Lexer.x 60.7 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 22
--
-- GHC's lexer.
--
-- This is a combination of an Alex-generated lexer from a regex
-- definition, with some hand-coded bits.
--
-- Completely accurate information about token-spans within the source
-- file is maintained.  Every token has a start and end SrcLoc attached to it.
--
-----------------------------------------------------------------------------

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

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
{-# OPTIONS -w #-}
36 37 38
-- The above warning supression flag is a temporary kludge.
-- While working on this module you are encouraged to remove it and fix
-- any warnings in the module. See
Ian Lynagh's avatar
Ian Lynagh committed
39
--     http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
40
-- for details
twanvl's avatar
twanvl committed
41 42 43
--
-- Note that Alex itself generates code with with some unused bindings and
-- without type signatures, so removing the flag might not be possible.
44

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

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

59 60
import Bag
import ErrUtils
61 62 63 64 65 66
import Outputable
import StringBuffer
import FastString
import FastTypes
import SrcLoc
import UniqFM
67
import DynFlags
68
import Ctype
69
import Util		( maybePrefixMatch, readRational )
70

71
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
72
import Data.Bits
73
import Data.Char 	( chr, ord, isSpace )
Simon Marlow's avatar
Simon Marlow committed
74 75
import Data.Ratio
import Debug.Trace
76

Ian Lynagh's avatar
Ian Lynagh committed
77
import Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper )
78 79
}

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

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

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

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

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

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

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

112 113
$docsym    = [\| \^ \* \$]

114 115 116 117 118 119
@varid     = $small $idchar*
@conid     = $large $idchar*

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

120
@decimal     = $decdigit+
121 122 123 124 125 126 127 128 129
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

130 131 132 133 134
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

135 136 137 138
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
139
$tab+         { warn Opt_WarnTabs (text "Tab character") }
140 141 142 143 144

-- 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
145 146 147 148
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
149 150 151 152

-- 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.
153 154 155 156 157 158 159 160 161 162

-- 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
163 164
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
165 166 167

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

Jedai's avatar
Jedai committed
168
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
169 170 171 172 173 174 175

-- 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
176
"---"\-* [^$symbol :] .* { lineCommentToken }
177 178 179 180

-- 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
181
"--"\-* / { atEOL } { lineCommentToken }
182 183 184 185

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

Jedai's avatar
Jedai committed
186
"-- " / { atEOL } { lineCommentToken }
187 188 189 190 191 192 193 194 195 196 197 198 199

-- '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 }
200
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
201
  ^\# \! .* \n				; -- #!, for scripts
202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223
  ()					{ do_bol }
}

-- after a layout keyword (let, where, do, of), we begin a new layout
-- context if the curly brace is missing.
-- Careful! This stuff is quite delicate.
<layout, layout_do> {
  \{ / { notFollowedBy '-' }		{ pop_and open_brace }
	-- we might encounter {-# here, but {- has been handled already
  \n					;
  ^\# (line)?				{ begin line_prag1 }
}

-- do is treated in a subtly different way, see new_layout_context
<layout>    ()				{ new_layout_context True }
<layout_do> ()				{ new_layout_context False }

-- after a new layout context which was found to be to the left of the
-- previous context, we have generated a '{' token, and we now need to
-- generate a matching '}' token.
<layout_left>  ()			{ do_layout_left }

224
<0,option_prags> \n				{ begin bol }
225 226 227 228 229

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

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
230
<line_prag1> $decdigit+			{ setLine line_prag1a }
231
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
232 233 234 235
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
236
<line_prag2> $decdigit+			{ setLine line_prag2a }
237
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
238 239 240
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
241

