Lexer.x 105 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 (
Alan Zimmerman's avatar
Alan Zimmerman committed
56
   Token(..), lexer, pragState, mkPState, PState(..),
Ian Lynagh's avatar
Ian Lynagh committed
57
   P(..), ParseResult(..), getSrcLoc,
58
   getPState, getDynFlags, withThisPackage,
59
   failLocMsgP, failSpanMsgP, srcParseFail,
Ian Lynagh's avatar
Ian Lynagh committed
60
   getMessages,
61
   popContext, pushCurrentContext, 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,
Alan Zimmerman's avatar
Alan Zimmerman committed
73
   addAnnotation,AddAnn,mkParensApiAnn
74 75
  ) where

rodlogic's avatar
rodlogic committed
76
-- base
77
#if __GLASGOW_HASKELL__ < 709
rodlogic's avatar
rodlogic committed
78
import Control.Applicative
79
#endif
rodlogic's avatar
rodlogic committed
80
import Control.Monad
quchen's avatar
quchen committed
81 82 83
#if __GLASGOW_HASKELL__ > 710
import Control.Monad.Fail
#endif
rodlogic's avatar
rodlogic committed
84
import Data.Bits
85
import Data.Char
rodlogic's avatar
rodlogic committed
86 87 88 89 90 91 92 93 94 95 96 97
import Data.List
import Data.Maybe
import Data.Word

-- bytestring
import Data.ByteString (ByteString)

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

-- compiler/utils
98
import Bag
99 100 101 102
import Outputable
import StringBuffer
import FastString
import UniqFM
rodlogic's avatar
rodlogic committed
103 104 105 106
import Util             ( readRational )

-- compiler/main
import ErrUtils
107
import DynFlags
rodlogic's avatar
rodlogic committed
108 109 110

-- compiler/basicTypes
import SrcLoc
111
import Module
Alan Zimmerman's avatar
Alan Zimmerman committed
112 113
import BasicTypes     ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..),
                        SourceText )
114

rodlogic's avatar
rodlogic committed
115 116
-- compiler/parser
import Ctype
Alan Zimmerman's avatar
Alan Zimmerman committed
117 118

import ApiAnnotation
119 120
}

121 122 123
-- -----------------------------------------------------------------------------
-- Alex "Character set macros"

124 125
-- NB: The logic behind these definitions is also reflected in basicTypes/Lexeme.hs
-- Any changes here should likely be reflected there.
126 127 128 129
$unispace    = \x05 -- Trick Alex into handling Unicode. See alexGetByte.
$nl          = [\n\r\f]
$whitechar   = [$nl\v\ $unispace]
$white_no_nl = $whitechar # \n -- TODO #8424
130
$tab         = \t
131 132

$ascdigit  = 0-9
133
$unidigit  = \x03 -- Trick Alex into handling Unicode. See alexGetByte.
134
$decdigit  = $ascdigit -- for now, should really be $digit (ToDo)
135 136 137
$digit     = [$ascdigit $unidigit]

$special   = [\(\)\,\;\[\]\`\{\}]
138 139 140
$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~\:]
$unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetByte.
$symbol    = [$ascsymbol $unisymbol] # [$special \_\"\']
141

142
$unilarge  = \x01 -- Trick Alex into handling Unicode. See alexGetByte.
143
$asclarge  = [A-Z]
144 145
$large     = [$asclarge $unilarge]

146
$unismall  = \x02 -- Trick Alex into handling Unicode. See alexGetByte.
147
$ascsmall  = [a-z]
148 149
$small     = [$ascsmall $unismall \_]

150 151
$unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetByte.
$graphic   = [$small $large $symbol $digit $special $unigraphic \"\']
152

153
$binit     = 0-1
Ian Lynagh's avatar
Ian Lynagh committed
154
$octit     = 0-7
155
$hexit     = [$decdigit A-F a-f]
156 157 158 159

$suffix    = \x07 -- Trick Alex into handling Unicode. See alexGetByte.
-- TODO #10196. Only allow modifier letters in the suffix of an identifier.
$idchar    = [$small $large $digit $suffix \']
160

161 162
$pragmachar = [$small $large $digit]

163 164
$docsym    = [\| \^ \* \$]

165

166 167 168 169 170 171 172 173
-- -----------------------------------------------------------------------------
-- 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
174

175
@decimal     = $decdigit+
176
@binary      = $binit+
177 178 179 180 181
@octal       = $octit+
@hexadecimal = $hexit+
@exponent    = [eE] [\-\+]? @decimal

@qual = (@conid \.)+
182 183 184 185
@qvarid = @qual @varid
@qconid = @qual @conid
@qvarsym = @qual @varsym
@qconsym = @qual @consym
186 187 188

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

189 190 191 192 193
-- normal signed numerical literals can only be explicitly negative,
-- not explicitly positive (contrast @exponent)
@negative = \-
@signed = @negative ?

