1 -----------------------------------------------------------------------------
2 -- (c) The University of Glasgow, 2006
6 -- This is a combination of an Alex-generated lexer from a regex
7 -- definition, with some hand-coded bits.
9 -- Completely accurate information about token-spans within the source
10 -- file is maintained. Every token has a start and end SrcLoc attached to it.
12 -----------------------------------------------------------------------------
16 -- - parsing integers is a bit slow
17 -- - readRational is a bit slow
19 -- Known bugs, that were also in the previous version:
20 -- - M... should be 3 tokens, not 1.
21 -- - pragma-end should be only valid in a pragma
25 -- The above warning supression flag is a temporary kludge.
26 -- While working on this module you are encouraged to remove it and fix
27 -- any warnings in the module. See
28 -- http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings
31 -- Note that Alex itself generates code with with some unused bindings and
32 -- without type signatures, so removing the flag might not be possible.
34 {-# OPTIONS_GHC -funbox-strict-fields #-}
37 Token(..), lexer, pragState, mkPState, PState(..),
38 P(..), ParseResult(..), getSrcLoc,
39 failLocMsgP, failSpanMsgP, srcParseFail,
41 popContext, pushCurrentContext, setLastToken, setSrcLoc,
42 getLexState, popLexState, pushLexState,
43 extension, standaloneDerivingEnabled, bangPatEnabled,
57 import Util ( maybePrefixMatch, readRational )
61 import Data.Char ( chr, ord, isSpace )
65 import Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper )
68 $unispace = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
69 $whitechar = [\ \n\r\f\v $unispace]
70 $white_no_nl = $whitechar # \n
74 $unidigit = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
75 $decdigit = $ascdigit -- for now, should really be $digit (ToDo)
76 $digit = [$ascdigit $unidigit]
78 $special = [\(\)\,\;\[\]\`\{\}]
79 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
80 $unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
81 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
83 $unilarge = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
85 $large = [$asclarge $unilarge]
87 $unismall = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
89 $small = [$ascsmall $unismall \_]
91 $unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
92 $graphic = [$small $large $symbol $digit $special $unigraphic \:\"\']
95 $hexit = [$decdigit A-F a-f]
96 $symchar = [$symbol \:]
98 $idchar = [$small $large $digit \']
100 $docsym = [\| \^ \* \$]
102 @varid = $small $idchar*
103 @conid = $large $idchar*
105 @varsym = $symbol $symchar*
106 @consym = \: $symchar*
108 @decimal = $decdigit+
110 @hexadecimal = $hexit+
111 @exponent = [eE] [\-\+]? @decimal
113 -- we support the hierarchical module name extension:
116 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
118 -- normal signed numerical literals can only be explicitly negative,
119 -- not explicitly positive (contrast @exponent)
121 @signed = @negative ?
125 -- everywhere: skip whitespace and comments
127 $tab+ { warn Opt_WarnTabs (text "Tab character") }
129 -- Everywhere: deal with nested comments. We explicitly rule out
130 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
131 -- (this can happen even though pragmas will normally take precedence due to
132 -- longest-match, because pragmas aren't valid in every state, but comments
133 -- are). We also rule out nested Haddock comments, if the -haddock flag is
136 "{-" / { isNormalComment } { nested_comment lexToken }
138 -- Single-line comments are a bit tricky. Haskell 98 says that two or
139 -- more dashes followed by a symbol should be parsed as a varsym, so we
140 -- have to exclude those.
142 -- Since Haddock comments aren't valid in every state, we need to rule them
145 -- The following two rules match comments that begin with two dashes, but
146 -- continue with a different character. The rules test that this character
147 -- is not a symbol (in which case we'd have a varsym), and that it's not a
148 -- space followed by a Haddock comment symbol (docsym) (in which case we'd
149 -- have a Haddock comment). The rules then munch the rest of the line.
151 "-- " ~[$docsym \#] .* ;
152 "--" [^$symbol : \ ] .* ;
154 -- Next, match Haddock comments if no -haddock flag
156 "-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } ;
158 -- Now, when we've matched comments that begin with 2 dashes and continue
159 -- with a different character, we need to match comments that begin with three
160 -- or more dashes (which clearly can't be Haddock comments). We only need to
161 -- make sure that the first non-dash character isn't a symbol, and munch the
164 "---"\-* [^$symbol :] .* ;
166 -- Since the previous rules all match dashes followed by at least one
167 -- character, we also need to match a whole line filled with just dashes.
169 "--"\-* / { atEOL } ;
171 -- We need this rule since none of the other single line comment rules
172 -- actually match this case.
176 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
177 -- blank lines) until we find a non-whitespace character, then do layout
180 -- One slight wibble here: what if the line begins with {-#? In
181 -- theory, we have to lex the pragma to see if it's one we recognise,
182 -- and if it is, then we backtrack and do_bol, otherwise we treat it
183 -- as a nested comment. We don't bother with this: if the line begins
184 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
187 ^\# (line)? { begin line_prag1 }
188 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
189 ^\# \! .* \n ; -- #!, for scripts
193 -- after a layout keyword (let, where, do, of), we begin a new layout
194 -- context if the curly brace is missing.
195 -- Careful! This stuff is quite delicate.
196 <layout, layout_do> {
197 \{ / { notFollowedBy '-' } { pop_and open_brace }
198 -- we might encounter {-# here, but {- has been handled already
200 ^\# (line)? { begin line_prag1 }
203 -- do is treated in a subtly different way, see new_layout_context
204 <layout> () { new_layout_context True }
205 <layout_do> () { new_layout_context False }
207 -- after a new layout context which was found to be to the left of the
208 -- previous context, we have generated a '{' token, and we now need to
209 -- generate a matching '}' token.
