Lexer.x 62.1 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 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
Simon Marlow's avatar
Simon Marlow committed
74 75
import Data.Ratio
import Debug.Trace
76 77
}

78
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
79
$whitechar   = [\ \n\r\f\v $unispace]
80
$white_no_nl = $whitechar # \n
81
$tab         = \t
82 83

$ascdigit  = 0-9
84
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
85
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
86 87 88
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
89
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
90
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
91 92
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

93
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
94
$asclarge  = [A-Z]
95 96
$large     = [$asclarge $unilarge]

97
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
98
$ascsmall  = [a-z]
99 100
$small     = [$ascsmall $unismall \_]

101
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
102
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
103 104

$octit	   = 0-7
105
$hexit     = [$decdigit A-F a-f]
106 107 108 109
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

110 111
$docsym    = [\| \^ \* \$]

112 113 114 115 116 117
@varid     = $small $idchar*
@conid     = $large $idchar*

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

118
@decimal     = $decdigit+
119 120 121 122 123 124 125 126 127
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

128 129 130 131 132
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

133 134 135 136
haskell :-

-- everywhere: skip whitespace and comments
$white_no_nl+ 				;
137
$tab+         { warn Opt_WarnTabs (text "Tab character") }
138 139 140 141 142

-- 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
143 144 145 146
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
147 148 149 150

-- 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.
151 152 153 154 155 156 157 158 159 160

-- 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
161 162
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
163 164 165

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

Jedai's avatar
Jedai committed
166
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
167 168 169 170 171 172 173

-- 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
174
"---"\-* [^$symbol :] .* { lineCommentToken }
175 176 177 178

-- 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
179
"--"\-* / { atEOL } { lineCommentToken }
180 181 182 183

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

Jedai's avatar
Jedai committed
184
"-- " / { atEOL } { lineCommentToken }
185 186 187 188 189 190 191 192 193 194 195 196 197

-- '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 }
198
  ^\# pragma .* \n			; -- GCC 3.3 CPP generated, apparently
199
  ^\# \! .* \n				; -- #!, for scripts
200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221
  ()					{ 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 }

222
<0,option_prags> \n				{ begin bol }
223

224 225
"{-#" $whitechar* (line|LINE) / { notFollowedByPragmaChar }
                            { begin line_prag2 }
226 227 228

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
229
<line_prag1> $decdigit+			{ setLine line_prag1a }
230
<line_prag1a> \" [$graphic \ ]* \"	{ setFile line_prag1b }
231 232 233 234
<line_prag1b> .*			{ pop }

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
235
<line_prag2> $decdigit+			{ setLine line_prag2a }
236
<line_prag2a> \" [$graphic \ ]* \"	{ setFile line_prag2b }
237 238 239
<line_prag2b> "#-}"|"-}"		{ pop }
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
240

241
<0,option_prags> {
242 243 244 245
  "{-#" $whitechar* (RULES|rules)  / { notFollowedByPragmaChar } { rulePrag }
  "{-#" $whitechar* (INLINE|inline)	 / { notFollowedByPragmaChar }
                    { token (ITinline_prag True) }
  "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline) / { notFollowedByPragmaChar }
246
  					{ token (ITinline_prag False) }
247 248 249 250 251 252
  "{-#" $whitechar* (INLINE|inline)
        $whitechar+ (CONLIKE|conlike) / { notFollowedByPragmaChar }
                                        { token (ITinline_conlike_prag True) }
  "{-#" $whitechar* (NO(T)?INLINE|no(t?)inline)
        $whitechar+ (CONLIKE|constructorlike) / { notFollowedByPragmaChar }
                                        { token (ITinline_conlike_prag False) }
253
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) / { notFollowedByPragmaChar }
254
  					{ token ITspec_prag }
255
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
256 257
	$whitechar+ (INLINE|inline) / { notFollowedByPragmaChar }
                    { token (ITspec_inline_prag True) }
258
  "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
259
	$whitechar+ (NO(T?)INLINE|no(t?)inline) / { notFollowedByPragmaChar }
260
					{ token (ITspec_inline_prag False) }
261 262 263
  "{-#" $whitechar* (SOURCE|source) / { notFollowedByPragmaChar }
                    { token ITsource_prag }
  "{-#" $whitechar* (WARNING|warning) / { notFollowedByPragmaChar }
Ian Lynagh's avatar
Ian Lynagh committed
264
  					{ token ITwarning_prag }
265
  "{-#" $whitechar* (DEPRECATED|deprecated) / { notFollowedByPragmaChar }
266
  					{ token ITdeprecated_prag }
267 268 269
  "{-#" $whitechar* (SCC|scc)  / { notFollowedByPragmaChar }
                    { token ITscc_prag }
  "{-#" $whitechar* (GENERATED|generated) / { notFollowedByPragmaChar }
andy@galois.com's avatar
andy@galois.com committed
270
  					{ token ITgenerated_prag }
271 272 273 274 275 276
  "{-#" $whitechar* (CORE|core) / { notFollowedByPragmaChar }
                    { token ITcore_prag }
  "{-#" $whitechar* (UNPACK|unpack) / { notFollowedByPragmaChar }
                    { token ITunpack_prag }
  "{-#" $whitechar* (ANN|ann) / { notFollowedByPragmaChar }
                    { token ITann_prag }
277

