Lexer.x 86.2 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3 4 5 6 7 8 9
--
-- GHC's lexer.
--
-- This is a combination of an Alex-generated lexer from a regex
-- definition, with some hand-coded bits.
--
-- Completely accurate information about token-spans within the source
Ian Lynagh's avatar
Ian Lynagh committed
10 11
-- file is maintained.  Every token has a start and end RealSrcLoc
-- attached to it.
12 13 14 15 16 17 18 19 20 21 22
--
-----------------------------------------------------------------------------

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

23
--   qualified operator NOTES.
Ian Lynagh's avatar
Ian Lynagh committed
24
--
25 26 27 28 29 30 31 32 33
--   - If M.(+) is a single lexeme, then..
--     - Probably (+) should be a single lexeme too, for consistency.
--       Otherwise ( + ) would be a prefix operator, but M.( + ) would not be.
--     - But we have to rule out reserved operators, otherwise (..) becomes
--       a different lexeme.
--     - Should we therefore also rule out reserved operators in the qualified
--       form?  This is quite difficult to achieve.  We don't do it for
--       qualified varids.

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

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

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

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

75
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
76
import Data.Bits
77
import Data.Char
78
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
79
import Data.Maybe
80 81
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
82
import Data.Ratio
83
import Data.Word
84 85
}

86
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
87
$whitechar   = [\ \n\r\f\v $unispace]
88
$white_no_nl = $whitechar # \n
89
$tab         = \t
90 91

$ascdigit  = 0-9
92
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
93
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
94 95 96
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
97
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
98
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
99 100
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

101
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
102
$asclarge  = [A-Z]
103 104
$large     = [$asclarge $unilarge]

105
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
106
$ascsmall  = [a-z]
107 108
$small     = [$ascsmall $unismall \_]

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

Ian Lynagh's avatar
Ian Lynagh committed
112
$octit     = 0-7
113
$hexit     = [$decdigit A-F a-f]
114 115 116 117
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

118 119
$pragmachar = [$small $large $digit]

120 121
$docsym    = [\| \^ \* \$]

122 123 124 125 126 127
@varid     = $small $idchar*
@conid     = $large $idchar*

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

128
@decimal     = $decdigit+
129 130 131 132 133 134 135 136 137
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

138 139 140 141 142
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

143 144 145
haskell :-

-- everywhere: skip whitespace and comments
Ian Lynagh's avatar
Ian Lynagh committed
146
$white_no_nl+ ;
147
$tab+         { warn Opt_WarnTabs (text "Warning: Tab character") }
148 149 150 151 152

-- 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
153 154 155 156
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
157 158 159 160

-- 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.
161 162

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
163
-- out here.
164 165 166 167 168 169 170

-- 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
171 172
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
173 174 175

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

Jedai's avatar
Jedai committed
176
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
177 178 179 180 181 182 183

-- 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
184
"---"\-* [^$symbol :] .* { lineCommentToken }
185 186 187 188

-- 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
189
"--"\-* / { atEOL } { lineCommentToken }
190 191 192 193

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

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

-- '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> {
Ian Lynagh's avatar
Ian Lynagh committed
206 207 208 209 210
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
211 212 213 214 215 216
}

-- 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> {
Ian Lynagh's avatar
Ian Lynagh committed
217 218 219 220
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
221 222 223
}

-- do is treated in a subtly different way, see new_layout_context
Ian Lynagh's avatar
Ian Lynagh committed
224 225
<layout>    ()                          { new_layout_context True }
<layout_do> ()                          { new_layout_context False }
226 227 228 229

-- 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.
Ian Lynagh's avatar
Ian Lynagh committed
230
<layout_left>  ()                       { do_layout_left }
231

Ian Lynagh's avatar
Ian Lynagh committed
232
<0,option_prags> \n                     { begin bol }
233

234 235
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
236 237 238

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
Ian Lynagh's avatar
Ian Lynagh committed
239 240 241
<line_prag1> $decdigit+                 { setLine line_prag1a }
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
242 243 244

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
Ian Lynagh's avatar
Ian Lynagh committed
245 246 247
<line_prag2> $decdigit+                 { setLine line_prag2a }
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
248 249
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
250