242
<0,option_prags> {
243
  "{-#" $whitechar* (RULES|rules)       { rulePrag }
244 245 246 247 248
  "{-#" $whitechar* (INLINE|inline)	{ token (ITinline_prag True) }
  "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
  					{ token (ITinline_prag False) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
  					{ token ITspec_prag }
249
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
250 251 252 253
	$whitechar* (INLINE|inline)	{ token (ITspec_inline_prag True) }
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
	$whitechar* (NO(T?)INLINE|no(t?)inline)
					{ token (ITspec_inline_prag False) }
254
  "{-#" $whitechar* (SOURCE|source)	{ token ITsource_prag }
Ian Lynagh's avatar
Ian Lynagh committed
255 256
  "{-#" $whitechar* (WARNING|warning)
  					{ token ITwarning_prag }
257 258 259
  "{-#" $whitechar* (DEPRECATED|deprecated)
  					{ token ITdeprecated_prag }
  "{-#" $whitechar* (SCC|scc)		{ token ITscc_prag }
andy@galois.com's avatar
andy@galois.com committed
260 261
  "{-#" $whitechar* (GENERATED|generated)
  					{ token ITgenerated_prag }
262
  "{-#" $whitechar* (CORE|core)		{ token ITcore_prag }
263
  "{-#" $whitechar* (UNPACK|unpack)	{ token ITunpack_prag }
264

265 266 267 268
  -- We ignore all these pragmas, but don't generate a warning for them
  -- CFILES is a hugs-only thing.
  "{-#" $whitechar* (OPTIONS_HUGS|options_hugs|OPTIONS_NHC98|options_nhc98|OPTIONS_JHC|options_jhc|CFILES|cfiles)
                    { nested_comment lexToken }
269 270

  -- ToDo: should only be valid inside a pragma:
271
  "#-}" 				{ endPrag }
272 273
}

274
<option_prags> {
David Waern's avatar
David Waern committed
275 276
  "{-#"  $whitechar* (OPTIONS|options)   { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc)
277
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
278 279 280 281 282 283 284
  "{-#"  $whitechar* (OPTIONS_HADDOCK|options_haddock)
                                         { lex_string_prag ITdocOptions }
  "-- #"                                 { multiline_doc_comment }
  "{-#"  $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
  "{-#"  $whitechar* (INCLUDE|include)   { lex_string_prag ITinclude_prag }
}

285 286 287 288 289 290
<0> {
  -- In the "0" mode we ignore these pragmas
  "{-#"  $whitechar* (OPTIONS|options|OPTIONS_GHC|options_ghc|OPTIONS_HADDOCK|options_haddock|LANGUAGE|language|INCLUDE|include)
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
291
<0> {
Jedai's avatar
Jedai committed
292
  "-- #" .* { lineCommentToken }
293
}
294

295
<0,option_prags> {
296 297
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
298 299
}

300 301
-- '0' state: ordinary lexemes

302 303
-- Haddock comments

304
<0> {
David Waern's avatar
David Waern committed
305 306
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
307 308
}

309 310
-- "special" symbols

311
<0> {
312 313 314 315
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
316
<0> {
317 318 319 320 321 322 323 324
  "[|"	    / { 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 }
325 326 327

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
328 329
}

330
<0> {
ross's avatar
ross committed
331 332
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
333 334 335
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

336
<0> {
337 338 339
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

340
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
341 342 343 344 345 346
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

347
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
348 349
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
350 351
}

352
<0,option_prags> {
353 354 355 356 357 358 359 360 361 362 363 364
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

365
<0,option_prags> {
366
  @qual @varid			{ idtoken qvarid }
367 368 369 370 371
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

372
<0> {
373 374 375 376
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
377 378
}

379 380
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
381
<0> {
382 383 384 385 386 387
  @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 }
388 389
}

390 391
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
392
<0> {
393
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
394 395 396
  @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 }
397 398

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
399
  @floating_point    { strtoken tok_float }
400 401
}

402
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
403
  -- Unboxed ints (:: Int#) and words (:: Word#)
404 405
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
406 407 408 409 410
  @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 }
411
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
412

Ian Lynagh's avatar
Ian Lynagh committed
413 414 415 416
  @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 }

417 418
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
419 420
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
421 422 423 424 425 426
}

-- 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.
427
<0> {
428 429 430 431 432 433 434 435
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