210 <layout_left> () { do_layout_left }
212 <0,option_prags> \n { begin bol }
214 "{-#" $whitechar* (line|LINE) { begin line_prag2 }
216 -- single-line line pragmas, of the form
217 -- # <line> "<file>" <extra-stuff> \n
218 <line_prag1> $decdigit+ { setLine line_prag1a }
219 <line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
220 <line_prag1b> .* { pop }
222 -- Haskell-style line pragmas, of the form
223 -- {-# LINE <line> "<file>" #-}
224 <line_prag2> $decdigit+ { setLine line_prag2a }
225 <line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
226 <line_prag2b> "#-}"|"-}" { pop }
227 -- NOTE: accept -} at the end of a LINE pragma, for compatibility
228 -- with older versions of GHC which generated these.
231 "{-#" $whitechar* (RULES|rules) { token ITrules_prag }
232 "{-#" $whitechar* (INLINE|inline) { token (ITinline_prag True) }
233 "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
234 { token (ITinline_prag False) }
235 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
236 { token ITspec_prag }
237 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
238 $whitechar* (INLINE|inline) { token (ITspec_inline_prag True) }
239 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
240 $whitechar* (NO(T?)INLINE|no(t?)inline)
241 { token (ITspec_inline_prag False) }
242 "{-#" $whitechar* (SOURCE|source) { token ITsource_prag }
243 "{-#" $whitechar* (WARNING|warning)
244 { token ITwarning_prag }
245 "{-#" $whitechar* (DEPRECATED|deprecated)
246 { token ITdeprecated_prag }
247 "{-#" $whitechar* (SCC|scc) { token ITscc_prag }
248 "{-#" $whitechar* (GENERATED|generated)
249 { token ITgenerated_prag }
250 "{-#" $whitechar* (CORE|core) { token ITcore_prag }
251 "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag }
253 -- We ignore all these pragmas, but don't generate a warning for them
254 -- CFILES is a hugs-only thing.
255 "{-#" $whitechar* (OPTIONS_HUGS|options_hugs|OPTIONS_NHC98|options_nhc98|OPTIONS_JHC|options_jhc|CFILES|cfiles)
256 { nested_comment lexToken }
258 -- ToDo: should only be valid inside a pragma:
263 "{-#" $whitechar* (OPTIONS|options) { lex_string_prag IToptions_prag }
264 "{-#" $whitechar* (OPTIONS_GHC|options_ghc)
265 { lex_string_prag IToptions_prag }
266 "{-#" $whitechar* (OPTIONS_HADDOCK|options_haddock)
267 { lex_string_prag ITdocOptions }
268 "-- #" { multiline_doc_comment }
269 "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
270 "{-#" $whitechar* (INCLUDE|include) { lex_string_prag ITinclude_prag }
274 -- In the "0" mode we ignore these pragmas
275 "{-#" $whitechar* (OPTIONS|options|OPTIONS_GHC|options_ghc|OPTIONS_HADDOCK|options_haddock|LANGUAGE|language|INCLUDE|include)
276 { nested_comment lexToken }
284 "{-#" { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
285 (nested_comment lexToken) }
288 -- '0' state: ordinary lexemes
293 "-- " $docsym / { ifExtension haddockEnabled } { multiline_doc_comment }
294 "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
300 "[:" / { ifExtension parrEnabled } { token ITopabrack }
301 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
305 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
306 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
307 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
308 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
309 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
310 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
311 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
312 "$(" / { ifExtension thEnabled } { token ITparenEscape }
314 "[$" @varid "|" / { ifExtension qqEnabled }
315 { lex_quasiquote_tok }
319 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
320 { special IToparenbar }
321 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
325 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
329 "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
330 { token IToubxparen }
331 "#)" / { ifExtension unboxedTuplesEnabled }
332 { token ITcubxparen }
336 "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
337 "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
341 \( { special IToparen }
342 \) { special ITcparen }
343 \[ { special ITobrack }
344 \] { special ITcbrack }
345 \, { special ITcomma }
346 \; { special ITsemi }
347 \` { special ITbackquote }
354 @qual @varid { idtoken qvarid }
355 @qual @conid { idtoken qconid }
357 @conid { idtoken conid }
361 @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
362 @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
363 @varid "#"+ / { ifExtension magicHashEnabled } { varid }
364 @conid "#"+ / { ifExtension magicHashEnabled } { idtoken conid }
370 @qual @varsym { idtoken qvarsym }
371 @qual @consym { idtoken qconsym }
376 -- For the normal boxed literals we need to be careful
377 -- when trying to be close to Haskell98
379 -- Normal integral literals (:: Num a => a, from Integer)
380 @decimal { tok_num positive 0 0 decimal }
381 0[oO] @octal { tok_num positive 2 2 octal }
382 0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal }
384 -- Normal rational literals (:: Fractional a => a, from Rational)
385 @floating_point { strtoken tok_float }
389 -- Unboxed ints (:: Int#) and words (:: Word#)
390 -- It's simpler (and faster?) to give separate cases to the negatives,
391 -- especially considering octal/hexadecimal prefixes.
392 @decimal \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
393 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
394 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
395 @negative @decimal \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
396 @negative 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
397 @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
399 @decimal \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
400 0[oO] @octal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
401 0[xX] @hexadecimal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }
403 -- Unboxed floats and doubles (:: Float#, :: Double#)
404 -- prim_{float,double} work with signed literals
405 @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
406 @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
409 -- Strings and chars are lexed by hand-written code. The reason is
410 -- that even if we recognise the string or char here in the regex
411 -- lexer, we would still have to parse the string afterward in order
412 -- to convert it to a String.