251
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
252
  "{-#" $whitechar* $pragmachar+
253 254 255 256 257
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
258

259
  -- We ignore all these pragmas, but don't generate a warning for them
260 261
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
262 263

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
264
  "#-}"                          { endPrag }
265 266
}

267
<option_prags> {
268 269 270
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

Ian Lynagh's avatar
Ian Lynagh committed
271
  "-- #"                           { multiline_doc_comment }
David Waern's avatar
David Waern committed
272 273
}

274 275
<0> {
  -- In the "0" mode we ignore these pragmas
276
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
277 278 279
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
280
<0> {
Jedai's avatar
Jedai committed
281
  "-- #" .* { lineCommentToken }
282
}
283

284
<0,option_prags> {
285 286
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
287 288
}

289 290
-- '0' state: ordinary lexemes

291 292
-- Haddock comments

293
<0,option_prags> {
David Waern's avatar
David Waern committed
294 295
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
296 297
}

298 299
-- "special" symbols

300
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
301 302
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
303
}
Ian Lynagh's avatar
Ian Lynagh committed
304

305
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
306 307 308 309 310 311 312 313
  "[|"      / { 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 }
314

315 316 317 318
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

319
  "[" @varid "|"  / { ifExtension qqEnabled }
320
                     { lex_quasiquote_tok }
321 322
}

323
<0> {
ross's avatar
ross committed
324
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
Ian Lynagh's avatar
Ian Lynagh committed
325
                                        { special IToparenbar }
326 327 328
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

329
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
330
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
331 332
}

333
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
334 335 336 337 338 339
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

340
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
341 342 343 344 345 346 347 348 349 350
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
351 352
}

353
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
354 355 356 357
  @qual @varid                  { idtoken qvarid }
  @qual @conid                  { idtoken qconid }
  @varid                        { varid }
  @conid                        { idtoken conid }
358 359
}

360
<0> {
361 362 363 364
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
365 366
}

367 368
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
369
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
370 371
  @qual @varsym                                    { idtoken qvarsym }
  @qual @consym                                    { idtoken qconsym }
372 373
  @varsym                                          { varsym }
  @consym                                          { consym }
374 375
}

376 377
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
378
<0> {
379
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
380 381 382
  @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 }
383 384

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
385
  @floating_point    { strtoken tok_float }
386 387
}

388
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
389
  -- Unboxed ints (:: Int#) and words (:: Word#)
390 391
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
392 393 394 395 396
  @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 }
397
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
398

Ian Lynagh's avatar
Ian Lynagh committed
399 400 401 402
  @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 }

403 404
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
405 406
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
407 408 409 410 411 412
}

-- 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.
413
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
414 415
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
416 417 418 419 420 421
}

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

422
data Token
Ian Lynagh's avatar
Ian Lynagh committed
423
  = ITas                        -- Haskell keywords
424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
  | ITdo
  | ITelse
  | IThiding
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere
Ian Lynagh's avatar
Ian Lynagh committed
447
  | ITscc                       -- ToDo: remove (we use {-# SCC "..." #-} now)
448

Ian Lynagh's avatar
Ian Lynagh committed
449
  | ITforall                    -- GHC extension keywords
450 451 452 453 454
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
455
  | ITinterruptible
456 457 458
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
459
  | ITprimcallconv
460
  | ITmdo
461
  | ITfamily
462 463 464
  | ITgroup
  | ITby
  | ITusing
465

Ian Lynagh's avatar
Ian Lynagh committed
466
  -- Pragmas
467
  | ITinline_prag InlineSpec RuleMatchInfo
Ian Lynagh's avatar
Ian Lynagh committed
468 469
  | ITspec_prag                 -- SPECIALISE
  | ITspec_inline_prag Bool     -- SPECIALISE INLINE (or NOINLINE)
470 471
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
472
  | ITwarning_prag
473 474 475
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
476
  | ITgenerated_prag
477
  | ITcore_prag                 -- hdaume: core annotations
478
  | ITunpack_prag
479
  | ITann_prag
480
  | ITclose_prag
481 482 483
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
484 485
  | ITvect_prag
  | ITvect_scalar_prag
486
  | ITnovect_prag
487

