Lexer.x 87.7 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,
Jedai's avatar
Jedai committed
59 60
   addWarning,
   lexTokenStream
61 62
  ) where

63 64
import Bag
import ErrUtils
65 66 67 68 69
import Outputable
import StringBuffer
import FastString
import SrcLoc
import UniqFM
70
import DynFlags
71
import Module
72
import Ctype
Ian Lynagh's avatar
Ian Lynagh committed
73 74
import BasicTypes       ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) )
import Util             ( readRational )
75

76
import Control.Monad
Simon Marlow's avatar
Simon Marlow committed
77
import Data.Bits
78
import Data.Char
79
import Data.List
Ian Lynagh's avatar
Ian Lynagh committed
80
import Data.Maybe
81 82
import Data.Map (Map)
import qualified Data.Map as Map
Simon Marlow's avatar
Simon Marlow committed
83
import Data.Ratio
84
import Data.Word
85 86
}

87
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
88
$whitechar   = [\ \n\r\f\v $unispace]
89
$white_no_nl = $whitechar # \n
90
$tab         = \t
91 92

$ascdigit  = 0-9
93
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
94
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
95 96 97
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
98
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
99
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
100 101
$symbol    = [$ascsymbol $unisymbol] # [$special \_\:\"\']

102
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
103
$asclarge  = [A-Z]
104 105
$large     = [$asclarge $unilarge]

106
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
107
$ascsmall  = [a-z]
108 109
$small     = [$ascsmall $unismall \_]

110
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
111
$graphic   = [$small $large $symbol $digit $special $unigraphic \:\"\']
112

Ian Lynagh's avatar
Ian Lynagh committed
113
$octit     = 0-7
114
$hexit     = [$decdigit A-F a-f]
115 116 117 118
$symchar   = [$symbol \:]
$nl        = [\n\r]
$idchar    = [$small $large $digit \']

119 120
$pragmachar = [$small $large $digit]

121 122
$docsym    = [\| \^ \* \$]

123 124 125 126 127 128
@varid     = $small $idchar*
@conid     = $large $idchar*

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

129
@decimal     = $decdigit+
130 131 132 133 134 135 136 137 138
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

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

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

139 140 141 142 143
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

144 145 146
haskell :-

-- everywhere: skip whitespace and comments
Ian Lynagh's avatar
Ian Lynagh committed
147
$white_no_nl+ ;
148
$tab+         { warn Opt_WarnTabs (text "Warning: Tab character") }
149 150 151 152 153

-- 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
154 155 156 157
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
158 159 160 161

-- 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.
162 163

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
164
-- out here.
165 166 167 168 169 170 171

-- 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
172 173
"-- " ~[$docsym \#] .* { lineCommentToken }
"--" [^$symbol : \ ] .* { lineCommentToken }
174 175 176

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

Jedai's avatar
Jedai committed
177
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
178 179 180 181 182 183 184

-- 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
185
"---"\-* [^$symbol :] .* { lineCommentToken }
186 187 188 189

-- 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
190
"--"\-* / { atEOL } { lineCommentToken }
191 192 193 194

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

Jedai's avatar
Jedai committed
195
"-- " / { atEOL } { lineCommentToken }
196 197 198 199 200 201 202 203 204 205 206

-- '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
207 208 209 210 211
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
212 213 214 215 216 217
}

-- 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
218 219 220 221
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
222 223 224
}

-- do is treated in a subtly different way, see new_layout_context
Ian Lynagh's avatar
Ian Lynagh committed
225 226
<layout>    ()                          { new_layout_context True }
<layout_do> ()                          { new_layout_context False }
227 228 229 230

-- 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
231
<layout_left>  ()                       { do_layout_left }
232

Ian Lynagh's avatar
Ian Lynagh committed
233
<0,option_prags> \n                     { begin bol }
234

235 236
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
237 238 239

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
Ian Lynagh's avatar
Ian Lynagh committed
240 241 242
<line_prag1> $decdigit+                 { setLine line_prag1a }
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
243 244 245

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
Ian Lynagh's avatar
Ian Lynagh committed
246 247 248
<line_prag2> $decdigit+                 { setLine line_prag2a }
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
249 250
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
251