194 195 196 197

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

198 199
haskell :-

200 201 202 203 204

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

-- everywhere: skip whitespace
Ian Lynagh's avatar
Ian Lynagh committed
205
$white_no_nl+ ;
206
$tab          { warnTab }
207 208 209 210 211

-- 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
212 213 214 215
-- are). We also rule out nested Haddock comments, if the -haddock flag is
-- set.

"{-" / { isNormalComment } { nested_comment lexToken }
216 217 218 219

-- 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.
220 221

-- Since Haddock comments aren't valid in every state, we need to rule them
Ian Lynagh's avatar
Ian Lynagh committed
222
-- out here.
223 224 225 226 227 228 229

-- 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
230
"-- " ~[$docsym \#] .* { lineCommentToken }
231
"--" [^$symbol \ ] .* { lineCommentToken }
232 233 234

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

Jedai's avatar
Jedai committed
235
"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
236 237 238 239 240 241 242

-- 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.

243
"---"\-* ~$symbol .* { lineCommentToken }
244 245 246 247

-- 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
248
"--"\-* / { atEOL } { lineCommentToken }
249 250 251 252

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

Jedai's avatar
Jedai committed
253
"-- " / { atEOL } { lineCommentToken }
254 255 256 257 258 259 260 261 262 263 264

-- '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
265
  \n                                    ;
Adam Gundry's avatar
Adam Gundry committed
266 267
  ^\# line                              { begin line_prag1 }
  ^\# / { followedByDigit }             { begin line_prag1 }
Ian Lynagh's avatar
Ian Lynagh committed
268 269 270
  ^\# pragma .* \n                      ; -- GCC 3.3 CPP generated, apparently
  ^\# \! .* \n                          ; -- #!, for scripts
  ()                                    { do_bol }
271 272 273 274 275
}

-- 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.
276
<layout, layout_do, layout_if> {
Ian Lynagh's avatar
Ian Lynagh committed
277 278 279 280
  \{ / { notFollowedBy '-' }            { hopefully_open_brace }
        -- we might encounter {-# here, but {- has been handled already
  \n                                    ;
  ^\# (line)?                           { begin line_prag1 }
281 282
}

283 284 285 286 287 288
-- after an 'if', a vertical bar starts a layout context for MultiWayIf
<layout_if> {
  \| / { notFollowedBySymbol }          { new_layout_context True ITvbar }
  ()                                    { pop }
}

289
-- do is treated in a subtly different way, see new_layout_context
290 291
<layout>    ()                          { new_layout_context True  ITvocurly }
<layout_do> ()                          { new_layout_context False ITvocurly }
292 293 294 295

-- 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
296
<layout_left>  ()                       { do_layout_left }
297

Ian Lynagh's avatar
Ian Lynagh committed
298
<0,option_prags> \n                     { begin bol }
299

300 301
"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
                                { dispatch_pragmas linePrags }
302 303 304

-- single-line line pragmas, of the form
--    # <line> "<file>" <extra-stuff> \n
305
<line_prag1> @decimal                   { setLine line_prag1a }
Ian Lynagh's avatar
Ian Lynagh committed
306 307
<line_prag1a> \" [$graphic \ ]* \"      { setFile line_prag1b }
<line_prag1b> .*                        { pop }
308 309 310

-- Haskell-style line pragmas, of the form
--    {-# LINE <line> "<file>" #-}
311
<line_prag2> @decimal                   { setLine line_prag2a }
Ian Lynagh's avatar
Ian Lynagh committed
312 313
<line_prag2a> \" [$graphic \ ]* \"      { setFile line_prag2b }
<line_prag2b> "#-}"|"-}"                { pop }
314 315
   -- NOTE: accept -} at the end of a LINE pragma, for compatibility
   -- with older versions of GHC which generated these.
316

317
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
318
  "{-#" $whitechar* $pragmachar+
319 320 321 322 323
        $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
                                 { dispatch_pragmas twoWordPrags }

  "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
                                 { dispatch_pragmas oneWordPrags }
324

325
  -- We ignore all these pragmas, but don't generate a warning for them
326 327
  "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
                                 { dispatch_pragmas ignoredPrags }
328 329

  -- ToDo: should only be valid inside a pragma:
Ian Lynagh's avatar
Ian Lynagh committed
330
  "#-}"                          { endPrag }
331 332
}

333
<option_prags> {
334 335 336
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
                                   { dispatch_pragmas fileHeaderPrags }

