Lexer.x 92 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,
58
   traditionalRecordSyntaxEnabled,
59
   typeLiteralsEnabled,
ian@well-typed.com's avatar
ian@well-typed.com committed
60
   explicitNamespacesEnabled, sccProfilingOn, hpcEnabled,
Jedai's avatar
Jedai committed
61 62
   addWarning,
   lexTokenStream
63 64
  ) where

65 66
import Bag
import ErrUtils
67 68 69 70 71
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
72
import DynFlags
73
import Module
74
import Ctype
Ian Lynagh's avatar
Ian Lynagh committed
75 76
import BasicTypes       ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) )
import Util             ( readRational )
77

78
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
79
import Data.Bits
80
import Data.ByteString (ByteString)
81
import Data.Char
82
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
83
import Data.Maybe
84 85
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
86
import Data.Ratio
87
import Data.Word
88 89
}

90
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
91
$whitechar   = [\ \n\r\f\v $unispace]
92
$white_no_nl = $whitechar # \n
93
$tab         = \t
94 95

$ascdigit  = 0-9
96
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
97
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
98 99 100
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
101
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
102
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
103 104
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

105
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
106
$asclarge  = [A-Z]
107 108
$large     = [$asclarge $unilarge]

109
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
110
$ascsmall  = [a-z]
111 112
$small     = [$ascsmall $unismall \_]

113
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
114
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
115

Ian Lynagh's avatar
Ian Lynagh committed
116
$octit     = 0-7
117
$hexit     = [$decdigit A-F a-f]
118 119 120 121
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

122 123
$pragmachar = [$small $large $digit]

124 125
$docsym    = [\| \^ \* \$]

126 127 128 129 130 131
@varid     = $small $idchar*
@conid     = $large $idchar*

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

132
@decimal     = $decdigit+
133 134 135 136 137 138 139 140 141
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

142 143 144 145 146
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

147 148 149
haskell :-

-- everywhere: skip whitespace and comments
Ian Lynagh's avatar
Ian Lynagh committed
150
$white_no_nl+ ;
151
$tab+         { warn Opt_WarnTabs (text "Tab character") }
152 153 154 155 156

-- 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
157 158 159 160
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
161 162 163 164

-- 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.
165 166

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
167
-- out here.
168 169 170 171 172 173 174

-- 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
175 176
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
177 178 179

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

Jedai's avatar
Jedai committed
180
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
181 182 183 184 185 186 187

-- 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
188
"---"\-* [^$symbol :] .* { lineCommentToken }
189 190 191 192

-- 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
193
"--"\-* / { atEOL } { lineCommentToken }
194 195 196 197

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

Jedai's avatar
Jedai committed
198
"-- " / { atEOL } { lineCommentToken }
199 200 201 202 203 204 205 206 207 208 209

-- '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
210 211 212 213 214
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
215 216 217 218 219 220
}

-- 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
221 222 223 224
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
225 226 227
}

-- do is treated in a subtly different way, see new_layout_context
Ian Lynagh's avatar
Ian Lynagh committed
228 229
<layout>    ()                          { new_layout_context True }
<layout_do> ()                          { new_layout_context False }
230 231 232 233

-- 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
234
<layout_left>  ()                       { do_layout_left }
235

Ian Lynagh's avatar
Ian Lynagh committed
236
<0,option_prags> \n                     { begin bol }
237

238 239
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
240 241 242

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
Ian Lynagh's avatar
Ian Lynagh committed
243 244 245
<line_prag1> $decdigit+                 { setLine line_prag1a }
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
246 247 248

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
Ian Lynagh's avatar
Ian Lynagh committed
249 250 251
<line_prag2> $decdigit+                 { setLine line_prag2a }
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
252 253
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
254

255
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
256
  "{-#" $whitechar* $pragmachar+
257 258 259 260 261
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
262

263
  -- We ignore all these pragmas, but don't generate a warning for them
264 265
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
266 267

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
268
  "#-}"                          { endPrag }
269 270
}

271
<option_prags> {
272 273 274
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

Ian Lynagh's avatar
Ian Lynagh committed
275
  "-- #"                           { multiline_doc_comment }
David Waern's avatar
David Waern committed
276 277
}

278 279
<0> {
  -- In the "0" mode we ignore these pragmas
280
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
281 282 283
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
284
<0> {
Jedai's avatar
Jedai committed
285
  "-- #" .* { lineCommentToken }
286
}
287

