Lexer.x 92.2 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,
58
   traditionalRecordSyntaxEnabled,
59
   typeLiteralsEnabled,
60 61
   explicitForallEnabled,
   inRulePrag,
ian@well-typed.com's avatar
ian@well-typed.com committed
62
   explicitNamespacesEnabled, sccProfilingOn, hpcEnabled,
Jedai's avatar
Jedai committed
63 64
   addWarning,
   lexTokenStream
65 66
  ) where

67 68
import Bag
import ErrUtils
69 70 71 72 73
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
74
import DynFlags
75
import Module
76
import Ctype
Ian Lynagh's avatar
Ian Lynagh committed
77 78
import BasicTypes       ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) )
import Util             ( readRational )
79

80
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
81
import Data.Bits
82
import Data.ByteString (ByteString)
83
import Data.Char
84
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
85
import Data.Maybe
86 87
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
88
import Data.Ratio
89
import Data.Word
90 91
}

92
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
93
$whitechar   = [\ \n\r\f\v $unispace]
94
$white_no_nl = $whitechar # \n
95
$tab         = \t
96 97

$ascdigit  = 0-9
98
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
99
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
100 101 102
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
103
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
104
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
105 106
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

107
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
108
$asclarge  = [A-Z]
109 110
$large     = [$asclarge $unilarge]

111
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
112
$ascsmall  = [a-z]
113 114
$small     = [$ascsmall $unismall \_]

115
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
116
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
117

Ian Lynagh's avatar
Ian Lynagh committed
118
$octit     = 0-7
119
$hexit     = [$decdigit A-F a-f]
120 121 122 123
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

124 125
$pragmachar = [$small $large $digit]

126 127
$docsym    = [\| \^ \* \$]

128 129 130 131 132 133
@varid     = $small $idchar*
@conid     = $large $idchar*

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

134
@decimal     = $decdigit+
135 136 137 138 139 140 141 142 143
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

144 145 146 147 148
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

149 150 151
haskell :-

-- everywhere: skip whitespace and comments
Ian Lynagh's avatar
Ian Lynagh committed
152
$white_no_nl+ ;
153
$tab+         { warn Opt_WarnTabs (text "Tab character") }
154 155 156 157 158

-- 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
159 160 161 162
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
163 164 165 166

-- 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.
167 168

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
169
-- out here.
170 171 172 173 174 175 176

-- 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
177 178
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
179 180 181

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

Jedai's avatar
Jedai committed
182
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
183 184 185 186 187 188 189

-- 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
190
"---"\-* [^$symbol :] .* { lineCommentToken }
191 192 193 194

-- 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
195
"--"\-* / { atEOL } { lineCommentToken }
196 197 198 199

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

Jedai's avatar
Jedai committed
200
"-- " / { atEOL } { lineCommentToken }
201 202 203 204 205 206 207 208 209 210 211

-- '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
212 213 214 215 216
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
217 218 219 220 221 222
}

-- 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
223 224 225 226
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
227 228 229
}

-- do is treated in a subtly different way, see new_layout_context
Ian Lynagh's avatar
Ian Lynagh committed
230 231
<layout>    ()                          { new_layout_context True }
<layout_do> ()                          { new_layout_context False }
232 233 234 235

-- 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
236
<layout_left>  ()                       { do_layout_left }
237

Ian Lynagh's avatar
Ian Lynagh committed
238
<0,option_prags> \n                     { begin bol }
239

240 241
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
242 243 244

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
Ian Lynagh's avatar
Ian Lynagh committed
245 246 247
<line_prag1> $decdigit+                 { setLine line_prag1a }
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
248 249 250

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
Ian Lynagh's avatar
Ian Lynagh committed
251 252 253
<line_prag2> $decdigit+                 { setLine line_prag2a }
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
254 255
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
256

257
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
258
  "{-#" $whitechar* $pragmachar+
259 260 261 262 263
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
264

265
  -- We ignore all these pragmas, but don't generate a warning for them
266 267
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
268 269

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
270
  "#-}"                          { endPrag }
271 272
}

273
<option_prags> {
274 275 276
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

Ian Lynagh's avatar
Ian Lynagh committed
277
  "-- #"                           { multiline_doc_comment }
David Waern's avatar
David Waern committed
278 279
}