Ian Lynagh's avatar
Ian Lynagh committed
337
  "-- #"                           { multiline_doc_comment }
David Waern's avatar
David Waern committed
338 339
}

340 341
<0> {
  -- In the "0" mode we ignore these pragmas
342
  "{-#"  $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
343 344 345
                     { nested_comment lexToken }
}

David Waern's avatar
David Waern committed
346
<0> {
Jedai's avatar
Jedai committed
347
  "-- #" .* { lineCommentToken }
348
}
349

350
<0,option_prags> {
351 352
  "{-#"  { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
                    (nested_comment lexToken) }
353 354
}

355 356
-- '0' state: ordinary lexemes

357 358
-- Haddock comments

359
<0,option_prags> {
David Waern's avatar
David Waern committed
360 361
  "-- " $docsym      / { ifExtension haddockEnabled } { multiline_doc_comment }
  "{-" \ ? $docsym   / { ifExtension haddockEnabled } { nested_doc_comment }
362 363
}

364 365
-- "special" symbols

366
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
367 368
  "[:" / { ifExtension parrEnabled }    { token ITopabrack }
  ":]" / { ifExtension parrEnabled }    { token ITcpabrack }
369
}
Ian Lynagh's avatar
Ian Lynagh committed
370

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

386 387 388 389
-- For backward compatibility, accept the old dollar syntax
  "[$" @varid "|"  / { ifExtension qqEnabled }
                     { lex_quasiquote_tok }

390
  "[" @varid "|"  / { ifExtension qqEnabled }
391
                     { lex_quasiquote_tok }
392 393

  -- qualified quasi-quote (#5555)
394 395
  "[" @qvarid "|"  / { ifExtension qqEnabled }
                     { lex_qquasiquote_tok }
396 397
}

398
<0> {
399
  "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
Ian Lynagh's avatar
Ian Lynagh committed
400
                                        { special IToparenbar }
401 402 403
  "|)" / { ifExtension arrowsEnabled }  { special ITcparenbar }
}

404
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
405
  \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
406 407
}

Adam Gundry's avatar
Adam Gundry committed
408 409 410 411 412
<0> {
  "#" @varid / { ifExtension overloadedLabelsEnabled }
               { skip_one_varid ITlabelvarid }
}

413
<0> {
414
  "(#" / { ifExtension unboxedTuplesEnabled }
Ian Lynagh's avatar
Ian Lynagh committed
415 416 417 418 419
         { token IToubxparen }
  "#)" / { ifExtension unboxedTuplesEnabled }
         { token ITcubxparen }
}

420
<0,option_prags> {
Ian Lynagh's avatar
Ian Lynagh committed
421 422 423 424 425 426 427 428 429 430
  \(                                    { special IToparen }
  \)                                    { special ITcparen }
  \[                                    { special ITobrack }
  \]                                    { special ITcbrack }
  \,                                    { special ITcomma }
  \;                                    { special ITsemi }
  \`                                    { special ITbackquote }

  \{                                    { open_brace }
  \}                                    { close_brace }
431 432
}

433
<0,option_prags> {
434 435
  @qvarid                       { idtoken qvarid }
  @qconid                       { idtoken qconid }
Ian Lynagh's avatar
Ian Lynagh committed
436
  @varid                        { varid }
437
  @conid                        { idtoken conid }
438 439
}

440
<0> {
441 442
  @qvarid "#"+      / { ifExtension magicHashEnabled } { idtoken qvarid }
  @qconid "#"+      / { ifExtension magicHashEnabled } { idtoken qconid }
443
  @varid "#"+       / { ifExtension magicHashEnabled } { varid }
444
  @conid "#"+       / { ifExtension magicHashEnabled } { idtoken conid }
445 446
}

447 448
-- ToDo: - move `var` and (sym) into lexical syntax?
--       - remove backquote from $special?
449
<0> {
450 451
  @qvarsym                                         { idtoken qvarsym }
  @qconsym                                         { idtoken qconsym }
452 453
  @varsym                                          { varsym }
  @consym                                          { consym }
454 455
}

456 457
-- For the normal boxed literals we need to be careful
-- when trying to be close to Haskell98
458
<0> {
459
  -- Normal integral literals (:: Num a => a, from Integer)
460
  @decimal                                                               { tok_num positive 0 0 decimal }
461
  0[bB] @binary                / { ifExtension binaryLiteralsEnabled }   { tok_num positive 2 2 binary }
462 463 464
  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 }
465 466
  @negative 0[bB] @binary      / { ifExtension negativeLiteralsEnabled `alexAndPred`
                                   ifExtension binaryLiteralsEnabled }   { tok_num negative 3 3 binary }
467 468
  @negative 0[oO] @octal       / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 octal }
  @negative 0[xX] @hexadecimal / { ifExtension negativeLiteralsEnabled } { tok_num negative 3 3 hexadecimal }
469 470