415 \" { lex_string_tok }
419 -- -----------------------------------------------------------------------------
423 = ITas -- Haskell keywords
447 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
449 | ITforall -- GHC extension keywords
467 | ITinline_prag Bool -- True <=> INLINE, False <=> NOINLINE
468 | ITspec_prag -- SPECIALISE
469 | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE)
477 | ITcore_prag -- hdaume: core annotations
480 | IToptions_prag String
481 | ITinclude_prag String
484 | ITdotdot -- reserved symbols
500 | ITbiglam -- GHC-extension symbols
502 | ITocurly -- special symbols
504 | ITocurlybar -- {|, for type applications
505 | ITccurlybar -- |}, for type applications
509 | ITopabrack -- [:, for parallel arrays with -XParr
510 | ITcpabrack -- :], for parallel arrays with -XParr
521 | ITvarid FastString -- identifiers
523 | ITvarsym FastString
524 | ITconsym FastString
525 | ITqvarid (FastString,FastString)
526 | ITqconid (FastString,FastString)
527 | ITqvarsym (FastString,FastString)
528 | ITqconsym (FastString,FastString)
530 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
532 | ITpragma StringBuffer
535 | ITstring FastString
537 | ITrational Rational
540 | ITprimstring FastString
543 | ITprimfloat Rational
544 | ITprimdouble Rational
546 -- MetaHaskell extension tokens
547 | ITopenExpQuote -- [| or [e|
548 | ITopenPatQuote -- [p|
549 | ITopenDecQuote -- [d|
550 | ITopenTypQuote -- [t|
552 | ITidEscape FastString -- $x
553 | ITparenEscape -- $(
556 | ITquasiQuote (FastString,FastString,SrcSpan) -- [:...|...|]
558 -- Arrow notation extension
565 | ITLarrowtail -- -<<
566 | ITRarrowtail -- >>-
568 | ITunknown String -- Used when the lexer can't make sense of it
569 | ITeof -- end of file token
571 -- Documentation annotations
572 | ITdocCommentNext String -- something beginning '-- |'
573 | ITdocCommentPrev String -- something beginning '-- ^'
574 | ITdocCommentNamed String -- something beginning '-- $'
575 | ITdocSection Int String -- a section heading
576 | ITdocOptions String -- doc options (prune, ignore-exports, etc)
577 | ITdocOptionsOld String -- doc options declared "-- # ..."-style
580 deriving Show -- debugging
584 isSpecial :: Token -> Bool
585 -- If we see M.x, where x is a keyword, but
586 -- is special, we treat is as just plain M.x,
588 isSpecial ITas = True
589 isSpecial IThiding = True
590 isSpecial ITqualified = True
591 isSpecial ITforall = True
592 isSpecial ITexport = True
593 isSpecial ITlabel = True
594 isSpecial ITdynamic = True
595 isSpecial ITsafe = True
596 isSpecial ITthreadsafe = True
597 isSpecial ITunsafe = True
598 isSpecial ITccallconv = True
599 isSpecial ITstdcallconv = True
600 isSpecial ITmdo = True
601 isSpecial ITfamily = True
602 isSpecial ITgroup = True
603 isSpecial ITby = True
604 isSpecial ITusing = True
608 -- the bitmap provided as the third component indicates whether the
609 -- corresponding extension keyword is valid under the extension options
610 -- provided to the compiler; if the extension corresponding to *any* of the
611 -- bits set in the bitmap is enabled, the keyword is valid (this setup
612 -- facilitates using a keyword in two different extensions that can be
613 -- activated independently)
615 reservedWordsFM = listToUFM $
616 map (\(x, y, z) -> (mkFastString x, (y, z)))
617 [( "_", ITunderscore, 0 ),
619 ( "case", ITcase, 0 ),
620 ( "class", ITclass, 0 ),
621 ( "data", ITdata, 0 ),
622 ( "default", ITdefault, 0 ),
623 ( "deriving", ITderiving, 0 ),
625 ( "else", ITelse, 0 ),
626 ( "hiding", IThiding, 0 ),
628 ( "import", ITimport, 0 ),
630 ( "infix", ITinfix, 0 ),
631 ( "infixl", ITinfixl, 0 ),
632 ( "infixr", ITinfixr, 0 ),
633 ( "instance", ITinstance, 0 ),
635 ( "module", ITmodule, 0 ),
636 ( "newtype", ITnewtype, 0 ),
638 ( "qualified", ITqualified, 0 ),
639 ( "then", ITthen, 0 ),
640 ( "type", ITtype, 0 ),
641 ( "where", ITwhere, 0 ),
642 ( "_scc_", ITscc, 0 ), -- ToDo: remove
644 ( "forall", ITforall, bit explicitForallBit .|. bit inRulePragBit),
645 ( "mdo", ITmdo, bit recursiveDoBit),
646 ( "family", ITfamily, bit tyFamBit),
647 ( "group", ITgroup, bit transformComprehensionsBit),
648 ( "by", ITby, bit transformComprehensionsBit),
649 ( "using", ITusing, bit transformComprehensionsBit),
651 ( "foreign", ITforeign, bit ffiBit),
652 ( "export", ITexport, bit ffiBit),
653 ( "label", ITlabel, bit ffiBit),
654 ( "dynamic", ITdynamic, bit ffiBit),
655 ( "safe", ITsafe, bit ffiBit),
656 ( "threadsafe", ITthreadsafe, bit ffiBit),
657 ( "unsafe", ITunsafe, bit ffiBit),
658 ( "stdcall", ITstdcallconv, bit ffiBit),
659 ( "ccall", ITccallconv, bit ffiBit),
660 ( "dotnet", ITdotnet, bit ffiBit),
662 ( "rec", ITrec, bit arrowsBit),
663 ( "proc", ITproc, bit arrowsBit)