Ian Lynagh's avatar
Ian Lynagh committed
488
  | ITdotdot                    -- reserved symbols
489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
504
  | ITbiglam                    -- GHC-extension symbols
505

Ian Lynagh's avatar
Ian Lynagh committed
506
  | ITocurly                    -- special symbols
507 508 509 510 511 512
  | ITccurly
  | ITocurlybar                 -- {|, for type applications
  | ITccurlybar                 -- |}, for type applications
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
513 514
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
515 516 517 518 519 520 521 522 523 524
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote

Ian Lynagh's avatar
Ian Lynagh committed
525
  | ITvarid   FastString        -- identifiers
526 527 528 529 530 531 532
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)
533 534
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
535

Ian Lynagh's avatar
Ian Lynagh committed
536
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
537 538 539 540

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
541
  | ITrational   FractionalLit
542 543 544 545

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
546
  | ITprimword   Integer
547 548
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
549

550
  -- Template Haskell extension tokens
Ian Lynagh's avatar
Ian Lynagh committed
551 552 553 554 555 556 557 558 559
  | ITopenExpQuote              --  [| or [e|
  | ITopenPatQuote              --  [p|
  | ITopenDecQuote              --  [d|
  | ITopenTypQuote              --  [t|
  | ITcloseQuote                --  |]
  | ITidEscape   FastString     --  $x
  | ITparenEscape               --  $(
  | ITvarQuote                  --  '
  | ITtyQuote                   --  ''
Ian Lynagh's avatar
Ian Lynagh committed
560
  | ITquasiQuote (FastString,FastString,RealSrcSpan) --  [:...|...|]
561 562 563 564

  -- Arrow notation extension
  | ITproc
  | ITrec
Ian Lynagh's avatar
Ian Lynagh committed
565 566 567 568 569 570
  | IToparenbar                 --  (|
  | ITcparenbar                 --  |)
  | ITlarrowtail                --  -<
  | ITrarrowtail                --  >-
  | ITLarrowtail                --  -<<
  | ITRarrowtail                --  >>-
571

Ian Lynagh's avatar
Ian Lynagh committed
572 573
  | ITunknown String            -- Used when the lexer can't make sense of it
  | ITeof                       -- end of file token
574 575 576 577 578 579 580

  -- 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
581
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
582 583
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
584

585 586 587 588 589 590 591 592 593 594 595
#ifdef DEBUG
  deriving Show -- debugging
#endif

-- the bitmap provided as the third component indicates whether the
-- corresponding extension keyword is valid under the extension options
-- provided to the compiler; if the extension corresponding to *any* of the
-- bits set in the bitmap is enabled, the keyword is valid (this setup
-- facilitates using a keyword in two different extensions that can be
-- activated independently)
--
Ian Lynagh's avatar
Ian Lynagh committed
596
reservedWordsFM :: UniqFM (Token, Int)
597
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647
    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

         ( "forall",         ITforall,        bit explicitForallBit .|.
                                              bit inRulePragBit),
         ( "mdo",            ITmdo,           bit recursiveDoBit),
         ( "family",         ITfamily,        bit tyFamBit),
         ( "group",          ITgroup,         bit transformComprehensionsBit),
         ( "by",             ITby,            bit transformComprehensionsBit),
         ( "using",          ITusing,         bit transformComprehensionsBit),

         ( "foreign",        ITforeign,       bit ffiBit),
         ( "export",         ITexport,        bit ffiBit),
         ( "label",          ITlabel,         bit ffiBit),
         ( "dynamic",        ITdynamic,       bit ffiBit),
         ( "safe",           ITsafe,          bit ffiBit .|.
                                              bit safeHaskellBit),
         ( "interruptible",  ITinterruptible, bit interruptibleFfiBit),
         ( "unsafe",         ITunsafe,        bit ffiBit),
         ( "stdcall",        ITstdcallconv,   bit ffiBit),
         ( "ccall",          ITccallconv,     bit ffiBit),
         ( "prim",           ITprimcallconv,  bit ffiBit),