  -- Normal rational literals (:: Fractional a => a, from Rational)
471 472
  @floating_point                                                        { strtoken tok_float }
  @negative @floating_point    / { ifExtension negativeLiteralsEnabled } { strtoken tok_float }
473 474
}

475
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
476
  -- Unboxed ints (:: Int#) and words (:: Word#)
477 478
  -- It's simpler (and faster?) to give separate cases to the negatives,
  -- especially considering octal/hexadecimal prefixes.
Ian Lynagh's avatar
Ian Lynagh committed
479
  @decimal                     \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
480 481
  0[bB] @binary                \# / { ifExtension magicHashEnabled `alexAndPred`
                                      ifExtension binaryLiteralsEnabled } { tok_primint positive 2 3 binary }
Ian Lynagh's avatar
Ian Lynagh committed
482 483 484
  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 }
485 486
  @negative 0[bB] @binary      \# / { ifExtension magicHashEnabled `alexAndPred`
                                      ifExtension binaryLiteralsEnabled } { tok_primint negative 3 4 binary }
Ian Lynagh's avatar
Ian Lynagh committed
487
  @negative 0[oO] @octal       \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
488
  @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
489

Ian Lynagh's avatar
Ian Lynagh committed
490
  @decimal                     \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
491 492
  0[bB] @binary                \# \# / { ifExtension magicHashEnabled `alexAndPred`
                                         ifExtension binaryLiteralsEnabled } { tok_primword 2 4 binary }
Ian Lynagh's avatar
Ian Lynagh committed
493 494 495
  0[oO] @octal                 \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
  0[xX] @hexadecimal           \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }

496 497
  -- Unboxed floats and doubles (:: Float#, :: Double#)
  -- prim_{float,double} work with signed literals
498 499
  @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
  @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
500 501 502 503 504 505
}

-- 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.
506
<0> {
Ian Lynagh's avatar
Ian Lynagh committed
507 508
  \'                            { lex_char_tok }
  \"                            { lex_string_tok }
509 510
}

511 512 513 514

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

515
{
516

517 518 519
-- -----------------------------------------------------------------------------
-- The token type

520
data Token
Ian Lynagh's avatar
Ian Lynagh committed
521
  = ITas                        -- Haskell keywords
522 523 524 525 526 527 528 529
  | ITcase
  | ITclass
  | ITdata
  | ITdefault
  | ITderiving
  | ITdo
  | ITelse
  | IThiding
530
  | ITforeign
531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546
  | ITif
  | ITimport
  | ITin
  | ITinfix
  | ITinfixl
  | ITinfixr
  | ITinstance
  | ITlet
  | ITmodule
  | ITnewtype
  | ITof
  | ITqualified
  | ITthen
  | ITtype
  | ITwhere

547
  | ITforall            IsUnicodeSyntax -- GHC extension keywords
548 549 550 551
  | ITexport
  | ITlabel
  | ITdynamic
  | ITsafe
552
  | ITinterruptible
553 554 555
  | ITunsafe
  | ITstdcallconv
  | ITccallconv
556
  | ITcapiconv
557
  | ITprimcallconv
thoughtpolice's avatar
thoughtpolice committed
558
  | ITjavascriptcallconv
559
  | ITmdo
560
  | ITfamily
561
  | ITrole
562 563 564
  | ITgroup
  | ITby
  | ITusing
cactus's avatar
cactus committed
565
  | ITpattern
566
  | ITstatic
567

Alan Zimmerman's avatar
Alan Zimmerman committed
568 569 570 571 572 573 574 575
  -- 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
576
  | ITline_prag
Alan Zimmerman's avatar
Alan Zimmerman committed
577 578 579 580 581 582
  | ITscc_prag          SourceText
  | ITgenerated_prag    SourceText
  | ITcore_prag         SourceText         -- hdaume: core annotations
  | ITunpack_prag       SourceText
  | ITnounpack_prag     SourceText
  | ITann_prag          SourceText
583
  | ITclose_prag
584 585 586
  | IToptions_prag String
  | ITinclude_prag String
  | ITlanguage_prag
Alan Zimmerman's avatar
Alan Zimmerman committed
587 588 589 590 591 592 593 594 595
  | 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
596