252
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
253
  "{-#" $whitechar* $pragmachar+
254 255 256 257 258
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
259

260
  -- We ignore all these pragmas, but don't generate a warning for them
261 262
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
263 264

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
265
  "#-}"                          { endPrag }
266 267
}

268
<option_prags> {
269 270 271
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

Ian Lynagh's avatar
Ian Lynagh committed
272
  "-- #"                           { multiline_doc_comment }
David Waern's avatar
David Waern committed
273 274
}

275 276
<0> {
  -- In the "0" mode we ignore these pragmas
277
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
278 279 280
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
281
<0> {
Jedai's avatar
Jedai committed
282
  "-- #" .* { lineCommentToken }
283
}
284

285
<0,option_prags> {
286 287
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
288 289
}

290 291
-- '0' state: ordinary lexemes

292 293
-- Haddock comments

294
<0,option_prags> {
David Waern's avatar
David Waern committed
295 296
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
297 298
}

299 300
-- "special" symbols

301
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
302 303
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
304
}
Ian Lynagh's avatar
Ian Lynagh committed
305

306
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
307 308 309 310 311 312 313 314
  "[|"      / { 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 }
315

316 317 318 319
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

320
  "[" @varid "|"  / { ifExtension qqEnabled }
321
                     { lex_quasiquote_tok }
322 323
}

324
<0> {
ross's avatar
ross committed
325
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
Ian Lynagh's avatar
Ian Lynagh committed
326
                                        { special IToparenbar }
327 328 329
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

330
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
331
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
332 333
}

334
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
335 336 337 338 339 340
  "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

341
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
342 343 344 345 346 347 348 349 350 351
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
352 353
}

354
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
355 356 357 358
  @qual @varid                  { idtoken qvarid }
  @qual @conid                  { idtoken qconid }
  @varid                        { varid }
  @conid                        { idtoken conid }
359 360
}

361
<0> {
362 363 364 365
  @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
366 367
}

368 369
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
370
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
371 372
  @qual @varsym                                    { idtoken qvarsym }
  @qual @consym                                    { idtoken qconsym }
373 374
  @varsym                                          { varsym }
  @consym                                          { consym }
375 376
}

377 378
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
379
<0> {
380
  -- Normal integral literals (:: Num a => a, from Integer)
Ian Lynagh's avatar
Ian Lynagh committed
381 382 383
  @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 }
384 385

  -- Normal rational literals (:: Fractional a => a, from Rational)
Ian Lynagh's avatar
Ian Lynagh committed
386
  @floating_point    { strtoken tok_float }
387 388
}

389
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
390
  -- Unboxed ints (:: Int#) and words (:: Word#)
391 392
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
393 394 395 396 397
  @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 }
398
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
399

Ian Lynagh's avatar
Ian Lynagh committed
400 401 402 403
  @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 }

404 405
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
406 407
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
408 409 410 411 412 413
}

-- 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.
414
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
415 416
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
417 418 419 420 421 422
}

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

423
data Token
Ian Lynagh's avatar
Ian Lynagh committed
424
  = ITas                        -- Haskell keywords
425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447
  | 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
448
  | ITscc                       -- ToDo: remove (we use {-# SCC "..." #-} now)
449

Ian Lynagh's avatar
Ian Lynagh committed
450
  | ITforall                    -- GHC extension keywords
451 452 453 454 455
  | ITforeign
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
456
  | ITinterruptible
457 458 459
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
460
  | ITcapiconv
461
  | ITprimcallconv
462
  | ITmdo
463
  | ITfamily
464 465 466
  | ITgroup
  | ITby
  | ITusing
467

Ian Lynagh's avatar
Ian Lynagh committed
468
  -- Pragmas
469
  | ITinline_prag InlineSpec RuleMatchInfo
Ian Lynagh's avatar
Ian Lynagh committed
470 471
  | ITspec_prag                 -- SPECIALISE
  | ITspec_inline_prag Bool     -- SPECIALISE INLINE (or NOINLINE)
472 473
  | ITsource_prag
  | ITrules_prag
Ian Lynagh's avatar
Ian Lynagh committed
474
  | ITwarning_prag
475 476 477
  | ITdeprecated_prag
  | ITline_prag
  | ITscc_prag
andy@galois.com's avatar
andy@galois.com committed
478
  | ITgenerated_prag
479
  | ITcore_prag                 -- hdaume: core annotations
480
  | ITunpack_prag
481
  | ITnounpack_prag
482
  | ITann_prag
483
  | ITclose_prag
484 485 486
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
487 488
  | ITvect_prag
  | ITvect_scalar_prag
489
  | ITnovect_prag
490