278 279
  -- We ignore all these pragmas, but don't generate a warning for them
  -- CFILES is a hugs-only thing.
280
  "{-#" $whitechar* (OPTIONS_(HUGS|hugs|NHC98|nhc98|JHC|jhc|YHC|yhc|CATCH|catch|DERIVE|derive)|CFILES|cfiles|CONTRACT|contract) / { notFollowedByPragmaChar }
281
                    { nested_comment lexToken }
282 283

  -- ToDo: should only be valid inside a pragma:
284
  "#-}" 				{ endPrag }
285 286
}

287
<option_prags> {
288 289 290
  "{-#"  $whitechar* (OPTIONS|options) / { notFollowedByPragmaChar }
                                        { lex_string_prag IToptions_prag }
  "{-#"  $whitechar* (OPTIONS_GHC|options_ghc) / { notFollowedByPragmaChar }
291
                                        { lex_string_prag IToptions_prag }
David Waern's avatar
David Waern committed
292
  "{-#"  $whitechar* (OPTIONS_HADDOCK|options_haddock)
293
                   / { notFollowedByPragmaChar }
David Waern's avatar
David Waern committed
294 295
                                         { lex_string_prag ITdocOptions }
  "-- #"                                 { multiline_doc_comment }
296 297 298 299
  "{-#"  $whitechar* (LANGUAGE|language) / { notFollowedByPragmaChar }
                                         { token ITlanguage_prag }
  "{-#"  $whitechar* (INCLUDE|include) / { notFollowedByPragmaChar }
                                         { lex_string_prag ITinclude_prag }
David Waern's avatar
David Waern committed
300 301
}

302 303
<0> {
  -- In the "0" mode we ignore these pragmas
304
  "{-#"  $whitechar* (OPTIONS|options|OPTIONS_GHC|options_ghc|OPTIONS_HADDOCK|options_haddock|LANGUAGE|language|INCLUDE|include) / { notFollowedByPragmaChar }
305 306 307
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
308
<0> {
Jedai's avatar
Jedai committed
309
  "-- #" .* { lineCommentToken }
310
}
311

312
<0,option_prags> {
313 314
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
315 316
}

317 318
-- '0' state: ordinary lexemes

319 320
-- Haddock comments

321
<0> {
David Waern's avatar
David Waern committed
322 323
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
324 325
}

326 327
-- "special" symbols

328
<0> {
329 330 331 332
  "[:" / { ifExtension parrEnabled }	{ token ITopabrack }
  ":]" / { ifExtension parrEnabled }	{ token ITcpabrack }
}
  
333
<0> {
334 335 336 337 338 339 340 341
  "[|"	    / { 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 }
342 343 344

  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }
345 346
}

347
<0> {
ross's avatar
ross committed
348 349
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
					{ special IToparenbar }
350 351 352
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

353
<0> {
354 355 356
  \? @varid / { ifExtension ipEnabled }	{ skip_one_varid ITdupipvarid }
}

357
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
358 359 360 361 362 363
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

364
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
365 366
  "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
  "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
367 368
}

369
<0,option_prags> {
370 371 372 373 374 375 376 377 378 379 380 381
  \(					{ special IToparen }
  \)					{ special ITcparen }
  \[					{ special ITobrack }
  \]					{ special ITcbrack }
  \,					{ special ITcomma }
  \;					{ special ITsemi }
  \`					{ special ITbackquote }
 				
  \{					{ open_brace }
  \}					{ close_brace }
}

382
<0,option_prags> {
383
  @qual @varid			{ idtoken qvarid }
384 385 386 387 388
  @qual @conid			{ idtoken qconid }
  @varid			{ varid }
  @conid			{ idtoken conid }
}