         ( "rec",            ITrec,           bit recBit),
         ( "proc",           ITproc,          bit arrowsBit)
648 649
     ]

Ian Lynagh's avatar
Ian Lynagh committed
650
reservedSymsFM :: UniqFM (Token, Int -> Bool)
651
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668
    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
669
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
670
        -- For 'forall a . t'
671
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
672 673 674 675 676

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
677

Ian Lynagh's avatar
Ian Lynagh committed
678 679
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
680 681
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
682 683
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
684 685 686 687 688 689 690 691

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

       ,("★", ITstar, unicodeSyntaxEnabled)

Simon Marlow's avatar
Simon Marlow committed
692 693 694
        -- 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).
695 696 697 698 699
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
700
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
701

702
special :: Token -> Action
twanvl's avatar
twanvl committed
703
special tok span _buf _len = return (L span tok)
704

705
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
706 707
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
708

709 710
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
711

712
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
713
skip_one_varid f span buf len
714
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
715

716
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
717
strtoken f span buf len =
718
  return (L span $! (f $! lexemeToString buf len))
719

720
init_strtoken :: Int -> (String -> Token) -> Action
721
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
722
init_strtoken drop f span buf len =
723
  return (L span $! (f $! lexemeToString buf (len-drop)))
724 725

begin :: Int -> Action
726
begin code _span _str _len = do pushLexState code; lexToken
727 728

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
729 730
pop _span _buf _len = do _ <- popLexState
                         lexToken
731

732 733 734 735 736 737 738 739 740 741 742
hopefully_open_brace :: Action
hopefully_open_brace span buf len
 = do relaxed <- extension relaxedLayout
      ctx <- getContext
      (AI l _) <- getInput
      let offset = srcLocCol l
          isOK = relaxed ||
                 case ctx of
                 Layout prev_off : _ -> prev_off < offset
                 _                   -> True
      if isOK then pop_and open_brace span buf len
Ian Lynagh's avatar
Ian Lynagh committed
743
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
744

745
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
746 747
pop_and act span buf len = do _ <- popLexState
                              act span buf len
748

749
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
750
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
751 752
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
753
notFollowedBy :: Char -> AlexAccPred Int
Ian Lynagh's avatar
Ian Lynagh committed
754
notFollowedBy char _ _ _ (AI _ buf)
755
  = nextCharIs buf (/=char)
756

Ian Lynagh's avatar
Ian Lynagh committed
757
notFollowedBySymbol :: AlexAccPred Int
758
notFollowedBySymbol _ _ _ (AI _ buf)
759
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
760

761 762 763 764 765
-- We must reject doc comments as being ordinary comments everywhere.
-- In some cases the doc comment will be selected as the lexeme due to
-- maximal munch, but not always, because the nested comment rule is
-- valid in all states, but the doc-comment rules are only valid in
-- the non-layout states.
Ian Lynagh's avatar
Ian Lynagh committed
766
isNormalComment :: AlexAccPred Int
767
isNormalComment bits _ _ (AI _ buf)
768 769 770 771 772 773
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
774
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
775
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
776

Ian Lynagh's avatar
Ian Lynagh committed
777
atEOL :: AlexAccPred Int
778
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
779

Ian Lynagh's avatar
Ian Lynagh committed
780
ifExtension :: (Int -> Bool) -> AlexAccPred Int
781 782
ifExtension pred bits _ _ _ = pred bits

783 784 785 786
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
    worker commentAcc input docType oneLine = case alexGetChar input of
Ian Lynagh's avatar
Ian Lynagh committed
787
      Just ('\n', input')
788 789 790 791 792 793
        | 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
Ian Lynagh's avatar
Ian Lynagh committed
794

795 796 797 798 799 800 801 802 803 804 805
    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
Ian Lynagh's avatar
Ian Lynagh committed
806
          Just (c, input')
807 808 809 810
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
811 812 813 814 815
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