Ian Lynagh's avatar
Ian Lynagh committed
491
  | ITdotdot                    -- reserved symbols
492 493 494 495 496 497 498 499 500
  | ITcolon
  | ITdcolon
  | ITequal
  | ITlam
  | ITvbar
  | ITlarrow
  | ITrarrow
  | ITat
  | ITtilde
batterseapower's avatar
batterseapower committed
501
  | ITtildehsh
502 503 504 505 506 507
  | ITdarrow
  | ITminus
  | ITbang
  | ITstar
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
508
  | ITbiglam                    -- GHC-extension symbols
509

Ian Lynagh's avatar
Ian Lynagh committed
510
  | ITocurly                    -- special symbols
511 512 513 514 515 516
  | ITccurly
  | ITocurlybar                 -- {|, for type applications
  | ITccurlybar                 -- |}, for type applications
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
517 518
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
519 520 521 522 523 524 525 526 527
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote
dreixel's avatar
dreixel committed
528
  | ITsimpleQuote               --  '
529

Ian Lynagh's avatar
Ian Lynagh committed
530
  | ITvarid   FastString        -- identifiers
531 532 533 534 535 536 537
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)
538 539
  | ITprefixqvarsym (FastString,FastString)
  | ITprefixqconsym (FastString,FastString)
540

Ian Lynagh's avatar
Ian Lynagh committed
541
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
542 543 544 545

  | ITchar       Char
  | ITstring     FastString
  | ITinteger    Integer
546
  | ITrational   FractionalLit
547 548 549 550

  | ITprimchar   Char
  | ITprimstring FastString
  | ITprimint    Integer
Ian Lynagh's avatar
Ian Lynagh committed
551
  | ITprimword   Integer
552 553
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
554

555
  -- Template Haskell extension tokens
Ian Lynagh's avatar
Ian Lynagh committed
556 557 558 559 560 561 562 563
  | ITopenExpQuote              --  [| or [e|
  | ITopenPatQuote              --  [p|
  | ITopenDecQuote              --  [d|
  | ITopenTypQuote              --  [t|
  | ITcloseQuote                --  |]
  | ITidEscape   FastString     --  $x
  | ITparenEscape               --  $(
  | ITtyQuote                   --  ''
Ian Lynagh's avatar
Ian Lynagh committed
564
  | ITquasiQuote (FastString,FastString,RealSrcSpan) --  [:...|...|]
565 566 567 568

  -- Arrow notation extension
  | ITproc
  | ITrec
Ian Lynagh's avatar
Ian Lynagh committed
569 570 571 572 573 574
  | IToparenbar                 --  (|
  | ITcparenbar                 --  |)
  | ITlarrowtail                --  -<
  | ITrarrowtail                --  >-
  | ITLarrowtail                --  -<<
  | ITRarrowtail                --  >>-
575

Ian Lynagh's avatar
Ian Lynagh committed
576 577
  | ITunknown String            -- Used when the lexer can't make sense of it
  | ITeof                       -- end of file token
578 579 580 581 582 583 584

  -- 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
585
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
586 587
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
588

589
  deriving Show
590 591 592 593 594 595 596 597

-- 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
598
reservedWordsFM :: UniqFM (Token, Int)
599
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645
    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),
646
         ( "capi",           ITcapiconv,      bit cApiFfiBit),
Ian Lynagh's avatar
Ian Lynagh committed
647 648 649 650
         ( "prim",           ITprimcallconv,  bit ffiBit),

         ( "rec",            ITrec,           bit recBit),
         ( "proc",           ITproc,          bit arrowsBit)
651 652
     ]