436
data Token
437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
  = ITas  			-- Haskell keywords
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
  | ITdo
  | ITelse
  | IThiding
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere
  | ITscc			-- ToDo: remove (we use {-# SCC "..." #-} now)

  | ITforall			-- GHC extension keywords
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
  | ITthreadsafe
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
  | ITdotnet
  | ITmdo
475
  | ITfamily
476 477 478
  | ITgroup
  | ITby
  | ITusing
479

480 481 482 483
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
484 485
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
486
  | ITwarning_prag
487 488 489
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
490
  | ITgenerated_prag
491
  | ITcore_prag                 -- hdaume: core annotations
492
  | ITunpack_prag
493
  | ITclose_prag
494 495 496
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522

  | 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
523 524
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542
  | 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)
543 544
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
545 546 547 548 549 550 551 552 553 554 555 556 557

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

  | ITpragma StringBuffer

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

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
558
  | ITprimword   Integer
559 560 561 562
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
563 564 565 566 567 568 569 570 571
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
572
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
573 574 575 576

  -- Arrow notation extension
  | ITproc
  | ITrec
577 578 579 580 581 582
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
583 584 585

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
586 587 588 589 590 591 592

  -- 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
593
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
594 595
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
596

597 598 599 600
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
601
{-
602
isSpecial :: Token -> Bool
603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618
-- If we see M.x, where x is a keyword, but
-- is special, we treat is as just plain M.x, 
-- not as a keyword.
isSpecial ITas        	= True
isSpecial IThiding    	= True
isSpecial ITqualified 	= True
isSpecial ITforall    	= True
isSpecial ITexport    	= True
isSpecial ITlabel     	= True
isSpecial ITdynamic   	= True
isSpecial ITsafe    	= True
isSpecial ITthreadsafe 	= True
isSpecial ITunsafe    	= True
isSpecial ITccallconv   = True
isSpecial ITstdcallconv = True
isSpecial ITmdo		= True
619
isSpecial ITfamily	= True
620 621 622
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
623
isSpecial _             = False
twanvl's avatar
twanvl committed
624
-}
625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661

-- the bitmap provided as the third component indicates whether the
-- corresponding extension keyword is valid under the extension options
-- provided to the compiler; if the extension corresponding to *any* of the
-- bits set in the bitmap is enabled, the keyword is valid (this setup
-- facilitates using a keyword in two different extensions that can be
-- activated independently)
--
reservedWordsFM = listToUFM $
	map (\(x, y, z) -> (mkFastString x, (y, z)))
       [( "_",		ITunderscore, 	0 ),
	( "as",		ITas, 		0 ),
	( "case",	ITcase, 	0 ),     
	( "class",	ITclass, 	0 ),    
	( "data",	ITdata, 	0 ),     
	( "default",	ITdefault, 	0 ),  
	( "deriving",	ITderiving, 	0 ), 
	( "do",		ITdo, 		0 ),       
	( "else",	ITelse, 	0 ),     
	( "hiding",	IThiding, 	0 ),
	( "if",		ITif, 		0 ),       
	( "import",	ITimport, 	0 ),   
	( "in",		ITin, 		0 ),       
	( "infix",	ITinfix, 	0 ),    
	( "infixl",	ITinfixl, 	0 ),   
	( "infixr",	ITinfixr, 	0 ),   
	( "instance",	ITinstance, 	0 ), 
	( "let",	ITlet, 		0 ),      
	( "module",	ITmodule, 	0 ),   
	( "newtype",	ITnewtype, 	0 ),  
	( "of",		ITof, 		0 ),       
	( "qualified",	ITqualified, 	0 ),
	( "then",	ITthen, 	0 ),     
	( "type",	ITtype, 	0 ),     
	( "where",	ITwhere, 	0 ),
	( "_scc_",	ITscc, 		0 ),		-- ToDo: remove

662
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
663
	( "mdo",	ITmdo,		 bit recursiveDoBit),
664
	( "family",	ITfamily,	 bit tyFamBit),
665 666 667
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
668 669 670 671 672 673 674 675 676 677 678 679 680