816 817 818 819
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
820
nested_comment :: P (RealLocated Token) -> Action
821
nested_comment cont span _str _len = do
822
  input <- getInput
Jedai's avatar
Jedai committed
823
  go "" (1::Int) input
824
  where
Jedai's avatar
Jedai committed
825 826 827 828 829 830
    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
831 832 833
      Nothing -> errBrace input span
      Just ('-',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
834 835
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
836 837
      Just ('\123',input) -> case alexGetChar input of
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
838 839 840
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
841 842 843 844 845 846 847 848

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
849
        Just ('\125',input) ->
850
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
851
        Just (_,_) -> go ('-':commentAcc) input docType False
852 853 854 855 856 857
      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
858
        Just (_,_) -> go ('\123':commentAcc) input docType False
859 860
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
861 862
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
863
withLexedDocType lexDocComment = do
864
  input@(AI _ buf) <- getInput
865 866 867 868
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
869 870
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
871
    _ -> panic "withLexedDocType: Bad doc type"
Ian Lynagh's avatar
Ian Lynagh committed
872 873
 where
    lexDocSection n input = case alexGetChar input of
874
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
875
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
876 877
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

878 879 880
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
881
rulePrag span _buf _len = do
882
  setExts (.|. bit inRulePragBit)
883 884 885
  return (L span ITrules_prag)

endPrag :: Action
886
endPrag span _buf _len = do
887 888 889
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

890 891 892 893
-- 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
Ian Lynagh's avatar
Ian Lynagh committed
894
-- than what was lexed by Alex, and the lexToken function doesn't know this, so
895
-- it writes the wrong token length to the parser state. This function is
Ian Lynagh's avatar
Ian Lynagh committed
896
-- called afterwards, so it can just update the state.
897 898

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
899
                 RealSrcSpan -> P (RealLocated Token)
900 901
docCommentEnd input commentAcc docType buf span = do
  setInput input
902
  let (AI loc nextBuf) = input
903
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
904
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
905
      last_len = byteDiff buf nextBuf
Ian Lynagh's avatar
Ian Lynagh committed
906

907
  span `seq` setLastToken span' last_len
908
  return (L span' (docType comment))
Ian Lynagh's avatar
Ian Lynagh committed
909

Ian Lynagh's avatar
Ian Lynagh committed
910 911
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
912

913
open_brace, close_brace :: Action
Ian Lynagh's avatar
Ian Lynagh committed
914
open_brace span _str _len = do
915 916
  ctx <- getContext
  setContext (NoLayout:ctx)
917
  return (L span ITocurly)
Ian Lynagh's avatar
Ian Lynagh committed
918
close_brace span _str _len = do
919
  popContext
920
  return (L span ITccurly)
921

Ian Lynagh's avatar
Ian Lynagh committed
922
qvarid, qconid :: StringBuffer -> Int -> Token
923 924
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
925

926
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
927 928 929
-- 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.
930
splitQualName orig_buf len parens = split orig_buf orig_buf
931
  where
932
    split buf dot_buf
Ian Lynagh's avatar
Ian Lynagh committed
933 934 935
        | orig_buf `byteDiff` buf >= len  = done dot_buf
        | c == '.'                        = found_dot buf'
        | otherwise                       = split buf' dot_buf
936 937
      where
       (c,buf') = nextChar buf
Ian Lynagh's avatar
Ian Lynagh committed
938

939 940 941
    -- 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.
942
    found_dot buf -- buf points after the '.'
Ian Lynagh's avatar
Ian Lynagh committed
943 944
        | isUpper c    = split buf' buf
        | otherwise    = done buf
945 946 947 948
      where
       (c,buf') = nextChar buf

    done dot_buf =
Ian Lynagh's avatar
Ian Lynagh committed
949 950
        (lexemeToFastString orig_buf (qual_size - 1),
         if parens -- Prelude.(+)
951 952
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
953
      where
Ian Lynagh's avatar
Ian Lynagh committed
954
        qual_size = orig_buf `byteDiff` dot_buf
955