Ian Lynagh's avatar
Ian Lynagh committed
653
reservedSymsFM :: UniqFM (Token, Int -> Bool)
654
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
655 656 657 658 659 660 661 662 663 664 665 666
    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)
batterseapower's avatar
batterseapower committed
667
       ,("~#",  ITtildehsh, always)
Ian Lynagh's avatar
Ian Lynagh committed
668 669 670 671 672
       ,("=>",  ITdarrow,   always)
       ,("-",   ITminus,    always)
       ,("!",   ITbang,     always)

        -- For data T (a::*) = MkT
673
       ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
674
        -- For 'forall a . t'
675
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
676 677 678 679 680

       ,("-<",  ITlarrowtail, arrowsEnabled)
       ,(">-",  ITrarrowtail, arrowsEnabled)
       ,("-<<", ITLarrowtail, arrowsEnabled)
       ,(">>-", ITRarrowtail, arrowsEnabled)
681

Ian Lynagh's avatar
Ian Lynagh committed
682 683
       ,("∷",   ITdcolon, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow, unicodeSyntaxEnabled)
Ian Lynagh's avatar
Ian Lynagh committed
684 685
       ,("∀",   ITforall, \i -> unicodeSyntaxEnabled i &&
                                explicitForallEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
686 687
       ,("→",   ITrarrow, unicodeSyntaxEnabled)
       ,("←",   ITlarrow, unicodeSyntaxEnabled)
688 689 690 691 692 693 694 695

       ,("⤙",   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
696 697 698
        -- 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).
699 700 701 702 703
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
704
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
705

706
special :: Token -> Action
twanvl's avatar
twanvl committed
707
special tok span _buf _len = return (L span tok)
708

709
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
710 711
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
712

713 714
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
715

716
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
717
skip_one_varid f span buf len
718
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
719

720
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
721
strtoken f span buf len =
722
  return (L span $! (f $! lexemeToString buf len))
723

724
init_strtoken :: Int -> (String -> Token) -> Action
725
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
726
init_strtoken drop f span buf len =
727
  return (L span $! (f $! lexemeToString buf (len-drop)))
728 729

begin :: Int -> Action
730
begin code _span _str _len = do pushLexState code; lexToken
731 732

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
733 734
pop _span _buf _len = do _ <- popLexState
                         lexToken
735

736 737 738 739 740 741 742 743 744 745 746
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
747
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
748

749
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
750 751
pop_and act span buf len = do _ <- popLexState
                              act span buf len
752

753
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
754
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
755 756
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

Ian Lynagh's avatar
Ian Lynagh committed
757
notFollowedBy :: Char -> AlexAccPred Int
Ian Lynagh's avatar
Ian Lynagh committed
758
notFollowedBy char _ _ _ (AI _ buf)
759
  = nextCharIs buf (/=char)
760

Ian Lynagh's avatar
Ian Lynagh committed
761
notFollowedBySymbol :: AlexAccPred Int
762
notFollowedBySymbol _ _ _ (AI _ buf)
763
  = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
764

765 766 767 768 769
-- 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
770
isNormalComment :: AlexAccPred Int
771
isNormalComment bits _ _ (AI _ buf)
772 773 774 775 776 777
  | haddockEnabled bits = notFollowedByDocOrPragma
  | otherwise           = nextCharIs buf (/='#')
  where
    notFollowedByDocOrPragma
       = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))

Ian Lynagh's avatar
Ian Lynagh committed
778
spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
779
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
780

Ian Lynagh's avatar
Ian Lynagh committed
781
atEOL :: AlexAccPred Int
782
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
783

Ian Lynagh's avatar
Ian Lynagh committed
784
ifExtension :: (Int -> Bool) -> AlexAccPred Int
785 786
ifExtension pred bits _ _ _ = pred bits

787 788 789
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
790
    worker commentAcc input docType oneLine = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
791
      Just ('\n', input')
792 793 794 795 796 797
        | 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
798

799 800
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
801 802 803
        check input = case alexGetChar' input of
          Just ('-', input) -> case alexGetChar' input of
            Just ('-', input) -> case alexGetChar' input of
804 805 806 807 808
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

809
        dropNonNewlineSpace input = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
810
          Just (c, input')
811 812 813 814
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
815 816 817 818 819
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