Ian Lynagh's avatar
Ian Lynagh committed
597
  | ITdotdot                    -- reserved symbols
598
  | ITcolon
599
  | ITdcolon            IsUnicodeSyntax
600 601
  | ITequal
  | ITlam
602
  | ITlcase
603
  | ITvbar
604 605
  | ITlarrow            IsUnicodeSyntax
  | ITrarrow            IsUnicodeSyntax
606 607
  | ITat
  | ITtilde
batterseapower's avatar
batterseapower committed
608
  | ITtildehsh
609
  | ITdarrow            IsUnicodeSyntax
610 611
  | ITminus
  | ITbang
612
  | ITstar              IsUnicodeSyntax
613 614
  | ITdot

Ian Lynagh's avatar
Ian Lynagh committed
615
  | ITbiglam                    -- GHC-extension symbols
616

Ian Lynagh's avatar
Ian Lynagh committed
617
  | ITocurly                    -- special symbols
618 619 620 621
  | ITccurly
  | ITvocurly
  | ITvccurly
  | ITobrack
Ian Lynagh's avatar
Ian Lynagh committed
622 623
  | ITopabrack                  -- [:, for parallel arrays with -XParallelArrays
  | ITcpabrack                  -- :], for parallel arrays with -XParallelArrays
624 625 626 627 628 629 630 631 632
  | ITcbrack
  | IToparen
  | ITcparen
  | IToubxparen
  | ITcubxparen
  | ITsemi
  | ITcomma
  | ITunderscore
  | ITbackquote
dreixel's avatar
dreixel committed
633
  | ITsimpleQuote               --  '
634

Ian Lynagh's avatar
Ian Lynagh committed
635
  | ITvarid   FastString        -- identifiers
636 637 638 639 640 641 642 643
  | 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
644
  | ITdupipvarid   FastString   -- GHC extension: implicit param: ?x
Adam Gundry's avatar
Adam Gundry committed
645
  | ITlabelvarid   FastString   -- Overloaded label: #x
646

647 648 649
  | 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
650
  | ITrational FractionalLit
651

652 653 654 655
  | 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
656 657
  | ITprimfloat  FractionalLit
  | ITprimdouble FractionalLit
658

659
  -- Template Haskell extension tokens
660
  | ITopenExpQuote HasE         --  [| or [e|
Ian Lynagh's avatar
Ian Lynagh committed
661 662 663 664
  | ITopenPatQuote              --  [p|
  | ITopenDecQuote              --  [d|
  | ITopenTypQuote              --  [t|
  | ITcloseQuote                --  |]
665
  | ITopenTExpQuote HasE        --  [|| or [e||
666
  | ITcloseTExpQuote            --  ||]
Ian Lynagh's avatar
Ian Lynagh committed
667 668
  | ITidEscape   FastString     --  $x
  | ITparenEscape               --  $(
669 670
  | ITidTyEscape   FastString   --  $$x
  | ITparenTyEscape             --  $$(
Ian Lynagh's avatar
Ian Lynagh committed
671
  | ITtyQuote                   --  ''
672 673 674 675 676 677 678 679
  | 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 |]
680 681 682 683

  -- Arrow notation extension
  | ITproc
  | ITrec
684 685 686 687 688 689
  | IToparenbar                  --  (|
  | ITcparenbar                  --  |)
  | ITlarrowtail IsUnicodeSyntax --  -<
  | ITrarrowtail IsUnicodeSyntax --  >-
  | ITLarrowtail IsUnicodeSyntax --  -<<
  | ITRarrowtail IsUnicodeSyntax --  >>-
690

691 692
  | ITunknown String             -- Used when the lexer can't make sense of it
  | ITeof                        -- end of file token
693 694 695 696 697 698 699

  -- 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
700
  | ITdocOptionsOld   String     -- doc options declared "-- # ..."-style
Jedai's avatar
Jedai committed
701 702
  | ITlineComment     String     -- comment starting by "--"
  | ITblockComment    String     -- comment in {- -}
703

704
  deriving Show
705

Alan Zimmerman's avatar
Alan Zimmerman committed
706 707 708
instance Outputable Token where
  ppr x = text (show x)

Alan Zimmerman's avatar
Alan Zimmerman committed
709

710 711 712 713 714 715 716
-- 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)
--
717
reservedWordsFM :: UniqFM (Token, ExtsBitmap)
718
reservedWordsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745
    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 ),

746 747
         ( "forall",         ITforall NormalSyntax,
                                              xbit ExplicitForallBit .|.
748 749
                                              xbit InRulePragBit),
         ( "mdo",            ITmdo,           xbit RecursiveDoBit),
750 751 752
             -- See Note [Lexing type pseudo-keywords]
         ( "family",         ITfamily,        0 ),
         ( "role",           ITrole,          0 ),
753
         ( "pattern",        ITpattern,       xbit PatternSynonymsBit),
754
         ( "static",         ITstatic,        0 ),
755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775
         ( "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),

