Lexer.x 112 KB
Newer Older
1
-----------------------------------------------------------------------------
2
-- (c) The University of Glasgow, 2006
3
--
4
-- GHC's lexer for Haskell 2010 [1].
5
--
6 7
-- This is a combination of an Alex-generated lexer [2] from a regex
-- definition, with some hand-coded bits. [3]
8 9
--
-- 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
-- References:
-- [1] https://www.haskell.org/onlinereport/haskell2010/haskellch2.html
-- [2] http://www.haskell.org/alex/
-- [3] https://ghc.haskell.org/trac/ghc/wiki/Commentary/Compiler/Parser
--
18 19 20 21 22 23 24 25 26 27
-----------------------------------------------------------------------------

--   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

28
--   qualified operator NOTES.
Ian Lynagh's avatar
Ian Lynagh committed
29
--
30 31 32 33 34 35 36 37 38
--   - 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.

39 40 41 42

-- -----------------------------------------------------------------------------
-- Alex "Haskell code fragment top"

43
{
44
{-# LANGUAGE BangPatterns #-}
45 46

-- See Note [Warnings in code generated by Alex] in compiler/parser/Lexer.x
47 48
{-# OPTIONS_GHC -fno-warn-unused-matches #-}
{-# OPTIONS_GHC -fno-warn-unused-binds #-}
49
{-# OPTIONS_GHC -fno-warn-unused-imports #-}
50
{-# OPTIONS_GHC -fno-warn-tabs #-}
51
{-# OPTIONS_GHC -fno-warn-missing-signatures #-}
52

53 54
{-# OPTIONS_GHC -funbox-strict-fields #-}

55
module Lexer (
56 57 58
   Token(..), lexer, pragState, mkPState, mkPStatePure, PState(..),
   P(..), ParseResult(..), mkParserFlags, ParserFlags(..), getSrcLoc,
   getPState, extopt, withThisPackage,
59
   failLocMsgP, failSpanMsgP, srcParseFail,
Ian Lynagh's avatar
Ian Lynagh committed
60
   getMessages,
61
   popContext, pushModuleContext, setLastToken, setSrcLoc,
vivian's avatar
vivian committed
62
   activeContext, nextIsEOF,
simonpj@microsoft.com's avatar
simonpj@microsoft.com committed
63
   getLexState, popLexState, pushLexState,
64
   extension, bangPatEnabled, datatypeContextsEnabled,
65
   traditionalRecordSyntaxEnabled,
66 67
   explicitForallEnabled,
   inRulePrag,
cactus's avatar
cactus committed
68 69 70
   explicitNamespacesEnabled,
   patternSynonymsEnabled,
   sccProfilingOn, hpcEnabled,
Jedai's avatar
Jedai committed
71
   addWarning,
Alan Zimmerman's avatar
Alan Zimmerman committed
72
   lexTokenStream,
73 74
   addAnnotation,AddAnn,addAnnsAt,mkParensApiAnn,
   moveAnnotations
75 76
  ) where

rodlogic's avatar
rodlogic committed
77 78
-- base
import Control.Monad
quchen's avatar
quchen committed
79 80 81
#if __GLASGOW_HASKELL__ > 710
import Control.Monad.Fail
#endif
rodlogic's avatar
rodlogic committed
82
import Data.Bits
83
import Data.Char
rodlogic's avatar
rodlogic committed
84 85 86 87
import Data.List
import Data.Maybe
import Data.Word

88 89 90
import Data.IntSet (IntSet)
import qualified Data.IntSet as IntSet

91 92 93
-- ghc-boot
import qualified GHC.LanguageExtensions as LangExt

rodlogic's avatar
rodlogic committed
94 95 96 97 98 99 100 101
-- bytestring
import Data.ByteString (ByteString)

-- containers
import Data.Map (Map)
import qualified Data.Map as Map

-- compiler/utils
102
import Bag
103 104 105 106
import Outputable
import StringBuffer
import FastString
import UniqFM
rodlogic's avatar
rodlogic committed
107 108 109 110
import Util             ( readRational )

-- compiler/main
import ErrUtils
111
import DynFlags
rodlogic's avatar
rodlogic committed
112 113 114

-- compiler/basicTypes
import SrcLoc
115
import Module
Alan Zimmerman's avatar
Alan Zimmerman committed
116 117
import BasicTypes     ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..),
                        SourceText )
118

rodlogic's avatar
rodlogic committed
119 120
-- compiler/parser
import Ctype
Alan Zimmerman's avatar
Alan Zimmerman committed
121 122

import ApiAnnotation
123 124
}

125 126 127
-- -----------------------------------------------------------------------------
-- Alex "Character set macros"