288
<0,option_prags> {
289 290
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
291 292
}

293 294
-- '0' state: ordinary lexemes

295 296
-- Haddock comments

297
<0,option_prags> {
David Waern's avatar
David Waern committed
298 299
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
300 301
}

302 303
-- "special" symbols

304
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
305 306
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
307
}
Ian Lynagh's avatar
Ian Lynagh committed
308

309
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
310 311 312 313 314 315 316 317
  "[|"      / { 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 }
318

319 320 321 322
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

323
  "[" @varid "|"  / { ifExtension qqEnabled }
324
                     { lex_quasiquote_tok }
325 326 327 328

  -- qualified quasi-quote (#5555)
  "[" @qual @varid "|"  / { ifExtension qqEnabled }
                          { lex_qquasiquote_tok }
329 330
}

331
<0> {
ross's avatar
ross committed
332
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
Ian Lynagh's avatar
Ian Lynagh committed
333
                                        { special IToparenbar }
334 335 336
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

337
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
338
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
339 340
}

341
<0> {
342
  "(#" / { ifExtension unboxedTuplesEnabled }
Ian Lynagh's avatar
Ian Lynagh committed
343 344 345 346 347
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

348
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
349 350 351 352 353 354 355 356 357 358
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
359 360
}

361
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
362 363 364
  @qual @varid                  { idtoken qvarid }
  @qual @conid                  { idtoken qconid }
  @varid                        { varid }
365
  @conid                        { conid }
366 367
}

368
<0> {
369 370 371
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
372
  @conid "#"+       / { ifExtension magicHashEnabled } { conid }
373 374
}

375 376
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
377
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
378 379
  @qual @varsym                                    { idtoken qvarsym }
  @qual @consym                                    { idtoken qconsym }
380 381
  @varsym                                          { varsym }
  @consym                                          { consym }
382 383
}

384 385
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
386
<0> {
387
  -- Normal integral literals (:: Num a => a, from Integer)
388 389 390 391 392 393
  @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 }
  @negative @decimal           / { ifExtension negativeLiteralsEnabled } { tok_num negative 1 1 decimal }
  @negative 0[oO] @octal       / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 octal }
  @negative 0[xX] @hexadecimal / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 hexadecimal }
394 395

  -- Normal rational literals (:: Fractional a => a, from Rational)
396 397
  @floating_point                                                        { strtoken tok_float }
  @negative @floating_point    / { ifExtension negativeLiteralsEnabled } { strtoken tok_float }
398 399
}

400
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
401
  -- Unboxed ints (:: Int#) and words (:: Word#)
402 403
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
404 405 406 407 408
  @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 }
409
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
410

Ian Lynagh's avatar
Ian Lynagh committed
411 412 413 414
  @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 }

415 416
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
417 418
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
419 420 421 422 423 424
}

-- 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.
425
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
426 427
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
428 429 430 431 432 433
}

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

434
data Token
Ian Lynagh's avatar
Ian Lynagh committed
435
  = ITas                        -- Haskell keywords
436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458
  | 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
459
  | ITscc                       -- ToDo: remove (we use {-# SCC "..." #-} now)
460

Ian Lynagh's avatar
Ian Lynagh committed
461
  | ITforall                    -- GHC extension keywords
462 463 464 465 466
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
467
  | ITinterruptible
468 469 470
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
471
  | ITcapiconv
472
  | ITprimcallconv
473
  | ITmdo
474
  | ITfamily
475 476 477
  | ITgroup
  | ITby
  | ITusing
478 479 480
  | ITnominal
  | ITrepresentational
  | ITphantom
481

Ian Lynagh's avatar
Ian Lynagh committed
482
  -- Pragmas
483
  | ITinline_prag InlineSpec RuleMatchInfo
Ian Lynagh's avatar
Ian Lynagh committed
484 485
  | ITspec_prag                 -- SPECIALISE
  | ITspec_inline_prag Bool     -- SPECIALISE INLINE (or NOINLINE)
486 487
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
488
  | ITwarning_prag
489 490 491
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
492
  | ITgenerated_prag
493
  | ITcore_prag                 -- hdaume: core annotations
494
  | ITunpack_prag
495
  | ITnounpack_prag
496
  | ITann_prag
497
  | ITclose_prag
498 499 500
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
501 502
  | ITvect_prag
  | ITvect_scalar_prag
503
  | ITnovect_prag
504
  | ITctype
505