         ( "rec",            ITrec,           xbit ArrowsBit .|.
                                              xbit RecursiveDoBit),
         ( "proc",           ITproc,          xbit ArrowsBit)
776 777
     ]

778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793
{-----------------------------------
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.
-------------------------------------}
794

795
reservedSymsFM :: UniqFM (Token, ExtsBitmap -> Bool)
796
reservedSymsFM = listToUFM $
Ian Lynagh's avatar
Ian Lynagh committed
797
    map (\ (x,y,z) -> (mkFastString x,(y,z)))
798
      [ ("..",  ITdotdot,              always)
Ian Lynagh's avatar
Ian Lynagh committed
799
        -- (:) is a reserved op, meaning only list cons
800 801 802 803 804 805 806 807 808 809 810 811 812
       ,(":",   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
813 814

        -- For data T (a::*) = MkT
815 816
       ,("*", ITstar NormalSyntax, always)
                                  -- \i -> kindSigsEnabled i || tyFamEnabled i)
Ian Lynagh's avatar
Ian Lynagh committed
817
        -- For 'forall a . t'
818
       ,(".", ITdot,  always) -- \i -> explicitForallEnabled i || inRulePrag i)
Ian Lynagh's avatar
Ian Lynagh committed
819

820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840
       ,("-<",  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)

       ,("★", ITstar UnicodeSyntax, unicodeSyntaxEnabled)
841

Simon Marlow's avatar
Simon Marlow committed
842 843 844
        -- 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).
845 846 847 848 849
       ]

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

Ian Lynagh's avatar
Ian Lynagh committed
850
type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
851

852
special :: Token -> Action
twanvl's avatar
twanvl committed
853
special tok span _buf _len = return (L span tok)
854

855
token, layout_token :: Token -> Action
twanvl's avatar
twanvl committed
856 857
token t span _buf _len = return (L span t)
layout_token t span _buf _len = pushLexState layout >> return (L span t)
858

859 860
idtoken :: (StringBuffer -> Int -> Token) -> Action
idtoken f span buf len = return (L span $! (f buf len))
861

862
skip_one_varid :: (FastString -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
863
skip_one_varid f span buf len
864
  = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
865

866 867 868 869
skip_two_varid :: (FastString -> Token) -> Action
skip_two_varid f span buf len
  = return (L span $! f (lexemeToFastString (stepOn (stepOn buf)) (len-2)))

870
strtoken :: (String -> Token) -> Action
Ian Lynagh's avatar
Ian Lynagh committed
871
strtoken f span buf len =
872
  return (L span $! (f $! lexemeToString buf len))
873

874
init_strtoken :: Int -> (String -> Token) -> Action
875
-- like strtoken, but drops the last N character(s)
Ian Lynagh's avatar
Ian Lynagh committed
876
init_strtoken drop f span buf len =
877
  return (L span $! (f $! lexemeToString buf (len-drop)))
878 879

begin :: Int -> Action
880
begin code _span _str _len = do pushLexState code; lexToken
881 882

pop :: Action
Ian Lynagh's avatar
Ian Lynagh committed
883 884
pop _span _buf _len = do _ <- popLexState
                         lexToken
885

886 887 888 889 890 891 892 893 894 895 896
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
897
              else failSpanMsgP (RealSrcSpan span) (text "Missing block")
898

899
pop_and :: Action -> Action
Ian Lynagh's avatar
Ian Lynagh committed
900 901
pop_and act span buf len = do _ <- popLexState
                              act span buf len
902

903
{-# INLINE nextCharIs #-}
Ian Lynagh's avatar
Ian Lynagh committed
904
nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
905 906
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)

907 908 909 910
{-# INLINE nextCharIsNot #-}
nextCharIsNot :: StringBuffer -> (Char -> Bool) -> Bool
nextCharIsNot buf p = not (nextCharIs buf p)

911
notFollowedBy :: Char -> AlexAccPred ExtsBitmap
Ian Lynagh's avatar
Ian Lynagh committed
912
notFollowedBy char _ _ _ (AI _ buf)
913
  = nextCharIsNot buf (== char)
914

915
notFollowedBySymbol :: AlexAccPred ExtsBitmap
916
notFollowedBySymbol _ _ _ (AI _ buf)
917
  = nextCharIsNot buf (`elem` "!#$%&*+./<=>?@\\^|-~")
918