280 281
<0> {
  -- In the "0" mode we ignore these pragmas
282
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
283 284 285
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
286
<0> {
Jedai's avatar
Jedai committed
287
  "-- #" .* { lineCommentToken }
288
}
289

290
<0,option_prags> {
291 292
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
293 294
}

295 296
-- '0' state: ordinary lexemes

297 298
-- Haddock comments

299
<0,option_prags> {
David Waern's avatar
David Waern committed
300 301
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
302 303
}

304 305
-- "special" symbols

306
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
307 308
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
309
}
Ian Lynagh's avatar
Ian Lynagh committed
310

311
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
312 313 314 315 316 317 318 319
  "[|"      / { 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 }
320

321 322 323 324
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

325
  "[" @varid "|"  / { ifExtension qqEnabled }
326
                     { lex_quasiquote_tok }
327 328 329 330

  -- qualified quasi-quote (#5555)
  "[" @qual @varid "|"  / { ifExtension qqEnabled }
                          { lex_qquasiquote_tok }
331 332
}

333
<0> {
ross's avatar
ross committed
334
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
Ian Lynagh's avatar
Ian Lynagh committed
335
                                        { special IToparenbar }
336 337 338
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

339
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
340
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
341 342
}

343
<0> {
344
  "(#" / { ifExtension unboxedTuplesEnabled }
Ian Lynagh's avatar
Ian Lynagh committed
345 346 347 348 349
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

350
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
351 352 353 354 355 356 357 358 359 360
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
361 362
}

363
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
364 365 366
  @qual @varid                  { idtoken qvarid }
  @qual @conid                  { idtoken qconid }
  @varid                        { varid }
367
  @conid                        { idtoken conid }
368 369
}

370
<0> {
371 372 373
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
374
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
375 376
}

377 378
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
379
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
380 381
  @qual @varsym                                    { idtoken qvarsym }
  @qual @consym                                    { idtoken qconsym }
382 383
  @varsym                                          { varsym }
  @consym                                          { consym }
384 385
}

386 387
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
388
<0> {
389
  -- Normal integral literals (:: Num a => a, from Integer)
390 391 392 393 394 395
  @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 }
396 397

  -- Normal rational literals (:: Fractional a => a, from Rational)
398 399
  @floating_point                                                        { strtoken tok_float }
  @negative @floating_point    / { ifExtension negativeLiteralsEnabled } { strtoken tok_float }
400 401
}

402
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
403
  -- Unboxed ints (:: Int#) and words (:: Word#)
404 405
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
406 407 408 409 410
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
  0[oO] @octal                 \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
  0[xX] @hexadecimal           \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
  @negative @decimal           \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
411
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
412

Ian Lynagh's avatar
Ian Lynagh committed
413 414 415 416
  @decimal                     \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
  0[oO] @octal                 \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
  0[xX] @hexadecimal           \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }

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

-- Strings and chars are lexed by hand-written code.  The reason is
-- that even if we recognise the string or char here in the regex
-- lexer, we would still have to parse the string afterward in order
-- to convert it to a String.
427
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
428 429
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
430 431 432 433 434 435
}

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

436
data Token
Ian Lynagh's avatar
Ian Lynagh committed
437
  = ITas                        -- Haskell keywords
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460
  | 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
461
  | ITscc                       -- ToDo: remove (we use {-# SCC "..." #-} now)
462

Ian Lynagh's avatar
Ian Lynagh committed
463
  | ITforall                    -- GHC extension keywords
464 465 466 467 468
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
469
  | ITinterruptible
470 471 472
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
473
  | ITcapiconv
474
  | ITprimcallconv
thoughtpolice's avatar
thoughtpolice committed
475
  | ITjavascriptcallconv
476
  | ITmdo
477
  | ITfamily
478
  | ITrole
479 480 481
  | ITgroup
  | ITby
  | ITusing
482