Ian Lynagh's avatar
Ian Lynagh committed
506
  | ITdotdot                    -- reserved symbols
507 508 509 510
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
511
  | ITlcase
512 513 514 515 516
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
batterseapower's avatar
batterseapower committed
517
  | ITtildehsh
518 519 520 521 522 523
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
524
  | ITbiglam                    -- GHC-extension symbols
525

Ian Lynagh's avatar
Ian Lynagh committed
526
  | ITocurly                    -- special symbols
527 528 529 530
  | ITccurly
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
531 532
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
533 534 535 536 537 538 539 540 541
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote
dreixel's avatar
dreixel committed
542
  | ITsimpleQuote               --  '
543

Ian Lynagh's avatar
Ian Lynagh committed
544
  | ITvarid   FastString        -- identifiers
545 546 547 548 549 550 551
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)
552 553
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
554

Ian Lynagh's avatar
Ian Lynagh committed
555
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
556 557 558 559

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
560
  | ITrational   FractionalLit
561 562

  | ITprimchar   Char
563
  | ITprimstring ByteString
564
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
565
  | ITprimword   Integer
566 567
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
568

569
  -- Template Haskell extension tokens
Ian Lynagh's avatar
Ian Lynagh committed
570 571 572 573 574 575 576 577
  | ITopenExpQuote              --  [| or [e|
  | ITopenPatQuote              --  [p|
  | ITopenDecQuote              --  [d|
  | ITopenTypQuote              --  [t|
  | ITcloseQuote                --  |]
  | ITidEscape   FastString     --  $x
  | ITparenEscape               --  $(
  | ITtyQuote                   --  ''
578 579 580 581 582 583 584 585
  | ITquasiQuote (FastString,FastString,RealSrcSpan)
    -- ITquasiQuote(quoter, quote, loc)
    -- represents a quasi-quote of the form
    -- [quoter| quote |]
  | ITqQuasiQuote (FastString,FastString,FastString,RealSrcSpan)
    -- ITqQuasiQuote(Qual, quoter, quote, loc)
    -- represents a qualified quasi-quote of the form
    -- [Qual.quoter| quote |]
586 587 588 589

  -- Arrow notation extension
  | ITproc
  | ITrec
Ian Lynagh's avatar
Ian Lynagh committed
590 591 592 593 594 595
  | IToparenbar                 --  (|
  | ITcparenbar                 --  |)
  | ITlarrowtail                --  -<
  | ITrarrowtail                --  >-
  | ITLarrowtail                --  -<<
  | ITRarrowtail                --  >>-
596

Ian Lynagh's avatar
Ian Lynagh committed
597 598
  | ITunknown String            -- Used when the lexer can't make sense of it
  | ITeof                       -- end of file token
599 600 601 602 603 604 605

  -- 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
606
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
607 608
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
609

610
  deriving Show
611 612 613 614 615 616 617 618

-- 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
619
reservedWordsFM :: UniqFM (Token, Int)
620
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666
    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),
667
         ( "capi",           ITcapiconv,      bit cApiFfiBit),
Ian Lynagh's avatar
Ian Lynagh committed
668 669
         ( "prim",           ITprimcallconv,  bit ffiBit),

670
         ( "rec",            ITrec,           bit arrowsBit .|.
671
                                              bit recursiveDoBit),
Ian Lynagh's avatar
Ian Lynagh committed
672
         ( "proc",           ITproc,          bit arrowsBit)
673 674
     ]

675 676 677 678 679 680 681 682
reservedUpcaseWordsFM :: UniqFM (Token, Int)
reservedUpcaseWordsFM = listToUFM $
    map (\(x, y, z) -> (mkFastString x, (y, z)))
       [ ( "N",     ITnominal,          0 ), -- no extension bit for better error msgs
         ( "R",     ITrepresentational, 0 ),
         ( "P",     ITphantom,          0 )
       ]

Ian Lynagh's avatar
Ian Lynagh committed
683
reservedSymsFM :: UniqFM (Token, Int -> Bool)
684
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
685 686 687 688 689 690 691 692 693 694 695 696
    map (\ (x,y,z) -> (mkFastString x,(y,z)))
      [ ("..",  ITdotdot,   always)
        -- (:) is a reserved op, meaning only list cons
       ,(":",   ITcolon,    always)
       ,("::",  ITdcolon,   always)
       ,("=",   ITequal,    always)
       ,("\\",  ITlam,      always)
       ,("|",   ITvbar,     always)
       ,("<-",  ITlarrow,   always)
       ,("->",  ITrarrow,   always)
       ,("@",   ITat,       always)
       ,("~",   ITtilde,    always)
697
       ,("~#",  ITtildehsh, magicHashEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
698 699 700 701 702
       ,("=>",  ITdarrow,   always)
       ,("-",   ITminus,    always)
       ,("!",   ITbang,     always)