128 129
-- NB: The logic behind these definitions is also reflected in basicTypes/Lexeme.hs
-- Any changes here should likely be reflected there.
130 131 132 133
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetByte.
$nl          = [\n\r\f]
$whitechar   = [$nl\v\ $unispace]
$white_no_nl = $whitechar # \n -- TODO #8424
134
$tab         = \t
135 136

$ascdigit  = 0-9
137
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetByte.
138
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
139 140 141
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
142 143 144
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~\:]
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetByte.
$symbol    = [$ascsymbol $unisymbol] # [$special \_\"\']
145

146
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetByte.
147
$asclarge  = [A-Z]
148 149
$large     = [$asclarge $unilarge]

150
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetByte.
151
$ascsmall  = [a-z]
152 153
$small     = [$ascsmall $unismall \_]

154 155
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetByte.
$graphic   = [$small $large $symbol $digit $special $unigraphic \"\']
156

157
$binit     = 0-1
Ian Lynagh's avatar
Ian Lynagh committed
158
$octit     = 0-7
159
$hexit     = [$decdigit A-F a-f]
160

161 162
$uniidchar = \x07 -- Trick Alex into handling Unicode. See alexGetByte.
$idchar    = [$small $large $digit $uniidchar \']
163

164 165
$pragmachar = [$small $large $digit]

166 167
$docsym    = [\| \^ \* \$]

168

169 170 171 172 173 174 175 176
-- -----------------------------------------------------------------------------
-- Alex "Regular expression macros"

@varid     = $small $idchar*          -- variable identifiers
@conid     = $large $idchar*          -- constructor identifiers

@varsym    = ($symbol # \:) $symbol*  -- variable (operator) symbol
@consym    = \: $symbol*              -- constructor (operator) symbol
177

178
@decimal     = $decdigit+
179
@binary      = $binit+
180 181 182 183 184
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

@qual = (@conid \.)+
185 186 187 188
@qvarid = @qual @varid
@qconid = @qual @conid
@qvarsym = @qual @varsym
@qconsym = @qual @consym
189 190 191

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

192 193 194 195 196
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

197 198 199 200

-- -----------------------------------------------------------------------------
-- Alex "Identifier"

201 202
haskell :-

203 204 205 206 207

-- -----------------------------------------------------------------------------
-- Alex "Rules"

-- everywhere: skip whitespace
Ian Lynagh's avatar
Ian Lynagh committed
208
$white_no_nl+ ;
209
$tab          { warnTab }
210 211 212 213 214

-- 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
215 216 217 218
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
219 220 221 222

-- 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.
223 224

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
225
-- out here.
226 227 228 229 230 231 232

-- 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.

233
"-- " ~$docsym .* { lineCommentToken }
234
"--" [^$symbol \ ] .* { lineCommentToken }
235 236 237

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

238
"-- " $docsym .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
239 240 241 242 243 244 245

-- 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.

246
"---"\-* ~$symbol .* { lineCommentToken }
247 248 249 250

-- 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
251
"--"\-* / { atEOL } { lineCommentToken }
252 253 254 255

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

Jedai's avatar
Jedai committed
256
"-- " / { atEOL } { lineCommentToken }
257 258 259 260 261 262 263 264 265 266 267

-- '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
268
  \n                                    ;
Adam Gundry's avatar
Adam Gundry committed
269 270
  ^\# line                              { begin line_prag1 }
  ^\# / { followedByDigit }             { begin line_prag1 }
Ian Lynagh's avatar
Ian Lynagh committed
271 272 273
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
274 275 276 277 278
}

-- 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.
279
<layout, layout_do, layout_if> {
Ian Lynagh's avatar
Ian Lynagh committed
280 281 282 283
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
284 285
}

286 287
-- after an 'if', a vertical bar starts a layout context for MultiWayIf
<layout_if> {
288
  \| / { notFollowedBySymbol }          { new_layout_context True dontGenerateSemic ITvbar }
289 290 291
  ()                                    { pop }
}

292
-- do is treated in a subtly different way, see new_layout_context
293 294
<layout>    ()                          { new_layout_context True  generateSemic ITvocurly }
<layout_do> ()                          { new_layout_context False generateSemic ITvocurly }
295 296 297 298

-- 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
299
<layout_left>  ()                       { do_layout_left }
300

Ian Lynagh's avatar
Ian Lynagh committed
301
<0,option_prags> \n                     { begin bol }
302

303 304
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
305 306 307

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
308
<line_prag1> @decimal                   { setLine line_prag1a }
Ian Lynagh's avatar
Ian Lynagh committed
309 310
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
311 312 313