666 reservedSymsFM :: UniqFM (Token, Int -> Bool)
667 reservedSymsFM = listToUFM $
668 map (\ (x,y,z) -> (mkFastString x,(y,z)))
669 [ ("..", ITdotdot, always)
670 -- (:) is a reserved op, meaning only list cons
671 ,(":", ITcolon, always)
672 ,("::", ITdcolon, always)
673 ,("=", ITequal, always)
674 ,("\\", ITlam, always)
675 ,("|", ITvbar, always)
676 ,("<-", ITlarrow, always)
677 ,("->", ITrarrow, always)
679 ,("~", ITtilde, always)
680 ,("=>", ITdarrow, always)
681 ,("-", ITminus, always)
682 ,("!", ITbang, always)
684 -- For data T (a::*) = MkT
685 ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i)
686 -- For 'forall a . t'
687 ,(".", ITdot, \i -> explicitForallEnabled i || inRulePrag i)
689 ,("-<", ITlarrowtail, arrowsEnabled)
690 ,(">-", ITrarrowtail, arrowsEnabled)
691 ,("-<<", ITLarrowtail, arrowsEnabled)
692 ,(">>-", ITRarrowtail, arrowsEnabled)
694 #if __GLASGOW_HASKELL__ >= 605
695 ,("∷", ITdcolon, unicodeSyntaxEnabled)
696 ,("⇒", ITdarrow, unicodeSyntaxEnabled)
697 ,("∀", ITforall, \i -> unicodeSyntaxEnabled i &&
698 explicitForallEnabled i)
699 ,("→", ITrarrow, unicodeSyntaxEnabled)
700 ,("←", ITlarrow, unicodeSyntaxEnabled)
701 ,("⋯", ITdotdot, unicodeSyntaxEnabled)
702 -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
703 -- form part of a large operator. This would let us have a better
704 -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
708 -- -----------------------------------------------------------------------------
711 type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
713 special :: Token -> Action
714 special tok span _buf _len = return (L span tok)
716 token, layout_token :: Token -> Action
717 token t span _buf _len = return (L span t)
718 layout_token t span _buf _len = pushLexState layout >> return (L span t)
720 idtoken :: (StringBuffer -> Int -> Token) -> Action
721 idtoken f span buf len = return (L span $! (f buf len))
723 skip_one_varid :: (FastString -> Token) -> Action
724 skip_one_varid f span buf len
725 = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
727 strtoken :: (String -> Token) -> Action
728 strtoken f span buf len =
729 return (L span $! (f $! lexemeToString buf len))
731 init_strtoken :: Int -> (String -> Token) -> Action
732 -- like strtoken, but drops the last N character(s)
733 init_strtoken drop f span buf len =
734 return (L span $! (f $! lexemeToString buf (len-drop)))
736 begin :: Int -> Action
737 begin code _span _str _len = do pushLexState code; lexToken
740 pop _span _buf _len = do popLexState; lexToken
742 pop_and :: Action -> Action
743 pop_and act span buf len = do popLexState; act span buf len
745 {-# INLINE nextCharIs #-}
746 nextCharIs buf p = not (atEnd buf) && p (currentChar buf)
748 notFollowedBy char _ _ _ (AI _ _ buf)
749 = nextCharIs buf (/=char)
751 notFollowedBySymbol _ _ _ (AI _ _ buf)
752 = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
754 -- We must reject doc comments as being ordinary comments everywhere.
755 -- In some cases the doc comment will be selected as the lexeme due to
756 -- maximal munch, but not always, because the nested comment rule is
757 -- valid in all states, but the doc-comment rules are only valid in
758 -- the non-layout states.
759 isNormalComment bits _ _ (AI _ _ buf)
760 | haddockEnabled bits = notFollowedByDocOrPragma
761 | otherwise = nextCharIs buf (/='#')
763 notFollowedByDocOrPragma
764 = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))
766 spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
769 haddockDisabledAnd p bits _ _ (AI _ _ buf)
770 = if haddockEnabled bits then False else (p buf)
773 atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
775 ifExtension pred bits _ _ _ = pred bits
777 multiline_doc_comment :: Action
778 multiline_doc_comment span buf _len = withLexedDocType (worker "")
780 worker commentAcc input docType oneLine = case alexGetChar input of
782 | oneLine -> docCommentEnd input commentAcc docType buf span
783 | otherwise -> case checkIfCommentLine input' of
784 Just input -> worker ('\n':commentAcc) input docType False
785 Nothing -> docCommentEnd input commentAcc docType buf span
786 Just (c, input) -> worker (c:commentAcc) input docType oneLine
787 Nothing -> docCommentEnd input commentAcc docType buf span
789 checkIfCommentLine input = check (dropNonNewlineSpace input)
791 check input = case alexGetChar input of
792 Just ('-', input) -> case alexGetChar input of
793 Just ('-', input) -> case alexGetChar input of
794 Just (c, _) | c /= '-' -> Just input
799 dropNonNewlineSpace input = case alexGetChar input of
801 | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
806 nested comments require traversing by hand, they can't be parsed
807 using regular expressions.