	( "foreign",	ITforeign,	 bit ffiBit),
	( "export",	ITexport,	 bit ffiBit),
	( "label",	ITlabel,	 bit ffiBit),
	( "dynamic",	ITdynamic,	 bit ffiBit),
	( "safe",	ITsafe,		 bit ffiBit),
	( "threadsafe",	ITthreadsafe,	 bit ffiBit),
	( "unsafe",	ITunsafe,	 bit ffiBit),
	( "stdcall",    ITstdcallconv,	 bit ffiBit),
	( "ccall",      ITccallconv,	 bit ffiBit),
	( "dotnet",     ITdotnet,	 bit ffiBit),

	( "rec",	ITrec,		 bit arrowsBit),
681
	( "proc",	ITproc,		 bit arrowsBit)
682 683
     ]

Ian Lynagh's avatar
Ian Lynagh committed
684
reservedSymsFM :: UniqFM (Token, Int -> Bool)
685
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702
    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
703
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
704
        -- For 'forall a . t'
705
       ,(".", ITdot, \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
706 707 708 709 710

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
711 712

#if __GLASGOW_HASKELL__ >= 605
Ian Lynagh's avatar
Ian Lynagh committed
713 714
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
715 716
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
717 718 719
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, 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
#endif
724 725 726 727 728
       ]

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

729
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
730

731
special :: Token -> Action
twanvl's avatar
twanvl committed
732
special tok span _buf _len = return (L span tok)
733

734
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
735 736
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
737

738 739
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
740

741 742 743
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
744

745 746 747
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
748

749
init_strtoken :: Int -> (String -> Token) -> Action
750
-- like strtoken, but drops the last N character(s)
751 752
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
753 754

begin :: Int -> Action
755
begin code _span _str _len = do pushLexState code; lexToken
756 757

pop :: Action
758
pop _span _buf _len = do popLexState; lexToken
759 760

pop_and :: Action -> Action
761
pop_and act span buf len = do popLexState; act span buf len
762

763 764 765 766 767
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
768

769
notFollowedBySymbol _ _ _ (AI _ _ buf)
770
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
771

772 773 774 775 776
-- 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.
777
isNormalComment bits _ _ (AI _ _ buf)
778 779 780 781 782 783 784
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
785

twanvl's avatar
twanvl committed
786
{-
787 788
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
789
-}
790

791
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
792

793 794
ifExtension pred bits _ _ _ = pred bits

795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822
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
823 824 825 826 827
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

828 829 830 831
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
832 833
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
834
  input <- getInput
Jedai's avatar
Jedai committed
835
  go "" (1::Int) input
836
  where
Jedai's avatar
Jedai committed
837 838 839 840 841 842
    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
843 844 845
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
846 847
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
848 849
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
850 851 852
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
853 854 855 856 857 858 859 860

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
861
        Just ('\125',input) ->
862
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
863
        Just (_,_) -> go ('-':commentAcc) input docType False
864 865 866 867 868 869
      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
870
        Just (_,_) -> go ('\123':commentAcc) input docType False
871 872 873
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
874 875 876 877 878
  input@(AI _ _ buf) <- getInput
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
879 880
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
881 882 883
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
884
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
885 886
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

887 888 889 890
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
rulePrag span buf len = do
891
  setExts (.|. bit inRulePragBit)
892 893 894 895 896 897 898
  return (L span ITrules_prag)

endPrag :: Action
endPrag span buf len = do
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932
-- docCommentEnd
-------------------------------------------------------------------------------
-- This function is quite tricky. We can't just return a new token, we also
-- need to update the state of the parser. Why? Because the token is longer
-- than what was lexed by Alex, and the lexToken function doesn't know this, so 
-- it writes the wrong token length to the parser state. This function is
-- called afterwards, so it can just update the state. 