Adam Gundry's avatar
Adam Gundry committed
919 920 921 922
followedByDigit :: AlexAccPred ExtsBitmap
followedByDigit _ _ _ (AI _ buf)
  = afterOptionalSpace buf (\b -> nextCharIs b (`elem` ['0'..'9']))

923 924 925 926 927
-- 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.
928
isNormalComment :: AlexAccPred ExtsBitmap
929
isNormalComment bits _ _ (AI _ buf)
930
  | haddockEnabled bits = notFollowedByDocOrPragma
931
  | otherwise           = nextCharIsNot buf (== '#')
932 933
  where
    notFollowedByDocOrPragma
934
       = afterOptionalSpace buf (\b -> nextCharIsNot b (`elem` "|^*$#"))
935

936 937 938 939 940
afterOptionalSpace :: StringBuffer -> (StringBuffer -> Bool) -> Bool
afterOptionalSpace buf p
    = if nextCharIs buf (== ' ')
      then p (snd (nextChar buf))
      else p buf
941

942
atEOL :: AlexAccPred ExtsBitmap
943
atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
944

945
ifExtension :: (ExtsBitmap -> Bool) -> AlexAccPred ExtsBitmap
946 947
ifExtension pred bits _ _ _ = pred bits

948 949 950
multiline_doc_comment :: Action
multiline_doc_comment span buf _len = withLexedDocType (worker "")
  where
951
    worker commentAcc input docType oneLine = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
952
      Just ('\n', input')
953 954 955 956 957 958
        | 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
959

960 961
    checkIfCommentLine input = check (dropNonNewlineSpace input)
      where
962 963 964
        check input = case alexGetChar' input of
          Just ('-', input) -> case alexGetChar' input of
            Just ('-', input) -> case alexGetChar' input of
965 966 967 968 969
              Just (c, _) | c /= '-' -> Just input
              _ -> Nothing
            _ -> Nothing
          _ -> Nothing

970
        dropNonNewlineSpace input = case alexGetChar' input of
Ian Lynagh's avatar
Ian Lynagh committed
971
          Just (c, input')
972 973 974 975
            | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
            | otherwise -> input
          Nothing -> input

Jedai's avatar
Jedai committed
976 977 978 979 980
lineCommentToken :: Action
lineCommentToken span buf len = do
  b <- extension rawTokenStreamEnabled
  if b then strtoken ITlineComment span buf len else lexToken

981 982 983 984
{-
  nested comments require traversing by hand, they can't be parsed
  using regular expressions.
-}
Ian Lynagh's avatar
Ian Lynagh committed
985
nested_comment :: P (RealLocated Token) -> Action
Alan Zimmerman's avatar
Alan Zimmerman committed
986
nested_comment cont span buf len = do
987
  input <- getInput
988
  go (reverse $ lexemeToString buf len) (1::Int) input
989
  where
Alan Zimmerman's avatar
Alan Zimmerman committed
990 991 992 993 994 995
    go commentAcc 0 input = do
      setInput input
      b <- extension rawTokenStreamEnabled
      if b
        then docCommentEnd input commentAcc ITblockComment buf span
        else cont
996
    go commentAcc n input = case alexGetChar' input of
997
      Nothing -> errBrace input span
998
      Just ('-',input) -> case alexGetChar' input of
999
        Nothing  -> errBrace input span
1000
        Just ('\125',input) -> go ('\125':'-':commentAcc) (n-1) input -- '}'
Jedai's avatar
Jedai committed
1001
        Just (_,_)          -> go ('-':commentAcc) n input
1002
      Just ('\123',input) -> case alexGetChar' input of  -- '{' char
1003
        Nothing  -> errBrace input span
Jedai's avatar
Jedai committed
1004 1005 1006
        Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
        Just (_,_)       -> go ('\123':commentAcc) n input
      Just (c,input) -> go (c:commentAcc) n input
1007 1008 1009 1010

nested_doc_comment :: Action
nested_doc_comment span buf _len = withLexedDocType (go "")
  where
1011
    go commentAcc input docType _ = case alexGetChar' input of
1012
      Nothing -> errBrace input span
1013
      Just ('-',input) -> case alexGetChar' input of
1014
        Nothing -> errBrace input span
twanvl's avatar
twanvl committed
1015
        Just ('\125',input) ->
1016
          docCommentEnd input commentAcc docType buf span
twanvl's avatar
twanvl committed
1017
        Just (_,_) -> go ('-':commentAcc) input docType False
1018
      Just ('\123', input) -> case alexGetChar' input of
1019 1020 1021 1022 1023
        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
1024
        Just (_,_) -> go ('\123':commentAcc) input docType False
1025 1026
      Just (c,input) -> go (c:commentAcc) input docType False