809 nested_comment :: P (Located Token) -> Action
810 nested_comment cont span _str _len = do
814 go 0 input = do setInput input; cont
815 go n input = case alexGetChar input of
816 Nothing -> errBrace input span
817 Just ('-',input) -> case alexGetChar input of
818 Nothing -> errBrace input span
819 Just ('\125',input) -> go (n-1) input
820 Just (_,_) -> go n input
821 Just ('\123',input) -> case alexGetChar input of
822 Nothing -> errBrace input span
823 Just ('-',input) -> go (n+1) input
824 Just (_,_) -> go n input
825 Just (_,input) -> go n input
827 nested_doc_comment :: Action
828 nested_doc_comment span buf _len = withLexedDocType (go "")
830 go commentAcc input docType _ = case alexGetChar input of
831 Nothing -> errBrace input span
832 Just ('-',input) -> case alexGetChar input of
833 Nothing -> errBrace input span
834 Just ('\125',input) ->
835 docCommentEnd input commentAcc docType buf span
836 Just (_,_) -> go ('-':commentAcc) input docType False
837 Just ('\123', input) -> case alexGetChar input of
838 Nothing -> errBrace input span
839 Just ('-',input) -> do
841 let cont = do input <- getInput; go commentAcc input docType False
842 nested_comment cont span buf _len
843 Just (_,_) -> go ('\123':commentAcc) input docType False
844 Just (c,input) -> go (c:commentAcc) input docType False
846 withLexedDocType lexDocComment = do
847 input@(AI _ _ buf) <- getInput
848 case prevChar buf ' ' of
849 '|' -> lexDocComment input ITdocCommentNext False
850 '^' -> lexDocComment input ITdocCommentPrev False
851 '$' -> lexDocComment input ITdocCommentNamed False
852 '*' -> lexDocSection 1 input
853 '#' -> lexDocComment input ITdocOptionsOld False
855 lexDocSection n input = case alexGetChar input of
856 Just ('*', input) -> lexDocSection (n+1) input
857 Just (_, _) -> lexDocComment input (ITdocSection n) True
858 Nothing -> do setInput input; lexToken -- eof reached, lex it normally
860 -- RULES pragmas turn on the forall and '.' keywords, and we turn them
861 -- off again at the end of the pragma.
863 rulePrag span buf len = do
864 setExts (.|. inRulePragBit)
865 return (L span ITrules_prag)
868 endPrag span buf len = do
869 setExts (.&. complement (bit inRulePragBit))
870 return (L span ITclose_prag)
873 -------------------------------------------------------------------------------
874 -- This function is quite tricky. We can't just return a new token, we also
875 -- need to update the state of the parser. Why? Because the token is longer
876 -- than what was lexed by Alex, and the lexToken function doesn't know this, so
877 -- it writes the wrong token length to the parser state. This function is
878 -- called afterwards, so it can just update the state.
880 -- This is complicated by the fact that Haddock tokens can span multiple lines,
881 -- which is something that the original lexer didn't account for.
882 -- I have added last_line_len in the parser state which represents the length
883 -- of the part of the token that is on the last line. It is now used for layout
884 -- calculation in pushCurrentContext instead of last_len. last_len is, like it
885 -- was before, the full length of the token, and it is now only used for error
888 docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
889 SrcSpan -> P (Located Token)
890 docCommentEnd input commentAcc docType buf span = do
892 let (AI loc last_offs nextBuf) = input
893 comment = reverse commentAcc
894 span' = mkSrcSpan (srcSpanStart span) loc
895 last_len = byteDiff buf nextBuf
897 last_line_len = if (last_offs - last_len < 0)
901 span `seq` setLastToken span' last_len last_line_len
902 return (L span' (docType comment))
904 errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
906 open_brace, close_brace :: Action
907 open_brace span _str _len = do
909 setContext (NoLayout:ctx)
910 return (L span ITocurly)
911 close_brace span _str _len = do
913 return (L span ITccurly)
915 qvarid buf len = ITqvarid $! splitQualName buf len
916 qconid buf len = ITqconid $! splitQualName buf len
918 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
919 -- takes a StringBuffer and a length, and returns the module name
920 -- and identifier parts of a qualified name. Splits at the *last* dot,
921 -- because of hierarchical module names.
922 splitQualName orig_buf len = split orig_buf orig_buf
925 | orig_buf `byteDiff` buf >= len = done dot_buf
926 | c == '.' = found_dot buf'
927 | otherwise = split buf' dot_buf
929 (c,buf') = nextChar buf
931 -- careful, we might get names like M....
932 -- so, if the character after the dot is not upper-case, this is
933 -- the end of the qualifier part.
934 found_dot buf -- buf points after the '.'
935 | isUpper c = split buf' buf
936 | otherwise = done buf
938 (c,buf') = nextChar buf
941 (lexemeToFastString orig_buf (qual_size - 1),
942 lexemeToFastString dot_buf (len - qual_size))
944 qual_size = orig_buf `byteDiff` dot_buf
948 case lookupUFM reservedWordsFM fs of
949 Just (keyword,0) -> do
951 return (L span keyword)
952 Just (keyword,exts) -> do
953 b <- extension (\i -> exts .&. i /= 0)
954 if b then do maybe_layout keyword
955 return (L span keyword)
956 else return (L span (ITvarid fs))
957 _other -> return (L span (ITvarid fs))
959 fs = lexemeToFastString buf len
961 conid buf len = ITconid fs
962 where fs = lexemeToFastString buf len
964 qvarsym buf len = ITqvarsym $! splitQualName buf len
965 qconsym buf len = ITqconsym $! splitQualName buf len
967 varsym = sym ITvarsym
968 consym = sym ITconsym
970 sym con span buf len =
971 case lookupUFM reservedSymsFM fs of
972 Just (keyword,exts) -> do
974 if b then return (L span keyword)
975 else return (L span $! con fs)
976 _other -> return (L span $! con fs)
978 fs = lexemeToFastString buf len
980 -- Variations on the integral numeric literal.