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
314
<line_prag2> @decimal                   { setLine line_prag2a }
Ian Lynagh's avatar
Ian Lynagh committed
315 316
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
317 318
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
319

320
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
321
  "{-#" $whitechar* $pragmachar+
322 323 324 325 326
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
327

328
  -- We ignore all these pragmas, but don't generate a warning for them
329 330
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
331 332

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
333
  "#-}"                          { endPrag }
334 335
}

336
<option_prags> {
337 338
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }
David Waern's avatar
David Waern committed
339 340
}

341 342
<0> {
  -- In the "0" mode we ignore these pragmas
343
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
344 345 346
                     { nested_comment lexToken }
}

347
<0,option_prags> {
348 349
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
350 351
}

352 353
-- '0' state: ordinary lexemes

354 355
-- Haddock comments

356
<0,option_prags> {
David Waern's avatar
David Waern committed
357 358
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
359 360
}

361 362
-- "special" symbols

363
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
364 365
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
366
}
Ian Lynagh's avatar
Ian Lynagh committed
367

368
<0> {
369 370
  "[|"        / { ifExtension thQuotesEnabled } { token (ITopenExpQuote NoE
                                                                NormalSyntax) }
371
  "[||"       / { ifExtension thQuotesEnabled } { token (ITopenTExpQuote NoE) }
372 373
  "[e|"       / { ifExtension thQuotesEnabled } { token (ITopenExpQuote HasE
                                                                NormalSyntax) }
374 375 376 377
  "[e||"      / { ifExtension thQuotesEnabled } { token (ITopenTExpQuote HasE) }
  "[p|"       / { ifExtension thQuotesEnabled } { token ITopenPatQuote }
  "[d|"       / { ifExtension thQuotesEnabled } { layout_token ITopenDecQuote }
  "[t|"       / { ifExtension thQuotesEnabled } { token ITopenTypQuote }
378 379
  "|]"        / { ifExtension thQuotesEnabled } { token (ITcloseQuote
                                                                NormalSyntax) }
380
  "||]"       / { ifExtension thQuotesEnabled } { token ITcloseTExpQuote }
381 382 383 384
  \$ @varid   / { ifExtension thEnabled } { skip_one_varid ITidEscape }
  "$$" @varid / { ifExtension thEnabled } { skip_two_varid ITidTyEscape }
  "$("        / { ifExtension thEnabled } { token ITparenEscape }
  "$$("       / { ifExtension thEnabled } { token ITparenTyEscape }
385

386
  "[" @varid "|"  / { ifExtension qqEnabled }
387
                     { lex_quasiquote_tok }
388 389

  -- qualified quasi-quote (#5555)
390 391
  "[" @qvarid "|"  / { ifExtension qqEnabled }
                     { lex_qquasiquote_tok }
392 393 394 395 396 397 398 399 400

  $unigraphic -- ⟦
    / { ifCurrentChar '⟦' `alexAndPred`
        ifExtension (\i -> unicodeSyntaxEnabled i && thQuotesEnabled i) }
    { token (ITopenExpQuote NoE UnicodeSyntax) }
  $unigraphic -- ⟧
    / { ifCurrentChar '⟧' `alexAndPred`
        ifExtension (\i -> unicodeSyntaxEnabled i && thQuotesEnabled i) }
    { token (ITcloseQuote UnicodeSyntax) }
401 402
}

eir@cis.upenn.edu's avatar
eir@cis.upenn.edu committed
403 404 405 406 407 408 409 410
  -- See Note [Lexing type applications]
<0> {
    [^ $idchar \) ] ^
  "@"
    / { ifExtension typeApplicationEnabled `alexAndPred` notFollowedBySymbol }
    { token ITtypeApp }
}

411
<0> {
ross's avatar
ross committed
412
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
413 414 415 416 417 418 419 420 421 422 423
                                        { special (IToparenbar NormalSyntax) }
  "|)" / { ifExtension arrowsEnabled }  { special (ITcparenbar NormalSyntax) }

  $unigraphic -- ⦇
    / { ifCurrentChar '⦇' `alexAndPred`
        ifExtension (\i -> unicodeSyntaxEnabled i && arrowsEnabled i) }
    { special (IToparenbar UnicodeSyntax) }
  $unigraphic -- ⦈
    / { ifCurrentChar '⦈' `alexAndPred`
        ifExtension (\i -> unicodeSyntaxEnabled i && arrowsEnabled i) }
    { special (ITcparenbar UnicodeSyntax) }
424 425
}

426
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
427
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
428 429
}