        -- For data T (a::*) = MkT
703
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
704
        -- For 'forall a . t'
705
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
706 707 708 709 710

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

Ian Lynagh's avatar
Ian Lynagh committed
712 713
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
714 715
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
716 717
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
718 719 720 721 722 723 724 725

       ,("⤙",   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
726 727 728
        -- 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).
729 730 731 732 733
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
734
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
735

736
special :: Token -> Action
twanvl's avatar
twanvl committed
737
special tok span _buf _len = return (L span tok)
738

739
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
740 741
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
742

743 744
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
745

746
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
747
skip_one_varid f span buf len
748
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
749

750
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
751
strtoken f span buf len =
752
  return (L span $! (f $! lexemeToString buf len))
753

754
init_strtoken :: Int -> (String -> Token) -> Action
755
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
756
init_strtoken drop f span buf len =
757
  return (L span $! (f $! lexemeToString buf (len-drop)))
758 759

begin :: Int -> Action
760
begin code _span _str _len = do pushLexState code; lexToken
761 762

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
763 764
pop _span _buf _len = do _ <- popLexState
                         lexToken
765

766 767 768 769 770 771 772 773 774 775 776
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
777
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
778

779
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
780 781
pop_and act span buf len = do _ <- popLexState
                              act span buf len
782

783
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
784
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
785 786
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

787 788 789 790
{-# INLINE nextCharIsNot #-}
nextCharIsNot :: StringBuffer -> (Char -> Bool) -> Bool
nextCharIsNot buf p = not (nextCharIs buf p)

Ian Lynagh's avatar
Ian Lynagh committed
791
notFollowedBy :: Char -> AlexAccPred Int
Ian Lynagh's avatar
Ian Lynagh committed
792
notFollowedBy char _ _ _ (AI _ buf)
793
  = nextCharIsNot buf (== char)
794

Ian Lynagh's avatar
Ian Lynagh committed
795
notFollowedBySymbol :: AlexAccPred Int
796
notFollowedBySymbol _ _ _ (AI _ buf)
797
  = nextCharIsNot buf (`elem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
798

799 800 801 802 803
-- 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
804
isNormalComment :: AlexAccPred Int
805
isNormalComment bits _ _ (AI _ buf)
806
  | haddockEnabled bits = notFollowedByDocOrPragma
807
  | otherwise           = nextCharIsNot buf (== '#')
808 809
  where
    notFollowedByDocOrPragma
810
       = afterOptionalSpace buf (\b -> nextCharIsNot b (`elem` "|^*$#"))
811

812 813 814 815 816
afterOptionalSpace :: StringBuffer -> (StringBuffer -> Bool) -> Bool
afterOptionalSpace buf p
    = if nextCharIs buf (== ' ')
      then p (snd (nextChar buf))
      else p buf
817

Ian Lynagh's avatar
Ian Lynagh committed
818
atEOL :: AlexAccPred Int
819
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
820

Ian Lynagh's avatar
Ian Lynagh committed
821
ifExtension :: (Int -> Bool) -> AlexAccPred Int
822 823
ifExtension pred bits _ _ _ = pred bits

824 825 826
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
827
    worker commentAcc input docType oneLine = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
828
      Just ('\n', input')
829 830 831 832 833 834
        | 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
835

836 837
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
838 839 840
        check input = case alexGetChar' input of
          Just ('-', input) -> case alexGetChar' input of
            Just ('-', input) -> case alexGetChar' input of
841 842 843 844 845
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

846
        dropNonNewlineSpace input = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
847
          Just (c, input')
848 849 850 851
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
852 853 854 855 856
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

857 858 859 860
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
861
nested_comment :: P (RealLocated Token) -> Action
862
nested_comment cont span _str _len = do
863
  input <- getInput
Jedai's avatar
Jedai committed
864
  go "" (1::Int) input
865
  where
Jedai's avatar
Jedai committed
866 867 868 869 870
    go commentAcc 0 input = do setInput input
                               b <- extension rawTokenStreamEnabled
                               if b
                                 then docCommentEnd input commentAcc ITblockComment _str span
                                 else cont
871
    go commentAcc n input = case alexGetChar' input of
872
      Nothing -> errBrace input span
873
      Just ('-',input) -> case alexGetChar' input of
874
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
875 876
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
877
      Just ('\123',input) -> case alexGetChar' input of
878
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
879 880 881
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
882 883 884 885