Ian Lynagh's avatar
Ian Lynagh committed
483
  -- Pragmas
484
  | ITinline_prag InlineSpec RuleMatchInfo
Ian Lynagh's avatar
Ian Lynagh committed
485 486
  | ITspec_prag                 -- SPECIALISE
  | ITspec_inline_prag Bool     -- SPECIALISE INLINE (or NOINLINE)
487 488
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
489
  | ITwarning_prag
490 491 492
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
493
  | ITgenerated_prag
494
  | ITcore_prag                 -- hdaume: core annotations
495
  | ITunpack_prag
496
  | ITnounpack_prag
497
  | ITann_prag
498
  | ITclose_prag
499 500 501
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
502 503
  | ITvect_prag
  | ITvect_scalar_prag
504
  | ITnovect_prag
505
  | ITminimal_prag
506
  | ITctype
507

Ian Lynagh's avatar
Ian Lynagh committed
508
  | ITdotdot                    -- reserved symbols
509 510 511 512
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
513
  | ITlcase
514 515 516 517 518
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
batterseapower's avatar
batterseapower committed
519
  | ITtildehsh
520 521 522 523 524 525
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
526
  | ITbiglam                    -- GHC-extension symbols
527

Ian Lynagh's avatar
Ian Lynagh committed
528
  | ITocurly                    -- special symbols
529 530 531 532
  | ITccurly
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
533 534
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
535 536 537 538 539 540 541 542 543
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote
dreixel's avatar
dreixel committed
544
  | ITsimpleQuote               --  '
545

Ian Lynagh's avatar
Ian Lynagh committed
546
  | ITvarid   FastString        -- identifiers
547 548 549 550 551 552 553
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)
554 555
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
556

Ian Lynagh's avatar
Ian Lynagh committed
557
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
558 559 560 561

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
562
  | ITrational   FractionalLit
563 564

  | ITprimchar   Char
565
  | ITprimstring ByteString
566
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
567
  | ITprimword   Integer
568 569
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
570

571
  -- Template Haskell extension tokens
Ian Lynagh's avatar
Ian Lynagh committed
572 573 574 575 576 577 578 579
  | ITopenExpQuote              --  [| or [e|
  | ITopenPatQuote              --  [p|
  | ITopenDecQuote              --  [d|
  | ITopenTypQuote              --  [t|
  | ITcloseQuote                --  |]
  | ITidEscape   FastString     --  $x
  | ITparenEscape               --  $(
  | ITtyQuote                   --  ''
580 581 582 583 584 585 586 587
  | 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 |]
588 589 590 591

  -- Arrow notation extension
  | ITproc
  | ITrec
Ian Lynagh's avatar
Ian Lynagh committed
592 593 594 595 596 597
  | IToparenbar                 --  (|
  | ITcparenbar                 --  |)
  | ITlarrowtail                --  -<
  | ITrarrowtail                --  >-
  | ITLarrowtail                --  -<<
  | ITRarrowtail                --  >>-
598

Ian Lynagh's avatar
Ian Lynagh committed
599 600
  | ITunknown String            -- Used when the lexer can't make sense of it
  | ITeof                       -- end of file token
601 602 603 604 605 606 607

  -- 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
608
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
609 610
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
611

612
  deriving Show
613 614 615 616 617 618 619 620

-- 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
621
reservedWordsFM :: UniqFM (Token, Int)
622
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
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
    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),
654 655 656
             -- See Note [Lexing type pseudo-keywords]
         ( "family",         ITfamily,        0 ),
         ( "role",           ITrole,          0 ),
Ian Lynagh's avatar
Ian Lynagh committed
657 658 659 660 661 662 663 664 665 666 667 668 669 670
         ( "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),
671
         ( "capi",           ITcapiconv,      bit cApiFfiBit),
Ian Lynagh's avatar
Ian Lynagh committed
672
         ( "prim",           ITprimcallconv,  bit ffiBit),
thoughtpolice's avatar
thoughtpolice committed
673
         ( "javascript",     ITjavascriptcallconv, bit ffiBit),
Ian Lynagh's avatar
Ian Lynagh committed
674

675
         ( "rec",            ITrec,           bit arrowsBit .|.
676
                                              bit recursiveDoBit),
Ian Lynagh's avatar
Ian Lynagh committed
677
         ( "proc",           ITproc,          bit arrowsBit)
678 679
     ]

680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695
{-----------------------------------
Note [Lexing type pseudo-keywords]
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

One might think that we wish to treat 'family' and 'role' as regular old
varids whenever -XTypeFamilies and -XRoleAnnotations are off, respectively.
But, there is no need to do so. These pseudo-keywords are not stolen syntax:
they are only used after the keyword 'type' at the top-level, where varids are
not allowed. Furthermore, checks further downstream (TcTyClsDecls) ensure that
type families and role annotations are never declared without their extensions
on. In fact, by unconditionally lexing these pseudo-keywords as special, we
can get better error messages.