Adam Gundry's avatar
Adam Gundry committed
430 431 432 433 434
<0> {
  "#" @varid / { ifExtension overloadedLabelsEnabled }
               { skip_one_varid ITlabelvarid }
}

435
<0> {
436
  "(#" / { orExtensions unboxedTuplesEnabled unboxedSumsEnabled }
Ian Lynagh's avatar
Ian Lynagh committed
437
         { token IToubxparen }
438
  "#)" / { orExtensions unboxedTuplesEnabled unboxedSumsEnabled }
Ian Lynagh's avatar
Ian Lynagh committed
439 440 441
         { token ITcubxparen }
}

442
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
443 444 445 446 447 448 449 450 451 452
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
453 454
}

455
<0,option_prags> {
456 457
  @qvarid                       { idtoken qvarid }
  @qconid                       { idtoken qconid }
Ian Lynagh's avatar
Ian Lynagh committed
458
  @varid                        { varid }
459
  @conid                        { idtoken conid }
460 461
}

462
<0> {
463 464
  @qvarid "#"+      / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qconid "#"+      / { ifExtension magicHashEnabled } { idtoken qconid }
465
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
466
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
467 468
}

469 470
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
471
<0> {
472 473
  @qvarsym                                         { idtoken qvarsym }
  @qconsym                                         { idtoken qconsym }
474 475
  @varsym                                          { varsym }
  @consym                                          { consym }
476 477
}

478 479
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
480
<0> {
481
  -- Normal integral literals (:: Num a => a, from Integer)
482
  @decimal                                                               { tok_num positive 0 0 decimal }
483
  0[bB] @binary                / { ifExtension binaryLiteralsEnabled }   { tok_num positive 2 2 binary }
484 485 486
  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 }
487 488
  @negative 0[bB] @binary      / { ifExtension negativeLiteralsEnabled `alexAndPred`
                                   ifExtension binaryLiteralsEnabled }   { tok_num negative 3 3 binary }
489 490
  @negative 0[oO] @octal       / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 octal }
  @negative 0[xX] @hexadecimal / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 hexadecimal }
491 492

  -- Normal rational literals (:: Fractional a => a, from Rational)
493 494
  @floating_point                                                        { strtoken tok_float }
  @negative @floating_point    / { ifExtension negativeLiteralsEnabled } { strtoken tok_float }
495 496
}

497
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
498
  -- Unboxed ints (:: Int#) and words (:: Word#)
499 500
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
501
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
502 503
  0[bB] @binary                \# / { ifExtension magicHashEnabled `alexAndPred`
                                      ifExtension binaryLiteralsEnabled } { tok_primint positive 2 3 binary }
Ian Lynagh's avatar
Ian Lynagh committed
504 505 506
  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 }
507 508
  @negative 0[bB] @binary      \# / { ifExtension magicHashEnabled `alexAndPred`
                                      ifExtension binaryLiteralsEnabled } { tok_primint negative 3 4 binary }
Ian Lynagh's avatar
Ian Lynagh committed
509
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
510
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
511

Ian Lynagh's avatar
Ian Lynagh committed
512
  @decimal                     \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
513 514
  0[bB] @binary                \# \# / { ifExtension magicHashEnabled `alexAndPred`
                                         ifExtension binaryLiteralsEnabled } { tok_primword 2 4 binary }
Ian Lynagh's avatar
Ian Lynagh committed
515 516 517
  0[oO] @octal                 \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
  0[xX] @hexadecimal           \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }

518 519
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
520 521
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
522 523 524 525 526 527
}

-- 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.
528
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
529 530
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
531 532
}

eir@cis.upenn.edu's avatar
eir@cis.upenn.edu committed
533 534 535 536 537 538 539 540 541 542 543
-- Note [Lexing type applications]
-- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
-- The desired syntax for type applications is to prefix the type application
-- with '@', like this:
--
--   foo @Int @Bool baz bum
--
-- This, of course, conflicts with as-patterns. The conflict arises because
-- expressions and patterns use the same parser, and also because we want
-- to allow type patterns within expression patterns.
--
Gabor Greif's avatar
Gabor Greif committed
544
-- Disambiguation is accomplished by requiring *something* to appear between
eir@cis.upenn.edu's avatar
eir@cis.upenn.edu committed
545 546 547 548 549 550 551 552 553 554 555 556 557 558
-- type application and the preceding token. This something must end with
-- a character that cannot be the end of the variable bound in an as-pattern.
-- Currently (June 2015), this means that the something cannot end with a
-- $idchar or a close-paren. (The close-paren is necessary if the as-bound
-- identifier is symbolic.)
--
-- Note that looking for whitespace before the '@' is insufficient, because
-- of this pathological case:
--
--   foo {- hi -}@Int
--
-- This design is predicated on the fact that as-patterns are generally
-- whitespace-free, and also that this whole thing is opt-in, with the
-- TypeApplications extension.
559 560 561 562