389
<0> {
390 391 392 393
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
394 395
}

396 397
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
398
<0> {
399 400 401 402 403 404
  @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 }
405 406
}

407 408
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
409
<0> {
410
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
411 412 413
  @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 }
414 415

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
416
  @floating_point    { strtoken tok_float }
417 418
}

419
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
420
  -- Unboxed ints (:: Int#) and words (:: Word#)
421 422
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
423 424 425 426 427
  @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 }
428
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
429

Ian Lynagh's avatar
Ian Lynagh committed
430 431 432 433
  @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 }

434 435
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
436 437
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
438 439 440 441 442 443
}

-- 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.
444
<0> {
445 446 447 448 449 450 451 452
  \'				{ lex_char_tok }
  \" 				{ lex_string_tok }
}

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

453
data Token
454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491
  = 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
492
  | ITfamily
493 494 495
  | ITgroup
  | ITby
  | ITusing
496

497 498
	-- Pragmas
  | ITinline_prag Bool		-- True <=> INLINE, False <=> NOINLINE
499
  | ITinline_conlike_prag Bool  -- same
500 501
  | ITspec_prag			-- SPECIALISE	
  | ITspec_inline_prag Bool	-- SPECIALISE INLINE (or NOINLINE)
502 503
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
504
  | ITwarning_prag
505 506 507
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
508
  | ITgenerated_prag
509
  | ITcore_prag                 -- hdaume: core annotations
510
  | ITunpack_prag
511
  | ITann_prag
512
  | ITclose_prag
513 514 515
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541

  | 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
542 543
  | ITopabrack			-- [:, for parallel arrays with -XParr
  | ITcpabrack			-- :], for parallel arrays with -XParr
544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561
  | 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)
562 563
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
564 565 566 567 568 569 570 571 572 573 574 575 576

  | 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
577
  | ITprimword   Integer
578 579 580 581
  | ITprimfloat  Rational
  | ITprimdouble Rational

  -- MetaHaskell extension tokens
582 583 584 585 586 587 588 589 590
  | ITopenExpQuote  		--  [| or [e|
  | ITopenPatQuote		--  [p|
  | ITopenDecQuote		--  [d|
  | ITopenTypQuote		--  [t|         
  | ITcloseQuote		--  |]
  | ITidEscape   FastString	--  $x
  | ITparenEscape		--  $( 
  | ITvarQuote			--  '
  | ITtyQuote			--  ''
591
  | ITquasiQuote (FastString,FastString,SrcSpan) --  [:...|...|]
592 593 594 595

  -- Arrow notation extension
  | ITproc
  | ITrec
596 597 598 599 600 601
  | IToparenbar			--  (|
  | ITcparenbar			--  |)
  | ITlarrowtail		--  -<
  | ITrarrowtail		--  >-
  | ITLarrowtail		--  -<<
  | ITRarrowtail		--  >>-
602 603 604

  | ITunknown String		-- Used when the lexer can't make sense of it
  | ITeof			-- end of file token
605 606 607 608 609 610 611

  -- 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
612
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
613 614
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
615

616 617 618 619
#ifdef DEBUG
  deriving Show -- debugging
#endif

twanvl's avatar
twanvl committed
620
{-
621
isSpecial :: Token -> Bool
622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637
-- 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
638
isSpecial ITfamily	= True
639 640 641
isSpecial ITgroup   = True
isSpecial ITby      = True
isSpecial ITusing   = True
642
isSpecial _             = False
twanvl's avatar
twanvl committed
643
-}
644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680

-- 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

681
    ( "forall",	ITforall,	 bit explicitForallBit .|. bit inRulePragBit),
Ian Lynagh's avatar
Ian Lynagh committed
682
	( "mdo",	ITmdo,		 bit recursiveDoBit),
683
	( "family",	ITfamily,	 bit tyFamBit),
684 685 686
    ( "group",  ITgroup,     bit transformComprehensionsBit),
    ( "by",     ITby,        bit transformComprehensionsBit),
    ( "using",  ITusing,     bit transformComprehensionsBit),
687 688 689 690 691 692 693 694 695 696 697 698 699

	( "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),
700
	( "proc",	ITproc,		 bit arrowsBit)
701 702
     ]

Ian Lynagh's avatar
Ian Lynagh committed
703
reservedSymsFM :: UniqFM (Token, Int -> Bool)
704
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721
    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
722
       ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
723
        -- For 'forall a . t'
724
       ,(".", ITdot, \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
725 726 727 728 729