981 tok_integral :: (Integer -> Token)
982 -> (Integer -> Integer)
983 -- -> (StringBuffer -> StringBuffer) -> (Int -> Int)
985 -> (Integer, (Char->Int)) -> Action
986 tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
987 return $ L span $ itint $! transint $ parseUnsignedInteger
988 (offsetBytes transbuf buf) (subtract translen len) radix char_to_int
990 -- some conveniences for use with tok_integral
991 tok_num = tok_integral ITinteger
992 tok_primint = tok_integral ITprimint
993 tok_primword = tok_integral ITprimword positive
996 decimal = (10,octDecDigit)
997 octal = (8,octDecDigit)
998 hexadecimal = (16,hexDigit)
1000 -- readRational can understand negative rationals, exponents, everything.
1001 tok_float str = ITrational $! readRational str
1002 tok_primfloat str = ITprimfloat $! readRational str
1003 tok_primdouble str = ITprimdouble $! readRational str
1005 -- -----------------------------------------------------------------------------
1006 -- Layout processing
1008 -- we're at the first token on a line, insert layout tokens if necessary
1010 do_bol span _str _len = do
1014 --trace "layout: inserting '}'" $ do
1016 -- do NOT pop the lex state, we might have a ';' to insert
1017 return (L span ITvccurly)
1019 --trace "layout: inserting ';'" $ do
1021 return (L span ITsemi)
1026 -- certain keywords put us in the "layout" state, where we might
1027 -- add an opening curly brace.
1028 maybe_layout ITdo = pushLexState layout_do
1029 maybe_layout ITmdo = pushLexState layout_do
1030 maybe_layout ITof = pushLexState layout
1031 maybe_layout ITlet = pushLexState layout
1032 maybe_layout ITwhere = pushLexState layout
1033 maybe_layout ITrec = pushLexState layout
1034 maybe_layout _ = return ()
1036 -- Pushing a new implicit layout context. If the indentation of the
1037 -- next token is not greater than the previous layout context, then
1038 -- Haskell 98 says that the new layout context should be empty; that is
1039 -- the lexer must generate {}.
1041 -- We are slightly more lenient than this: when the new context is started
1042 -- by a 'do', then we allow the new context to be at the same indentation as
1043 -- the previous context. This is what the 'strict' argument is for.
1045 new_layout_context strict span _buf _len = do
1047 (AI _ offset _) <- getInput
1050 Layout prev_off : _ |
1051 (strict && prev_off >= offset ||
1052 not strict && prev_off > offset) -> do
1053 -- token is indented to the left of the previous context.
1054 -- we must generate a {} sequence now.
1055 pushLexState layout_left
1056 return (L span ITvocurly)
1058 setContext (Layout offset : ctx)
1059 return (L span ITvocurly)
1061 do_layout_left span _buf _len = do
1063 pushLexState bol -- we must be at the start of a line
1064 return (L span ITvccurly)
1066 -- -----------------------------------------------------------------------------
1069 setLine :: Int -> Action
1070 setLine code span buf len = do
1071 let line = parseUnsignedInteger buf len 10 octDecDigit
1072 setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1073 -- subtract one: the line number refers to the *following* line
1078 setFile :: Int -> Action
1079 setFile code span buf len = do
1080 let file = lexemeToFastString (stepOn buf) (len-2)
1081 setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1087 -- -----------------------------------------------------------------------------
1088 -- Options, includes and language pragmas.
1090 lex_string_prag :: (String -> Token) -> Action
1091 lex_string_prag mkTok span _buf _len
1092 = do input <- getInput
1096 return (L (mkSrcSpan start end) tok)
1098 = if isString input "#-}"
1099 then do setInput input
1100 return (mkTok (reverse acc))
1101 else case alexGetChar input of
1102 Just (c,i) -> go (c:acc) i
1103 Nothing -> err input
1104 isString _ [] = True
1106 = case alexGetChar i of
1107 Just (c,i') | c == x -> isString i' xs
1109 err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"
1112 -- -----------------------------------------------------------------------------
1115 -- This stuff is horrible. I hates it.
1117 lex_string_tok :: Action
1118 lex_string_tok span _buf _len = do
1119 tok <- lex_string ""
1121 return (L (mkSrcSpan (srcSpanStart span) end) tok)
1123 lex_string :: String -> P Token
1126 case alexGetChar' i of
1127 Nothing -> lit_error
1131 magicHash <- extension magicHashEnabled
1135 case alexGetChar' i of
1139 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1140 else let s' = mkZFastString (reverse s) in
1141 return (ITprimstring s')
1142 -- mkZFastString is a hack to avoid encoding the
1143 -- string in UTF-8. We just want the exact bytes.
1145 return (ITstring (mkFastString (reverse s)))
1147 return (ITstring (mkFastString (reverse s)))
1150 | Just ('&',i) <- next -> do
1151 setInput i; lex_string s
1152 | Just (c,i) <- next, is_space c -> do
1153 setInput i; lex_stringgap s
1154 where next = alexGetChar' i
1160 lex_stringgap s = do
1163 '\\' -> lex_string s
1164 c | is_space c -> lex_stringgap s
1168 lex_char_tok :: Action
1169 -- Here we are basically parsing character literals, such as 'x' or '\n'
1170 -- but, when Template Haskell is on, we additionally spot
1171 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
1172 -- but WIHTOUT CONSUMING the x or T part (the parser does that).