-- -----------------------------------------------------------------------------
-- Alex "Haskell code fragment bottom"

563
{
564

565 566 567
-- -----------------------------------------------------------------------------
-- The token type

568
data Token
Ian Lynagh's avatar
Ian Lynagh committed
569
  = ITas                        -- Haskell keywords
570 571 572 573 574 575 576 577
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
  | ITdo
  | ITelse
  | IThiding
578
  | ITforeign
579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere

595
  | ITforall            IsUnicodeSyntax -- GHC extension keywords
596 597 598 599
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
600
  | ITinterruptible
601 602 603
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
604
  | ITcapiconv
605
  | ITprimcallconv
thoughtpolice's avatar
thoughtpolice committed
606
  | ITjavascriptcallconv
607
  | ITmdo
608
  | ITfamily
609
  | ITrole
610 611 612
  | ITgroup
  | ITby
  | ITusing
cactus's avatar
cactus committed
613
  | ITpattern
Facundo Domínguez's avatar
Facundo Domínguez committed
614
  | ITstatic
Ryan Scott's avatar
Ryan Scott committed
615 616
  | ITstock
  | ITanyclass
617

Edward Z. Yang's avatar
Edward Z. Yang committed
618 619 620 621 622 623
  -- Backpack tokens
  | ITunit
  | ITsignature
  | ITdependency
  | ITrequires

Alan Zimmerman's avatar
Alan Zimmerman committed
624 625 626 627 628 629 630 631
  -- Pragmas, see  note [Pragma source text] in BasicTypes
  | ITinline_prag       SourceText InlineSpec RuleMatchInfo
  | ITspec_prag         SourceText                -- SPECIALISE
  | ITspec_inline_prag  SourceText Bool    -- SPECIALISE INLINE (or NOINLINE)
  | ITsource_prag       SourceText
  | ITrules_prag        SourceText
  | ITwarning_prag      SourceText
  | ITdeprecated_prag   SourceText
632
  | ITline_prag
Alan Zimmerman's avatar
Alan Zimmerman committed
633 634 635 636 637 638
  | ITscc_prag          SourceText
  | ITgenerated_prag    SourceText
  | ITcore_prag         SourceText         -- hdaume: core annotations
  | ITunpack_prag       SourceText
  | ITnounpack_prag     SourceText
  | ITann_prag          SourceText
639
  | ITclose_prag
640 641 642
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
Alan Zimmerman's avatar
Alan Zimmerman committed
643 644 645 646 647 648 649 650 651
  | ITvect_prag         SourceText
  | ITvect_scalar_prag  SourceText
  | ITnovect_prag       SourceText
  | ITminimal_prag      SourceText
  | IToverlappable_prag SourceText  -- instance overlap mode
  | IToverlapping_prag  SourceText  -- instance overlap mode
  | IToverlaps_prag     SourceText  -- instance overlap mode
  | ITincoherent_prag   SourceText  -- instance overlap mode
  | ITctype             SourceText
652

Ian Lynagh's avatar
Ian Lynagh committed
653
  | ITdotdot                    -- reserved symbols
654
  | ITcolon
655
  | ITdcolon            IsUnicodeSyntax
656 657
  | ITequal
  | ITlam
658
  | ITlcase
659
  | ITvbar
660 661
  | ITlarrow            IsUnicodeSyntax
  | ITrarrow            IsUnicodeSyntax
662 663
  | ITat
  | ITtilde
batterseapower's avatar
batterseapower committed
664
  | ITtildehsh
665
  | ITdarrow            IsUnicodeSyntax
666 667 668 669
  | ITminus
  | ITbang
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
670
  | ITbiglam                    -- GHC-extension symbols
671

Ian Lynagh's avatar
Ian Lynagh committed
672
  | ITocurly                    -- special symbols
673 674 675 676
  | ITccurly
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
677 678
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
679 680 681 682 683 684 685 686 687
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote
dreixel's avatar
dreixel committed
688
  | ITsimpleQuote               --  '
689

Ian Lynagh's avatar
Ian Lynagh committed
690
  | ITvarid   FastString        -- identifiers
691 692 693 694 695 696 697 698
  | ITconid   FastString
  | ITvarsym  FastString
  | ITconsym  FastString
  | ITqvarid  (FastString,FastString)
  | ITqconid  (FastString,FastString)
  | ITqvarsym (FastString,FastString)
  | ITqconsym (FastString,FastString)