Ian Lynagh's avatar
Ian Lynagh committed
956 957
varid :: Action
varid span buf len =
Ian Lynagh's avatar
Ian Lynagh committed
958
  fs `seq`
959
  case lookupUFM reservedWordsFM fs of
Ian Lynagh's avatar
Ian Lynagh committed
960 961 962 963 964 965 966 967 968
        Just (keyword,0)    -> do
                maybe_layout keyword
                return (L span keyword)
        Just (keyword,exts) -> do
                b <- extension (\i -> exts .&. i /= 0)
                if b then do maybe_layout keyword
                             return (L span keyword)
                     else return (L span (ITvarid fs))
        _other -> return (L span (ITvarid fs))
969
  where
Ian Lynagh's avatar
Ian Lynagh committed
970
        fs = lexemeToFastString buf len
971

Ian Lynagh's avatar
Ian Lynagh committed
972
conid :: StringBuffer -> Int -> Token
973 974 975
conid buf len = ITconid fs
  where fs = lexemeToFastString buf len

Ian Lynagh's avatar
Ian Lynagh committed
976
qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
977 978 979 980
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
981

Ian Lynagh's avatar
Ian Lynagh committed
982
varsym, consym :: Action
983 984 985
varsym = sym ITvarsym
consym = sym ITconsym

Ian Lynagh's avatar
Ian Lynagh committed
986 987
sym :: (FastString -> Token) -> RealSrcSpan -> StringBuffer -> Int
    -> P (RealLocated Token)
Ian Lynagh's avatar
Ian Lynagh committed
988
sym con span buf len =
989
  case lookupUFM reservedSymsFM fs of
Ian Lynagh's avatar
Ian Lynagh committed
990 991 992 993 994
        Just (keyword,exts) -> do
                b <- extension exts
                if b then return (L span keyword)
                     else return (L span $! con fs)
        _other -> return (L span $! con fs)
995
  where
Ian Lynagh's avatar
Ian Lynagh committed
996
        fs = lexemeToFastString buf len
997

998 999
-- Variations on the integral numeric literal.
tok_integral :: (Integer -> Token)
1000 1001 1002 1003 1004 1005 1006
             -> (Integer -> Integer)
             -> 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
1007 1008

-- some conveniences for use with tok_integral
Ian Lynagh's avatar
Ian Lynagh committed
1009 1010 1011
tok_num :: (Integer -> Integer)
        -> Int -> Int
        -> (Integer, (Char->Int)) -> Action
1012
tok_num = tok_integral ITinteger
Ian Lynagh's avatar
Ian Lynagh committed
1013 1014 1015
tok_primint :: (Integer -> Integer)
            -> Int -> Int
            -> (Integer, (Char->Int)) -> Action
1016
tok_primint = tok_integral ITprimint
Ian Lynagh's avatar
Ian Lynagh committed
1017 1018
tok_primword :: Int -> Int
             -> (Integer, (Char->Int)) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
1019
tok_primword = tok_integral ITprimword positive
Ian Lynagh's avatar
Ian Lynagh committed
1020
positive, negative :: (Integer -> Integer)
1021 1022
positive = id
negative = negate
Ian Lynagh's avatar
Ian Lynagh committed
1023
decimal, octal, hexadecimal :: (Integer, Char -> Int)
1024 1025 1026 1027 1028
decimal = (10,octDecDigit)
octal = (8,octDecDigit)
hexadecimal = (16,hexDigit)

-- readRational can understand negative rationals, exponents, everything.
Ian Lynagh's avatar
Ian Lynagh committed
1029
tok_float, tok_primfloat, tok_primdouble :: String -> Token
1030 1031 1032 1033 1034 1035
tok_float        str = ITrational   $! readFractionalLit str
tok_primfloat    str = ITprimfloat  $! readFractionalLit str
tok_primdouble   str = ITprimdouble $! readFractionalLit str

readFractionalLit :: String -> FractionalLit
readFractionalLit str = (FL $! str) $! readRational str
1036 1037 1038 1039 1040 1041

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

-- we're at the first token on a line, insert layout tokens if necessary
do_bol :: Action
1042
do_bol span _str _len = do
Ian Lynagh's avatar
Ian Lynagh committed
1043 1044 1045
        pos <- getOffside
        case pos of
            LT -> do
1046
                --trace "layout: inserting '}'" $ do