token is indented to the left of the previous context.
		-- we must generate a {} sequence now.
		pushLexState layout_left
1132
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1133
	_ -> do
1134
		setContext (Layout offset : ctx)
1135
		return (L span ITvocurly)