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
730

Ian Lynagh's avatar
Ian Lynagh committed
731 732
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
733 734
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
735 736 737
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
       ,("⋯",   ITdotdot, unicodeSyntaxEnabled)
Simon Marlow's avatar
Simon Marlow committed
738 739 740
        -- 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).
741 742 743 744 745
       ]

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

746
type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
747

748
special :: Token -> Action
twanvl's avatar
twanvl committed
749
special tok span _buf _len = return (L span tok)
750

751
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
752 753
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
754

755 756
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
757

758 759 760
skip_one_varid :: (FastString -> Token) -> Action
skip_one_varid f span buf len 
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
761

762 763 764
strtoken :: (String -> Token) -> Action
strtoken f span buf len = 
  return (L span $! (f $! lexemeToString buf len))
765

766
init_strtoken :: Int -> (String -> Token) -> Action
767
-- like strtoken, but drops the last N character(s)
768 769
init_strtoken drop f span buf len = 
  return (L span $! (f $! lexemeToString buf (len-drop)))
770 771

begin :: Int -> Action
772
begin code _span _str _len = do pushLexState code; lexToken
773 774

pop :: Action
775
pop _span _buf _len = do popLexState; lexToken
776 777

pop_and :: Action -> Action
778
pop_and act span buf len = do popLexState; act span buf len
779

780 781 782 783 784
{-# INLINE nextCharIs #-}
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

notFollowedBy char _ _ _ (AI _ _ buf) 
  = nextCharIs buf (/=char)
785

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

789 790 791
notFollowedByPragmaChar _ _ _ (AI _ _ buf)
  = nextCharIs buf (\c -> not (isAlphaNum c || c == '_'))

792 793 794 795 796
-- 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.
797
isNormalComment bits _ _ (AI _ _ buf)
798 799 800 801 802 803 804
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
805

twanvl's avatar
twanvl committed
806
{-
807 808
haddockDisabledAnd p bits _ _ (AI _ _ buf)
  = if haddockEnabled bits then False else (p buf)
twanvl's avatar
twanvl committed
809
-}
810

811
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
812

813 814
ifExtension pred bits _ _ _ = pred bits

815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842
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
843 844 845 846 847
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

848 849 850 851
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
852 853
nested_comment :: P (Located Token) -> Action
nested_comment cont span _str _len = do
854
  input <- getInput
Jedai's avatar
Jedai committed
855
  go "" (1::Int) input
856
  where
Jedai's avatar
Jedai committed
857 858 859 860 861 862
    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
863 864 865
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
866 867
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
868 869
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
870 871 872
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
873 874 875 876 877 878 879 880

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
881
        Just ('\125',input) ->
882
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
883
        Just (_,_) -> go ('-':commentAcc) input docType False
884 885 886 887 888 889
      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
890
        Just (_,_) -> go ('\123':commentAcc) input docType False
891 892 893
      Just (c,input) -> go (c:commentAcc) input docType False

withLexedDocType lexDocComment = do
894 895 896 897 898
  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
899 900
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
901 902 903
 where 
    lexDocSection n input = case alexGetChar input of 
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
904
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
905 906
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

907 908 909 910
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
rulePrag span buf len = do
911
  setExts (.|. bit inRulePragBit)
912 913 914 915 916 917 918
  return (L span ITrules_prag)

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

919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952
-- 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 `{-'"
 
953
open_brace, close_brace :: Action
954
open_brace span _str _len = do 
955 956
  ctx <- getContext
  setContext (NoLayout:ctx)
957 958
  return (L span ITocurly)
close_brace span _str _len = do 
959
  popContext
960
  return (L span ITccurly)
961

962 963
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
964

965
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
966 967 968
-- 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.
969
splitQualName orig_buf len parens = split orig_buf orig_buf
970
  where
971 972 973 974 975 976
    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
977 978 979 980
  
    -- 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.
981 982 983 984 985 986 987 988
    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),
989 990 991
	 if parens -- Prelude.(+)
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
992 993
      where
	qual_size = orig_buf `byteDiff` dot_buf
994

995
varid span buf len = 
Ian Lynagh's avatar
Ian Lynagh committed
996
  fs `seq`
997 998 999
  case lookupUFM reservedWordsFM fs of
	Just (keyword,0)    -> do
		maybe_layout keyword
1000
		return (L span keyword)
1001 1002 1003
	Just (keyword,exts) -> do
		b <- extension (\i -> exts .&. i /= 0)
		if b then do maybe_layout keyword
1004 1005 1006
			     return (L span keyword)
		     else return (L span (ITvarid fs))
	_other -> return (L span (ITvarid fs))
1007 1008 1009 1010 1011 1012
  where
	fs = lexemeToFastString buf len