Ian Lynagh's avatar
Ian Lynagh committed
699
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
Adam Gundry's avatar
Adam Gundry committed
700
  | ITlabelvarid   FastString   -- Overloaded label: #x
701

702 703 704
  | ITchar     SourceText Char       -- Note [Literal source text] in BasicTypes
  | ITstring   SourceText FastString -- Note [Literal source text] in BasicTypes
  | ITinteger  SourceText Integer    -- Note [Literal source text] in BasicTypes
Alan Zimmerman's avatar
Alan Zimmerman committed
705
  | ITrational FractionalLit
706

707 708 709 710
  | ITprimchar   SourceText Char     -- Note [Literal source text] in BasicTypes
  | ITprimstring SourceText ByteString -- Note [Literal source text] @BasicTypes
  | ITprimint    SourceText Integer  -- Note [Literal source text] in BasicTypes
  | ITprimword   SourceText Integer  -- Note [Literal source text] in BasicTypes
711 712
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
713

714
  -- Template Haskell extension tokens
715 716 717 718 719 720 721 722 723 724 725 726
  | ITopenExpQuote HasE IsUnicodeSyntax --  [| or [e|
  | ITopenPatQuote                      --  [p|
  | ITopenDecQuote                      --  [d|
  | ITopenTypQuote                      --  [t|
  | ITcloseQuote IsUnicodeSyntax        --  |]
  | ITopenTExpQuote HasE                --  [|| or [e||
  | ITcloseTExpQuote                    --  ||]
  | ITidEscape   FastString             --  $x
  | ITparenEscape                       --  $(
  | ITidTyEscape   FastString           --  $$x
  | ITparenTyEscape                     --  $$(
  | ITtyQuote                           --  ''
727 728 729 730 731 732 733 734
  | 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 |]
735 736 737 738

  -- Arrow notation extension
  | ITproc
  | ITrec
739 740
  | IToparenbar  IsUnicodeSyntax --  (|
  | ITcparenbar  IsUnicodeSyntax --  |)
741 742 743 744
  | ITlarrowtail IsUnicodeSyntax --  -<
  | ITrarrowtail IsUnicodeSyntax --  >-
  | ITLarrowtail IsUnicodeSyntax --  -<<
  | ITRarrowtail IsUnicodeSyntax --  >>-
745

eir@cis.upenn.edu's avatar
eir@cis.upenn.edu committed
746 747 748 749 750 751 752
  -- type application '@' (lexed differently than as-pattern '@',
  -- due to checking for preceding whitespace)
  | ITtypeApp


  | ITunknown String            -- Used when the lexer can't make sense of it
  | ITeof                       -- end of file token
753 754 755 756 757 758 759

  -- 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)
Jedai's avatar
Jedai committed
760 761
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
762

763
  deriving Show
764

Alan Zimmerman's avatar
Alan Zimmerman committed
765 766 767
instance Outputable Token where
  ppr x = text (show x)

Alan Zimmerman's avatar
Alan Zimmerman committed
768

769 770 771 772 773 774 775
-- 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)
--
776
reservedWordsFM :: UniqFM (Token, ExtsBitmap)
777
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804
    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 ),

805 806
         ( "forall",         ITforall NormalSyntax,
                                              xbit ExplicitForallBit .|.
807 808
                                              xbit InRulePragBit),
         ( "mdo",            ITmdo,           xbit RecursiveDoBit),
809 810 811
             -- See Note [Lexing type pseudo-keywords]
         ( "family",         ITfamily,        0 ),
         ( "role",           ITrole,          0 ),
812
         ( "pattern",        ITpattern,       xbit PatternSynonymsBit),
Facundo Domínguez's avatar
Facundo Domínguez committed
813
         ( "static",         ITstatic,        0 ),
Ryan Scott's avatar
Ryan Scott committed
814 815
         ( "stock",          ITstock,         0 ),
         ( "anyclass",       ITanyclass,      0 ),
816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833
         ( "group",          ITgroup,         xbit TransformComprehensionsBit),
         ( "by",             ITby,            xbit TransformComprehensionsBit),
         ( "using",          ITusing,         xbit TransformComprehensionsBit),