1173 -- So we have to do two characters of lookahead: when we see 'x we need to
1174 -- see if there's a trailing quote
1175 lex_char_tok span _buf _len = do -- We've seen '
1176 i1 <- getInput -- Look ahead to first character
1177 let loc = srcSpanStart span
1178 case alexGetChar' i1 of
1179 Nothing -> lit_error
1181 Just ('\'', i2@(AI end2 _ _)) -> do -- We've seen ''
1182 th_exts <- extension thEnabled
1185 return (L (mkSrcSpan loc end2) ITtyQuote)
1188 Just ('\\', i2@(AI _end2 _ _)) -> do -- We've seen 'backslash
1190 lit_ch <- lex_escape
1191 mc <- getCharOrFail -- Trailing quote
1192 if mc == '\'' then finish_char_tok loc lit_ch
1193 else do setInput i2; lit_error
1195 Just (c, i2@(AI _end2 _ _))
1196 | not (isAny c) -> lit_error
1199 -- We've seen 'x, where x is a valid character
1200 -- (i.e. not newline etc) but not a quote or backslash
1201 case alexGetChar' i2 of -- Look ahead one more character
1202 Nothing -> lit_error
1203 Just ('\'', i3) -> do -- We've seen 'x'
1205 finish_char_tok loc c
1206 _other -> do -- We've seen 'x not followed by quote
1207 -- If TH is on, just parse the quote only
1208 th_exts <- extension thEnabled
1209 let (AI end _ _) = i1
1210 if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1211 else do setInput i2; lit_error
1213 finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1214 finish_char_tok loc ch -- We've already seen the closing quote
1215 -- Just need to check for trailing #
1216 = do magicHash <- extension magicHashEnabled
1217 i@(AI end _ _) <- getInput
1218 if magicHash then do
1219 case alexGetChar' i of
1220 Just ('#',i@(AI end _ _)) -> do
1222 return (L (mkSrcSpan loc end) (ITprimchar ch))
1224 return (L (mkSrcSpan loc end) (ITchar ch))
1226 return (L (mkSrcSpan loc end) (ITchar ch))
1228 lex_char :: Char -> AlexInput -> P Char
1231 '\\' -> do setInput inp; lex_escape
1232 c | isAny c -> do setInput inp; return c
1235 isAny c | c > '\x7f' = isPrint c
1236 | otherwise = is_any c
1238 lex_escape :: P Char
1252 '^' -> do c <- getCharOrFail
1253 if c >= '@' && c <= '_'
1254 then return (chr (ord c - ord '@'))
1257 'x' -> readNum is_hexdigit 16 hexDigit
1258 'o' -> readNum is_octdigit 8 octDecDigit
1259 x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x)
1263 case alexGetChar' i of
1264 Nothing -> lit_error
1266 case alexGetChar' i2 of
1267 Nothing -> do setInput i2; lit_error
1269 let str = [c1,c2,c3] in
1270 case [ (c,rest) | (p,c) <- silly_escape_chars,
1271 Just rest <- [maybePrefixMatch p str] ] of
1272 (escape_char,[]):_ -> do
1275 (escape_char,_:_):_ -> do
1280 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
1281 readNum is_digit base conv = do
1285 then readNum2 is_digit base conv (conv c)
1286 else do setInput i; lit_error
1288 readNum2 is_digit base conv i = do
1291 where read i input = do
1292 case alexGetChar' input of
1293 Just (c,input') | is_digit c -> do
1294 read (i*base + conv c) input'
1296 if i >= 0 && i <= 0x10FFFF
1297 then do setInput input; return (chr i)
1300 silly_escape_chars = [
1337 -- before calling lit_error, ensure that the current input is pointing to
1338 -- the position of the error in the buffer. This is so that we can report
1339 -- a correct location to the user, but also so we can detect UTF-8 decoding
1340 -- errors if they occur.
1341 lit_error = lexError "lexical error in string/character literal"
1343 getCharOrFail :: P Char
1346 case alexGetChar' i of
1347 Nothing -> lexError "unexpected end-of-file in string/character literal"
1348 Just (c,i) -> do setInput i; return c
1350 -- -----------------------------------------------------------------------------
1353 lex_quasiquote_tok :: Action
1354 lex_quasiquote_tok span buf len = do
1355 let quoter = reverse $ takeWhile (/= '$')
1356 $ reverse $ lexemeToString buf (len - 1)
1357 quoteStart <- getSrcLoc
1358 quote <- lex_quasiquote ""
1360 return (L (mkSrcSpan (srcSpanStart span) end)
1361 (ITquasiQuote (mkFastString quoter,
1362 mkFastString (reverse quote),
1363 mkSrcSpan quoteStart end)))
1365 lex_quasiquote :: String -> P String
1366 lex_quasiquote s = do
1368 case alexGetChar' i of
1369 Nothing -> lit_error
1372 | Just ('|',i) <- next -> do
1373 setInput i; lex_quasiquote ('|' : s)
1374 | Just (']',i) <- next -> do
1375 setInput i; lex_quasiquote (']' : s)
1376 where next = alexGetChar' i
1379 | Just (']',i) <- next -> do
1380 setInput i; return s
1381 where next = alexGetChar' i
1384 setInput i; lex_quasiquote (c : s)
1386 -- -----------------------------------------------------------------------------
1389 warn :: DynFlag -> SDoc -> Action
1390 warn option warning srcspan _buf _len = do
1391 addWarning option srcspan warning
1394 warnThen :: DynFlag -> SDoc -> Action -> Action
1395 warnThen option warning action srcspan buf len = do
1396 addWarning option srcspan warning
1397 action srcspan buf len
1399 -- -----------------------------------------------------------------------------
1410 SrcSpan -- The start and end of the text span related to
1411 -- the error. Might be used in environments which can
1412 -- show this span, e.g. by highlighting it.