-- This is complicated by the fact that Haddock tokens can span multiple lines, 
-- which is something that the original lexer didn't account for. 
-- I have added last_line_len in the parser state which represents the length 
-- of the part of the token that is on the last line. It is now used for layout 
-- calculation in pushCurrentContext instead of last_len. last_len is, like it 
-- was before, the full length of the token, and it is now only used for error
-- messages. /Waern 

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
                 SrcSpan -> P (Located Token) 
docCommentEnd input commentAcc docType buf span = do
  setInput input
  let (AI loc last_offs nextBuf) = input
      comment = reverse commentAcc
      span' = mkSrcSpan (srcSpanStart span) loc
      last_len = byteDiff buf nextBuf
      
      last_line_len = if (last_offs - last_len < 0) 
        then last_offs
        else last_len  
  
  span `seq` setLastToken span' last_len last_line_len
  return (L span' (docType comment))
 
errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
 
933
open_brace, close_brace :: Action
934
open_brace span _str _len = do 
935 936
  ctx <- getContext
  setContext (NoLayout:ctx)
937 938
  return (L span ITocurly)
close_brace span _str _len = do 
939
  popContext
940
  return (L span ITccurly)
941

942 943
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
944

945
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
946 947 948
-- 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.
949
splitQualName orig_buf len parens = split orig_buf orig_buf
950
  where
951 952 953 954 955 956
    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
957 958 959 960
  
    -- 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.
961 962 963 964 965 966 967 968
    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),
969 970 971
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
972 973
      where
	qual_size = orig_buf `byteDiff` dot_buf
974

975
varid span buf len = 
Ian Lynagh's avatar
Ian Lynagh committed
976
  fs `seq`
977 978 979
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
980
		return (L span keyword)
981 982 983
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
984 985 986
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
987 988 989 990 991 992
  where
	fs = lexemeToFastString buf len

conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

993 994 995 996
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
997 998 999 1000

varsym = sym ITvarsym
consym = sym ITconsym

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

1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023
-- 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
tok_num = tok_integral ITinteger
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1024
tok_primword = tok_integral ITprimword positive
1025 1026 1027 1028 1029 1030 1031
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
1032
tok_float        str = ITrational   $! readRational str
1033 1034
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
1035 1036 1037 1038 1039 1040

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1041
do_bol span _str _len = do
1042
	pos <- getOffside
1043 1044 1045 1046 1047
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1048
		return (L span ITvccurly)
1049 1050 1051
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1052
		return (L span ITsemi)
1053 1054 1055 1056 1057 1058 1059
	    GT -> do
		popLexState
		lexToken

-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
maybe_layout ITdo	= pushLexState layout_do
1060
maybe_layout ITmdo	= pushLexState layout_do
1061 1062 1063
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1064
maybe_layout ITrec	= pushLexState layout
1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
maybe_layout _	        = return ()

-- Pushing a new implicit layout context.  If the indentation of the
-- next token is not greater than the previous layout context, then
-- Haskell 98 says that the new layout context should be empty; that is
-- the lexer must generate {}.
--
-- We are slightly more lenient than this: when the new context is started
-- by a 'do', then we allow the new context to be at the same indentation as
-- the previous context.  This is what the 'strict' argument is for.
--
1076
new_layout_context strict span _buf _len = do
1077
    popLexState
1078
    (AI _ offset _) <- getInput
1079 1080 1081 1082 1083 1084 1085 1086
    ctx <- getContext
    case ctx of
	Layout prev_off : _  | 
	   (strict     && prev_off >= offset  ||
	    not strict && prev_off > offset) -> do
		-- token is indented to the left of the previous context.
		-- we must generate a {} sequence now.
		pushLexState layout_left
1087
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1088
	_ -> do
1089
		setContext (Layout offset : ctx)
1090
		return (L span ITvocurly)
1091

1092
do_layout_left span _buf _len = do
1093 1094
    popLexState
    pushLexState bol  -- we must be at the start of a line
1095
    return (L span ITvccurly)
1096 1097 1098 1099

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

1100 1101
setLine :: Int -> Action
setLine code span buf len = do
1102
  let line = parseUnsignedInteger buf len 10 octDecDigit
1103
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1104 1105 1106 1107 1108
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1109 1110
setFile :: Int -> Action
setFile code span buf len = do
1111
  let file = lexemeToFastString (stepOn buf) (len-2)
1112
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1113 1114 1115 1116
  popLexState
  pushLexState code
  lexToken