         ( "foreign",        ITforeign,       xbit FfiBit),
         ( "export",         ITexport,        xbit FfiBit),
         ( "label",          ITlabel,         xbit FfiBit),
         ( "dynamic",        ITdynamic,       xbit FfiBit),
         ( "safe",           ITsafe,          xbit FfiBit .|.
                                              xbit SafeHaskellBit),
         ( "interruptible",  ITinterruptible, xbit InterruptibleFfiBit),
         ( "unsafe",         ITunsafe,        xbit FfiBit),
         ( "stdcall",        ITstdcallconv,   xbit FfiBit),
         ( "ccall",          ITccallconv,     xbit FfiBit),
         ( "capi",           ITcapiconv,      xbit CApiFfiBit),
         ( "prim",           ITprimcallconv,  xbit FfiBit),
         ( "javascript",     ITjavascriptcallconv, xbit FfiBit),

Edward Z. Yang's avatar
Edward Z. Yang committed
834 835 836 837
         ( "unit",           ITunit,          0 ),
         ( "dependency",     ITdependency,       0 ),
         ( "signature",      ITsignature,     0 ),

838 839 840
         ( "rec",            ITrec,           xbit ArrowsBit .|.
                                              xbit RecursiveDoBit),
         ( "proc",           ITproc,          xbit ArrowsBit)
841 842
     ]

843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858
{-----------------------------------
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.
-------------------------------------}
859

860
reservedSymsFM :: UniqFM (Token, ExtsBitmap -> Bool)
861
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
862
    map (\ (x,y,z) -> (mkFastString x,(y,z)))
863
      [ ("..",  ITdotdot,              always)
Ian Lynagh's avatar
Ian Lynagh committed
864
        -- (:) is a reserved op, meaning only list cons
865 866 867 868 869 870 871 872 873 874 875 876 877
       ,(":",   ITcolon,               always)
       ,("::",  ITdcolon NormalSyntax, always)
       ,("=",   ITequal,               always)
       ,("\\",  ITlam,                 always)
       ,("|",   ITvbar,                always)
       ,("<-",  ITlarrow NormalSyntax, always)
       ,("->",  ITrarrow NormalSyntax, always)
       ,("@",   ITat,                  always)
       ,("~",   ITtilde,               always)
       ,("~#",  ITtildehsh,            magicHashEnabled)
       ,("=>",  ITdarrow NormalSyntax, always)
       ,("-",   ITminus,               always)
       ,("!",   ITbang,                always)
Ian Lynagh's avatar
Ian Lynagh committed
878 879

        -- For 'forall a . t'
880
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
881

882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901
       ,("-<",  ITlarrowtail NormalSyntax, arrowsEnabled)
       ,(">-",  ITrarrowtail NormalSyntax, arrowsEnabled)
       ,("-<<", ITLarrowtail NormalSyntax, arrowsEnabled)
       ,(">>-", ITRarrowtail NormalSyntax, arrowsEnabled)

       ,("∷",   ITdcolon UnicodeSyntax, unicodeSyntaxEnabled)
       ,("⇒",   ITdarrow UnicodeSyntax, unicodeSyntaxEnabled)
       ,("∀",   ITforall UnicodeSyntax, unicodeSyntaxEnabled)
       ,("→",   ITrarrow UnicodeSyntax, unicodeSyntaxEnabled)
       ,("←",   ITlarrow UnicodeSyntax, unicodeSyntaxEnabled)

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

Simon Marlow's avatar
Simon Marlow committed
902 903 904
        -- 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).
905 906 907 908 909
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
910
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
911

912
special :: Token -> Action
twanvl's avatar
twanvl committed
913
special tok span _buf _len = return (L span tok)
914

915
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
916 917
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
918

919 920
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
921

922
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
923
skip_one_varid f span buf len
924
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
925

926 927 928 929
skip_two_varid :: (FastString -> Token) -> Action
skip_two_varid f span buf len
  = return (L span $! f (lexemeToFastString (stepOn (stepOn buf)) (len-2)))

930
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
931
strtoken f span buf len =
932
  return (L span $! (f $! lexemeToString buf len))
933

934
init_strtoken :: Int -> (String -> Token) -> Action
935
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
936
init_strtoken drop f span buf len =
937
  return (L span $! (f $! lexemeToString buf (len-drop)))
938 939

begin :: Int -> Action
940
begin code _span _str _len = do pushLexState code; lexToken
941 942

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
943 944
pop _span _buf _len = do _ <- popLexState
                         lexToken
945

946 947 948 949 950 951 952 953
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
954 955
                 Layout prev_off _ : _ -> prev_off < offset
                 _                     -> True
956
      if isOK then pop_and open_brace span buf len
Ian Lynagh's avatar
Ian Lynagh committed
957
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
958

959
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
960 961
pop_and act span buf len = do _ <- popLexState
                              act span buf len
962

963
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
964
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
965 966
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