Also, note that these are included in the `varid` production in the parser --
a key detail to make all this work.
-------------------------------------}
696

Ian Lynagh's avatar
Ian Lynagh committed
697
reservedSymsFM :: UniqFM (Token, Int -> Bool)
698
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
699 700 701 702 703 704 705 706 707 708 709 710
    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)
711
       ,("~#",  ITtildehsh, magicHashEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
712 713 714 715 716
       ,("=>",  ITdarrow,   always)
       ,("-",   ITminus,    always)
       ,("!",   ITbang,     always)

        -- For data T (a::*) = MkT
717
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
718
        -- For 'forall a . t'
719
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
720 721 722 723 724

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
725

Ian Lynagh's avatar
Ian Lynagh committed
726 727
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
728
       ,("∀",   ITforall, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
729 730
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
731 732 733 734 735 736 737 738

       ,("⤙",   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
739 740 741
        -- 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).
742 743 744 745 746
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
747
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
748

749
special :: Token -> Action
twanvl's avatar
twanvl committed
750
special tok span _buf _len = return (L span tok)
751

752
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
753 754
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
755

756 757
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
758

759
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
760
skip_one_varid f span buf len
761
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
762

763
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
764
strtoken f span buf len =
765
  return (L span $! (f $! lexemeToString buf len))
766

767
init_strtoken :: Int -> (String -> Token) -> Action
768
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
769
init_strtoken drop f span buf len =
770
  return (L span $! (f $! lexemeToString buf (len-drop)))
771 772

begin :: Int -> Action
773
begin code _span _str _len = do pushLexState code; lexToken
774 775

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
776 777
pop _span _buf _len = do _ <- popLexState
                         lexToken
778

779 780 781 782 783 784 785 786 787 788 789
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
790
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
791

792
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
793 794
pop_and act span buf len = do _ <- popLexState
                              act span buf len
795

796
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
797
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
798 799
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

800 801 802 803
{-# INLINE nextCharIsNot #-}
nextCharIsNot :: StringBuffer -> (Char -> Bool) -> Bool
nextCharIsNot buf p = not (nextCharIs buf p)

Ian Lynagh's avatar
Ian Lynagh committed
804
notFollowedBy :: Char -> AlexAccPred Int
Ian Lynagh's avatar
Ian Lynagh committed
805
notFollowedBy char _ _ _ (AI _ buf)
806
  = nextCharIsNot buf (== char)
807

Ian Lynagh's avatar
Ian Lynagh committed
808
notFollowedBySymbol :: AlexAccPred Int
809
notFollowedBySymbol _ _ _ (AI _ buf)
810
  = nextCharIsNot buf (`elem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
811

812 813 814 815 816
-- 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
817
isNormalComment :: AlexAccPred Int
818
isNormalComment bits _ _ (AI _ buf)
819
  | haddockEnabled bits = notFollowedByDocOrPragma
820
  | otherwise           = nextCharIsNot buf (== '#')
821 822
  where
    notFollowedByDocOrPragma
823
       = afterOptionalSpace buf (\b -> nextCharIsNot b (`elem` "|^*$#"))
824

825 826 827 828 829
afterOptionalSpace :: StringBuffer -> (StringBuffer -> Bool) -> Bool
afterOptionalSpace buf p
    = if nextCharIs buf (== ' ')
      then p (snd (nextChar buf))
      else p buf
830

Ian Lynagh's avatar
Ian Lynagh committed
831
atEOL :: AlexAccPred Int
832
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
833

Ian Lynagh's avatar
Ian Lynagh committed
834
ifExtension :: (Int -> Bool) -> AlexAccPred Int
835 836
ifExtension pred bits _ _ _ = pred bits

837 838 839
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
840
    worker commentAcc input docType oneLine = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
841
      Just ('\n', input')
842 843 844 845 846 847
        | 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
848

849 850
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
851 852 853
        check input = case alexGetChar' input of
          Just ('-', input) -> case alexGetChar' input of
            Just ('-', input) -> case alexGetChar' input of
854 855 856 857 858
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