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
886
    go commentAcc input docType _ = case alexGetChar' input of
887
      Nothing -> errBrace input span
888
      Just ('-',input) -> case alexGetChar' input of
889
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
890
        Just ('\125',input) ->
891
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
892
        Just (_,_) -> go ('-':commentAcc) input docType False
893
      Just ('\123', input) -> case alexGetChar' input of
894 895 896 897 898
        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
899
        Just (_,_) -> go ('\123':commentAcc) input docType False
900 901
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
902 903
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
904
withLexedDocType lexDocComment = do
905
  input@(AI _ buf) <- getInput
906 907 908 909
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
910 911
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
912
    _ -> panic "withLexedDocType: Bad doc type"
Ian Lynagh's avatar
Ian Lynagh committed
913
 where
914
    lexDocSection n input = case alexGetChar' input of
915
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
916
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
917 918
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

919 920 921
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
922
rulePrag span _buf _len = do
923
  setExts (.|. bit inRulePragBit)
924 925 926
  return (L span ITrules_prag)

endPrag :: Action
927
endPrag span _buf _len = do
928 929 930
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

931 932 933 934
-- docCommentEnd
-------------------------------------------------------------------------------
-- This function is quite tricky. We can't just return a new token, we also
-- need to update the state of the parser. Why? Because the token is longer
Ian Lynagh's avatar
Ian Lynagh committed
935
-- than what was lexed by Alex, and the lexToken function doesn't know this, so
936
-- it writes the wrong token length to the parser state. This function is
Ian Lynagh's avatar
Ian Lynagh committed
937
-- called afterwards, so it can just update the state.
938 939

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
940
                 RealSrcSpan -> P (RealLocated Token)
941 942
docCommentEnd input commentAcc docType buf span = do
  setInput input
943
  let (AI loc nextBuf) = input
944
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
945
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
946
      last_len = byteDiff buf nextBuf
Ian Lynagh's avatar
Ian Lynagh committed
947

948
  span `seq` setLastToken span' last_len
949
  return (L span' (docType comment))
Ian Lynagh's avatar
Ian Lynagh committed
950

Ian Lynagh's avatar
Ian Lynagh committed
951 952
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
953

954
open_brace, close_brace :: Action
Ian Lynagh's avatar
Ian Lynagh committed
955
open_brace span _str _len = do
956 957
  ctx <- getContext
  setContext (NoLayout:ctx)
958
  return (L span ITocurly)
Ian Lynagh's avatar
Ian Lynagh committed
959
close_brace span _str _len = do
960
  popContext
961
  return (L span ITccurly)
962

Ian Lynagh's avatar
Ian Lynagh committed
963
qvarid, qconid :: StringBuffer -> Int -> Token
964 965
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
966

967
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
968 969 970
-- 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.
971
splitQualName orig_buf len parens = split orig_buf orig_buf
972
  where
973
    split buf dot_buf
Ian Lynagh's avatar
Ian Lynagh committed
974 975 976
        | orig_buf `byteDiff` buf >= len  = done dot_buf
        | c == '.'                        = found_dot buf'
        | otherwise                       = split buf' dot_buf
977 978
      where
       (c,buf') = nextChar buf
Ian Lynagh's avatar
Ian Lynagh committed
979

980 981 982
    -- 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.
983
    found_dot buf -- buf points after the '.'
Ian Lynagh's avatar
Ian Lynagh committed
984 985
        | isUpper c    = split buf' buf
        | otherwise    = done buf
986 987 988 989
      where
       (c,buf') = nextChar buf

    done dot_buf =
Ian Lynagh's avatar
Ian Lynagh committed
990 991
        (lexemeToFastString orig_buf (qual_size - 1),
         if parens -- Prelude.(+)
992 993
            then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
            else lexemeToFastString dot_buf (len - qual_size))
994
      where
Ian Lynagh's avatar
Ian Lynagh committed
995
        qual_size = orig_buf `byteDiff` dot_buf
996

Ian Lynagh's avatar
Ian Lynagh committed
997 998
varid :: Action
varid span buf len =
999
  case lookupUFM reservedWordsFM fs of