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

1013 1014 1015 1016
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
1017 1018 1019 1020

varsym = sym ITvarsym
consym = sym ITconsym

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 1042 1043
-- 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
1044
tok_primword = tok_integral ITprimword positive
1045 1046 1047 1048 1049 1050 1051
positive = id
negative = negate
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
1052
tok_float        str = ITrational   $! readRational str
1053 1054
tok_primfloat    str = ITprimfloat  $! readRational str
tok_primdouble   str = ITprimdouble $! readRational str
1055 1056 1057 1058 1059 1060

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1061
do_bol span _str _len = do
1062
	pos <- getOffside
1063 1064 1065 1066 1067
	case pos of
	    LT -> do
                --trace "layout: inserting '}'" $ do
		popContext
		-- do NOT pop the lex state, we might have a ';' to insert
1068
		return (L span ITvccurly)
1069 1070 1071
	    EQ -> do
                --trace "layout: inserting ';'" $ do
		popLexState
1072
		return (L span ITsemi)
1073 1074 1075 1076 1077 1078 1079
	    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
1080
maybe_layout ITmdo	= pushLexState layout_do
1081 1082 1083
maybe_layout ITof	= pushLexState layout
maybe_layout ITlet	= pushLexState layout
maybe_layout ITwhere	= pushLexState layout
1084
maybe_layout ITrec	= pushLexState layout
1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095
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.
--
1096
new_layout_context strict span _buf _len = do
1097
    popLexState
1098
    (AI _ offset _) <- getInput
1099 1100 1101 1102 1103 1104 1105 1106
    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
1107
		return (L span ITvocurly)
twanvl's avatar
twanvl committed
1108
	_ -> do
1109
		setContext (Layout offset : ctx)
1110
		return (L span ITvocurly)
1111

1112
do_layout_left span _buf _len = do
1113 1114
    popLexState
    pushLexState bol  -- we must be at the start of a line
1115
    return (L span ITvccurly)
1116 1117 1118 1119

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

1120 1121
setLine :: Int -> Action
setLine code span buf len = do
1122
  let line = parseUnsignedInteger buf len 10 octDecDigit
1123
  setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1124 1125 1126 1127 1128
	-- subtract one: the line number refers to the *following* line
  popLexState
  pushLexState code
  lexToken

1129 1130
setFile :: Int -> Action
setFile code span buf len = do
1131
  let file = lexemeToFastString (stepOn buf) (len-2)
1132
  setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1133 1134 1135 1136
  popLexState
  pushLexState code
  lexToken

1137 1138 1139 1140 1141

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

lex_string_prag :: (String -> Token) -> Action
twanvl's avatar
twanvl committed
1142
lex_string_prag mkTok span _buf _len
1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154
    = 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
1155
          isString _ [] = True
1156 1157 1158 1159 1160 1161 1162
          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"


1163 1164 1165 1166 1167 1168
-- -----------------------------------------------------------------------------
-- Strings & Chars

-- This stuff is horrible.  I hates it.

lex_string_tok :: Action
twanvl's avatar
twanvl committed
1169
lex_string_tok span _buf _len = do
1170 1171
  tok <- lex_string ""
  end <- getSrcLoc 
1172
  return (L (mkSrcSpan (srcSpanStart span) end) tok)
1173

1174
lex_string :: String -> P Token
1175 1176
lex_string s = do
  i <- getInput
1177
  case alexGetChar' i of
1178 1179 1180 1181
    Nothing -> lit_error

    Just ('"',i)  -> do
	setInput i
1182 1183
	magicHash <- extension magicHashEnabled
	if magicHash
1184 1185
	  then do
	    i <- getInput
1186
	    case alexGetChar' i of
1187 1188 1189 1190
	      Just ('#',i) -> do
		   setInput i
		   if any (> '\xFF') s
                    then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1191
                    else let s' = mkZFastString (reverse s) in
1192
			 return (ITprimstring s')
1193 1194
			-- mkZFastString is a hack to avoid encoding the
			-- string in UTF-8.  We just want the exact bytes.
1195 1196 1197 1198 1199 1200 1201 1202 1203 1204
	      _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
1205
	where next = alexGetChar' i
1206

1207 1208 1209
    Just (c, i) -> do
	c' <- lex_char c i
	lex_string (c':s)
1210 1211 1212 1213 1214 1215