967 968 969 970
{-# INLINE nextCharIsNot #-}
nextCharIsNot :: StringBuffer -> (Char -> Bool) -> Bool
nextCharIsNot buf p = not (nextCharIs buf p)

971
notFollowedBy :: Char -> AlexAccPred ExtsBitmap
Ian Lynagh's avatar
Ian Lynagh committed
972
notFollowedBy char _ _ _ (AI _ buf)
973
  = nextCharIsNot buf (== char)
974

975
notFollowedBySymbol :: AlexAccPred ExtsBitmap
976
notFollowedBySymbol _ _ _ (AI _ buf)
977
  = nextCharIsNot buf (`elem` "!#$%&*+./<=>?@\\^|-~")
ross's avatar
ross committed
978

Adam Gundry's avatar
Adam Gundry committed
979 980 981 982
followedByDigit :: AlexAccPred ExtsBitmap
followedByDigit _ _ _ (AI _ buf)
  = afterOptionalSpace buf (\b -> nextCharIs b (`elem` ['0'..'9']))

983 984 985 986
ifCurrentChar :: Char -> AlexAccPred ExtsBitmap
ifCurrentChar char _ (AI _ buf) _ _
  = nextCharIs buf (== char)

987 988 989 990 991
-- 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.
992
isNormalComment :: AlexAccPred ExtsBitmap
993
isNormalComment bits _ _ (AI _ buf)
994
  | haddockEnabled bits = notFollowedByDocOrPragma
995
  | otherwise           = nextCharIsNot buf (== '#')
996 997
  where
    notFollowedByDocOrPragma
998
       = afterOptionalSpace buf (\b -> nextCharIsNot b (`elem` "|^*$#"))
999

1000 1001 1002 1003 1004
afterOptionalSpace :: StringBuffer -> (StringBuffer -> Bool) -> Bool
afterOptionalSpace buf p
    = if nextCharIs buf (== ' ')
      then p (snd (nextChar buf))
      else p buf
1005

1006
atEOL :: AlexAccPred ExtsBitmap
1007
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
1008

1009
ifExtension :: (ExtsBitmap -> Bool) -> AlexAccPred ExtsBitmap
1010 1011
ifExtension pred bits _ _ _ = pred bits

1012 1013 1014
orExtensions :: (ExtsBitmap -> Bool) -> (ExtsBitmap -> Bool) -> AlexAccPred ExtsBitmap
orExtensions pred1 pred2 bits _ _ _ = pred1 bits || pred2 bits

1015 1016 1017
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
1018
    worker commentAcc input docType checkNextLine = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
1019
      Just ('\n', input')
1020 1021
        | checkNextLine -> case checkIfCommentLine input' of
          Just input -> worker ('\n':commentAcc) input docType checkNextLine
1022
          Nothing -> docCommentEnd input commentAcc docType buf span
1023 1024
        | otherwise -> docCommentEnd input commentAcc docType buf span
      Just (c, input) -> worker (c:commentAcc) input docType checkNextLine
1025
      Nothing -> docCommentEnd input commentAcc docType buf span
Ian Lynagh's avatar
Ian Lynagh committed
1026

1027 1028 1029 1030 1031 1032 1033 1034
    -- Check if the next line of input belongs to this doc comment as well.
    -- A doc comment continues onto the next line when the following
    -- conditions are met:
    --   * The line starts with "--"
    --   * The line doesn't start with "---".
    --   * The line doesn't start with "-- $", because that would be the
    --     start of a /new/ named haddock chunk (#10398).
    checkIfCommentLine :: AlexInput -> Maybe AlexInput
1035 1036
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
        check input = do
          ('-', input) <- alexGetChar' input
          ('-', input) <- alexGetChar' input
          (c, after_c) <- alexGetChar' input
          case c of
            '-' -> Nothing
            ' ' -> case alexGetChar' after_c of
                     Just ('$', _) -> Nothing
                     _ -> Just input
            _   -> Just input
1047

1048
        dropNonNewlineSpace input = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
1049
          Just (c, input')
1050 1051 1052 1053
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
1054 1055 1056 1057 1058
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

1059 1060 1061 1062
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
1063
nested_comment :: P (RealLocated Token) -> Action
Alan Zimmerman's avatar
Alan Zimmerman committed
1064
nested_comment cont span buf len = do
1065
  input <- getInput
1066
  go (reverse $ lexemeToString buf len) (1::Int) input
1067
  where
Alan Zimmerman's avatar
Alan Zimmerman committed
1068 1069 1070 1071 1072 1073
    go commentAcc 0 input = do
      setInput input
      b <- extension rawTokenStreamEnabled
      if b
        then docCommentEnd input commentAcc ITblockComment buf span
        else cont