820 821 822 823
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
824
nested_comment :: P (RealLocated Token) -> Action
825
nested_comment cont span _str _len = do
826
  input <- getInput
Jedai's avatar
Jedai committed
827
  go "" (1::Int) input
828
  where
Jedai's avatar
Jedai committed
829 830 831 832 833
    go commentAcc 0 input = do setInput input
                               b <- extension rawTokenStreamEnabled
                               if b
                                 then docCommentEnd input commentAcc ITblockComment _str span
                                 else cont
834
    go commentAcc n input = case alexGetChar' input of
835
      Nothing -> errBrace input span
836
      Just ('-',input) -> case alexGetChar' input of
837
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
838 839
        Just ('\125',input) -> go commentAcc (n-1) input
        Just (_,_)          -> go ('-':commentAcc) n input
840
      Just ('\123',input) -> case alexGetChar' input of
841
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
842 843 844
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
845 846 847 848

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
849
    go commentAcc input docType _ = case alexGetChar' input of
850
      Nothing -> errBrace input span
851
      Just ('-',input) -> case alexGetChar' input of
852
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
853
        Just ('\125',input) ->
854
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
855
        Just (_,_) -> go ('-':commentAcc) input docType False
856
      Just ('\123', input) -> case alexGetChar' input of
857 858 859 860 861
        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
862
        Just (_,_) -> go ('\123':commentAcc) input docType False
863 864
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
865 866
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
867
withLexedDocType lexDocComment = do
868
  input@(AI _ buf) <- getInput
869 870 871 872
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
    '$' -> lexDocComment input ITdocCommentNamed False
David Waern's avatar
David Waern committed
873 874
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
875
    _ -> panic "withLexedDocType: Bad doc type"
Ian Lynagh's avatar
Ian Lynagh committed
876
 where
877
    lexDocSection n input = case alexGetChar' input of
878
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
879
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
880 881
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

882 883 884
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
885
rulePrag span _buf _len = do
886
  setExts (.|. bit inRulePragBit)
887 888 889
  return (L span ITrules_prag)

endPrag :: Action
890
endPrag span _buf _len = do
891 892 893
  setExts (.&. complement (bit inRulePragBit))
  return (L span ITclose_prag)

894 895 896 897
-- 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
898
-- than what was lexed by Alex, and the lexToken function doesn't know this, so
899
-- it writes the wrong token length to the parser state. This function is
Ian Lynagh's avatar
Ian Lynagh committed
900
-- called afterwards, so it can just update the state.
901 902

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
903
                 RealSrcSpan -> P (RealLocated Token)
904 905
docCommentEnd input commentAcc docType buf span = do
  setInput input
906
  let (AI loc nextBuf) = input
907
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
908
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
909
      last_len = byteDiff buf nextBuf
Ian Lynagh's avatar
Ian Lynagh committed
910

911
  span `seq` setLastToken span' last_len
912
  return (L span' (docType comment))
Ian Lynagh's avatar
Ian Lynagh committed
913

Ian Lynagh's avatar
Ian Lynagh committed
914 915
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
916

917
open_brace, close_brace :: Action
Ian Lynagh's avatar
Ian Lynagh committed
918
open_brace span _str _len = do
919 920
  ctx <- getContext
  setContext (NoLayout:ctx)
921
  return (L span ITocurly)
Ian Lynagh's avatar
Ian Lynagh committed
922
close_brace span _str _len = do
923
  popContext
924
  return (L span ITccurly)
925

Ian Lynagh's avatar
Ian Lynagh committed
926
qvarid, qconid :: StringBuffer -> Int -> Token
927 928
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
929

930
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
931 932 933
-- 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.
934
splitQualName orig_buf len parens = split orig_buf orig_buf
935
  where
936
    split buf dot_buf
Ian Lynagh's avatar
Ian Lynagh committed
937 938 939
        | orig_buf `byteDiff` buf >= len  = done dot_buf
        | c == '.'                        = found_dot buf'
        | otherwise                       = split buf' dot_buf
940 941
      where
       (c,buf') = nextChar buf
Ian Lynagh's avatar
Ian Lynagh committed