Ian Lynagh's avatar
Ian Lynagh committed
1027 1028
withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
                 -> P (RealLocated Token)
1029
withLexedDocType lexDocComment = do
1030
  input@(AI _ buf) <- getInput
1031 1032 1033
  case prevChar buf ' ' of
    '|' -> lexDocComment input ITdocCommentNext False
    '^' -> lexDocComment input ITdocCommentPrev False
1034
    '$' -> lexDocComment input ITdocCommentNamed True
David Waern's avatar
David Waern committed
1035 1036
    '*' -> lexDocSection 1 input
    '#' -> lexDocComment input ITdocOptionsOld False
Ian Lynagh's avatar
Ian Lynagh committed
1037
    _ -> panic "withLexedDocType: Bad doc type"
Ian Lynagh's avatar
Ian Lynagh committed
1038
 where
1039
    lexDocSection n input = case alexGetChar' input of
1040
      Just ('*', input) -> lexDocSection (n+1) input
twanvl's avatar
twanvl committed
1041
      Just (_,   _)     -> lexDocComment input (ITdocSection n) True
1042 1043
      Nothing -> do setInput input; lexToken -- eof reached, lex it normally

1044 1045 1046
-- RULES pragmas turn on the forall and '.' keywords, and we turn them
-- off again at the end of the pragma.
rulePrag :: Action
Alan Zimmerman's avatar
Alan Zimmerman committed
1047
rulePrag span buf len = do
1048
  setExts (.|. xbit InRulePragBit)
Alan Zimmerman's avatar
Alan Zimmerman committed
1049 1050
  let !src = lexemeToString buf len
  return (L span (ITrules_prag src))
1051 1052

endPrag :: Action
1053
endPrag span _buf _len = do
1054
  setExts (.&. complement (xbit InRulePragBit))
1055 1056
  return (L span ITclose_prag)

1057 1058 1059 1060
-- 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
1061
-- than what was lexed by Alex, and the lexToken function doesn't know this, so
1062
-- it writes the wrong token length to the parser state. This function is
Ian Lynagh's avatar
Ian Lynagh committed
1063
-- called afterwards, so it can just update the state.
1064 1065

docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
Ian Lynagh's avatar
Ian Lynagh committed
1066
                 RealSrcSpan -> P (RealLocated Token)
1067 1068
docCommentEnd input commentAcc docType buf span = do
  setInput input
1069
  let (AI loc nextBuf) = input
1070
      comment = reverse commentAcc
Ian Lynagh's avatar
Ian Lynagh committed
1071
      span' = mkRealSrcSpan (realSrcSpanStart span) loc
1072
      last_len = byteDiff buf nextBuf
Ian Lynagh's avatar
Ian Lynagh committed
1073

1074
  span `seq` setLastToken span' last_len
1075
  return (L span' (docType comment))
Ian Lynagh's avatar
Ian Lynagh committed
1076

Ian Lynagh's avatar
Ian Lynagh committed
1077 1078
errBrace :: AlexInput -> RealSrcSpan -> P a
errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
Ian Lynagh's avatar
Ian Lynagh committed
1079

1080
open_brace, close_brace :: Action
Ian Lynagh's avatar
Ian Lynagh committed
1081
open_brace span _str _len = do
1082 1083
  ctx <- getContext
  setContext (NoLayout:ctx)
1084
  return (L span ITocurly)
Ian Lynagh's avatar
Ian Lynagh committed
1085
close_brace span _str _len = do
1086
  popContext
1087
  return (L span ITccurly)
1088

Ian Lynagh's avatar
Ian Lynagh committed
1089
qvarid, qconid :: StringBuffer -> Int -> Token
1090 1091
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
1092

1093
splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
1094 1095 1096
-- 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.
1097
splitQualName orig_buf len parens = split orig_buf orig_buf
1098
  where
1099
    split buf dot_buf
Ian Lynagh's avatar
Ian Lynagh committed
1100 1101 1102
        | orig_buf `byteDiff` buf >= len  = done dot_buf
        | c == '.'                        = found_dot buf'
        | otherwise                       = split buf' dot_buf
1103 1104
      where
       (c,buf') = nextChar buf
Ian Lynagh's avatar
Ian Lynagh committed
1105

1106 1107 1108
    -- careful, we might get names like M....
    -- so, if the character after the dot is not upper-case, this is
    -- the end of the qualifier part.
1109
    found_dot buf -- buf points after the '.'
Ian Lynagh's avatar
Ian Lynagh committed
1110 1111
        | isUpper c    = split buf' buf
        | otherwise    = done buf
1112 1113 1114 1115
      where
       (c,buf') = nextChar buf

    done dot_buf =