1117 1118 1119 1120 1121

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1122
lex_string_prag mkTok span _buf _len
1123 1124 1125 1126 1127 1128 1129 1130 1131 1132 1133 1134
    = do input <- getInput
         start <- getSrcLoc
         tok <- go [] input
         end <- getSrcLoc
         return (L (mkSrcSpan start end) tok)
    where go acc input
              = if isString input "#-}"
                   then do setInput input
                           return (mkTok (reverse acc))
                   else case alexGetChar input of
                          Just (c,i) -> go (c:acc) i
                          Nothing -> err input
twanvl's avatar
twanvl committed
1135
          isString _ [] = True
1136 1137 1138 1139 1140 1141 1142
          isString i (x:xs)
              = case alexGetChar i of
                  Just (c,i') | c == x    -> isString i' xs
                  _other -> False
          err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"


1143 1144 1145 1146 1147 1148
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1149
lex_string_tok span _buf _len = do
1150 1151
  tok <- lex_string ""
  end <- getSrcLoc 
1152
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1153

1154
lex_string :: String -> P Token
1155 1156
lex_string s = do
  i <- getInput
1157
  case alexGetChar' i of
1158 1159 1160 1161
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1162 1163
	magicHash <- extension magicHashEnabled
	if magicHash
1164 1165
	  then do
	    i <- getInput
1166
	    case alexGetChar' i of
1167 1168 1169 1170
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1171
                    else let s' = mkZFastString (reverse s) in
1172
			 return (ITprimstring s')
1173 1174
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1175 1176 1177 1178 1179 1180 1181 1182 1183 1184
	      _other ->
		return (ITstring (mkFastString (reverse s)))
	  else
		return (ITstring (mkFastString (reverse s)))

    Just ('\\',i)
	| Just ('&',i) <- next -> do 
		setInput i; lex_string s
	| Just (c,i) <- next, is_space c -> do 
		setInput i; lex_stringgap s
1185
	where next = alexGetChar' i
1186

1187 1188 1189
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1190 1191 1192 1193 1194 1195 1196 1197 1198 1199

lex_stringgap s = do
  c <- getCharOrFail
  case c of
    '\\' -> lex_string s
    c | is_space c -> lex_stringgap s
    _other -> lit_error


lex_char_tok :: Action
1200 1201 1202 1203 1204 1205
-- Here we are basically parsing character literals, such as 'x' or '\n'
-- but, when Template Haskell is on, we additionally spot
-- 'x and ''T, returning ITvarQuote and ITtyQuote respectively, 
-- but WIHTOUT CONSUMING the x or T part  (the parser does that).
-- So we have to do two characters of lookahead: when we see 'x we need to
-- see if there's a trailing quote
twanvl's avatar
twanvl committed
1206
lex_char_tok span _buf _len = do	-- We've seen '
1207
   i1 <- getInput	-- Look ahead to first character
1208
   let loc = srcSpanStart span
1209
   case alexGetChar' i1 of
1210 1211
	Nothing -> lit_error 

1212
	Just ('\'', i2@(AI end2 _ _)) -> do 	-- We've seen ''
1213 1214 1215
		  th_exts <- extension thEnabled
		  if th_exts then do
			setInput i2
1216
			return (L (mkSrcSpan loc end2)  ITtyQuote)
1217 1218
		   else lit_error

twanvl's avatar
twanvl committed
1219
	Just ('\\', i2@(AI _end2 _ _)) -> do 	-- We've seen 'backslash
1220 1221 1222 1223
		  setInput i2
		  lit_ch <- lex_escape
		  mc <- getCharOrFail	-- Trailing quote
		  if mc == '\'' then finish_char_tok loc lit_ch
1224
			        else do setInput i2; lit_error 
1225

twanvl's avatar
twanvl committed
1226
        Just (c, i2@(AI _end2 _ _))