859
        dropNonNewlineSpace input = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
860
          Just (c, input')
861 862 863 864
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
865 866 867 868 869
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

870 871 872 873
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
874
nested_comment :: P (RealLocated Token) -> Action
875
nested_comment cont span _str _len = do
876
  input <- getInput
Jedai's avatar
Jedai committed
877
  go "" (1::Int) input
878
  where
Jedai's avatar
Jedai committed
879 880 881 882 883
    go commentAcc 0 input = do setInput input
                               b <- extension rawTokenStreamEnabled
                               if b
                                 then docCommentEnd input commentAcc ITblockComment _str span
                                 else cont
884
    go commentAcc n input = case alexGetChar' input of
885
      Nothing -> errBrace input span
886
      Just ('-',input) -> case alexGetChar' input of
887
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
888 889
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
890
      Just ('\123',input) -> case alexGetChar' input of
891
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
892 893 894
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
895 896 897 898

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
899
    go commentAcc input docType _ = case alexGetChar' input of
900
      Nothing -> errBrace input span
901
      Just ('-',input) -> case alexGetChar' input of
902
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
903
        Just ('\125',input) ->
904
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
905
        Just (_,_) -> go ('-':commentAcc) input docType False
906
      Just ('\123', input) -> case alexGetChar' input of
907 908 909 910 911
        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
912
        Just (_,_) -> go ('\123':commentAcc) input docType False
913 914
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
915 916
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
917
withLexedDocType lexDocComment = do
918
  input@(AI _ buf) <- getInput
919 920 921 922
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
923 924
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
925
    _ -> panic "withLexedDocType: Bad doc type"
Ian Lynagh's avatar
Ian Lynagh committed
926
 where
927
    lexDocSection n input = case alexGetChar' input of
928
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
929
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
930 931
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

932 933 934
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
935
rulePrag span _buf _len = do
936
  setExts (.|. bit inRulePragBit)
937 938 939
  return (L span ITrules_prag)

endPrag :: Action
940
endPrag span _buf _len = do
941 942 943
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

944 945 946 947
-- 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
948
-- than what was lexed by Alex, and the lexToken function doesn't know this, so
949
-- it writes the wrong token length to the parser state. This function is
Ian Lynagh's avatar
Ian Lynagh committed
950
-- called afterwards, so it can just update the state.
951 952

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
953
                 RealSrcSpan -> P (RealLocated Token)
954 955
docCommentEnd input commentAcc docType buf span = do
  setInput input
956
  let (AI loc nextBuf) = input
957
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
958
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
959
      last_len = byteDiff buf nextBuf
Ian Lynagh's avatar
Ian Lynagh committed
960

961
  span `seq` setLastToken span' last_len
962
  return (L span' (docType comment))
Ian Lynagh's avatar
Ian Lynagh committed
963

Ian Lynagh's avatar
Ian Lynagh committed
964 965
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
966

967
open_brace, close_brace :: Action
Ian Lynagh's avatar
Ian Lynagh committed
968
open_brace span _str _len = do
969 970
  ctx <- getContext
  setContext (NoLayout:ctx)
971
  return (L span ITocurly)
Ian Lynagh's avatar
Ian Lynagh committed
972
close_brace span _str _len = do
973
  popContext
974
  return (L span ITccurly)
975

Ian Lynagh's avatar
Ian Lynagh committed
976
qvarid, qconid :: StringBuffer -> Int -> Token
977 978
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
979

980
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
981 982 983
-- 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.
984
splitQualName orig_buf len parens = split orig_buf orig_buf
985
  where
986
    split buf dot_buf
Ian Lynagh's avatar
Ian Lynagh committed
987 988 989
        | orig_buf `byteDiff` buf >= len  = done dot_buf
        | c == '.'                        = found_dot buf'
        | otherwise                       = split buf' dot_buf
990 991
      where
       (c,buf') = nextChar buf
Ian Lynagh's avatar
Ian Lynagh committed
992

993 994 995
    -- 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.
996
    found_dot buf -- buf points after the '.'
Ian Lynagh's avatar
Ian Lynagh committed
997 998
        | isUpper c    = split buf' buf
        | otherwise    = done buf