1413 Message -- The error message
1415 data PState = PState {
1416 buffer :: StringBuffer,
1418 messages :: Messages,
1419 last_loc :: SrcSpan, -- pos of previous token
1420 last_offs :: !Int, -- offset of the previous token from the
1421 -- beginning of the current line.
1422 -- \t is equal to 8 spaces.
1423 last_len :: !Int, -- len of previous token
1424 last_line_len :: !Int,
1425 loc :: SrcLoc, -- current loc (end of prev token + 1)
1426 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1427 context :: [LayoutContext],
1430 -- last_loc and last_len are used when generating error messages,
1431 -- and in pushCurrentContext only. Sigh, if only Happy passed the
1432 -- current token to happyError, we could at least get rid of last_len.
1433 -- Getting rid of last_loc would require finding another way to
1434 -- implement pushCurrentContext (which is only called from one place).
1436 newtype P a = P { unP :: PState -> ParseResult a }
1438 instance Monad P where
1444 returnP a = a `seq` (P $ \s -> POk s a)
1446 thenP :: P a -> (a -> P b) -> P b
1447 (P m) `thenP` k = P $ \ s ->
1449 POk s1 a -> (unP (k a)) s1
1450 PFailed span err -> PFailed span err
1452 failP :: String -> P a
1453 failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1455 failMsgP :: String -> P a
1456 failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1458 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1459 failLocMsgP loc1 loc2 str = P $ \_ -> PFailed (mkSrcSpan loc1 loc2) (text str)
1461 failSpanMsgP :: SrcSpan -> SDoc -> P a
1462 failSpanMsgP span msg = P $ \_ -> PFailed span msg
1464 extension :: (Int -> Bool) -> P Bool
1465 extension p = P $ \s -> POk s (p $! extsBitmap s)
1468 getExts = P $ \s -> POk s (extsBitmap s)
1470 setExts :: (Int -> Int) -> P ()
1471 setExts f = P $ \s -> POk s{ extsBitmap = f (extsBitmap s) } ()
1473 setSrcLoc :: SrcLoc -> P ()
1474 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1476 getSrcLoc :: P SrcLoc
1477 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1479 setLastToken :: SrcSpan -> Int -> Int -> P ()
1480 setLastToken loc len line_len = P $ \s -> POk s {
1483 last_line_len=line_len
1486 data AlexInput = AI SrcLoc {-#UNPACK#-}!Int StringBuffer
1488 alexInputPrevChar :: AlexInput -> Char
1489 alexInputPrevChar (AI _ _ buf) = prevChar buf '\n'
1491 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1492 alexGetChar (AI loc ofs s)
1494 | otherwise = adj_c `seq` loc' `seq` ofs' `seq` s' `seq`
1495 --trace (show (ord c)) $
1496 Just (adj_c, (AI loc' ofs' s'))
1497 where (c,s') = nextChar s
1498 loc' = advanceSrcLoc loc c
1499 ofs' = advanceOffs c ofs
1507 other_graphic = '\x6'
1510 | c <= '\x06' = non_graphic
1512 -- Alex doesn't handle Unicode, so when Unicode
1513 -- character is encoutered we output these values
1514 -- with the actual character value hidden in the state.
1516 case generalCategory c of
1517 UppercaseLetter -> upper
1518 LowercaseLetter -> lower
1519 TitlecaseLetter -> upper
1520 ModifierLetter -> other_graphic
1521 OtherLetter -> lower -- see #1103
1522 NonSpacingMark -> other_graphic
1523 SpacingCombiningMark -> other_graphic
1524 EnclosingMark -> other_graphic
1525 DecimalNumber -> digit
1526 LetterNumber -> other_graphic
1527 OtherNumber -> other_graphic
1528 ConnectorPunctuation -> other_graphic
1529 DashPunctuation -> other_graphic
1530 OpenPunctuation -> other_graphic
1531 ClosePunctuation -> other_graphic
1532 InitialQuote -> other_graphic
1533 FinalQuote -> other_graphic
1534 OtherPunctuation -> other_graphic
1535 MathSymbol -> symbol
1536 CurrencySymbol -> symbol
1537 ModifierSymbol -> symbol
1538 OtherSymbol -> symbol
1540 _other -> non_graphic
1542 -- This version does not squash unicode characters, it is used when
1544 alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
1545 alexGetChar' (AI loc ofs s)
1547 | otherwise = c `seq` loc' `seq` ofs' `seq` s' `seq`
1548 --trace (show (ord c)) $
1549 Just (c, (AI loc' ofs' s'))
1550 where (c,s') = nextChar s
1551 loc' = advanceSrcLoc loc c
1552 ofs' = advanceOffs c ofs
1554 advanceOffs :: Char -> Int -> Int
1555 advanceOffs '\n' _ = 0
1556 advanceOffs '\t' offs = (offs `quot` 8 + 1) * 8
1557 advanceOffs _ offs = offs + 1
1559 getInput :: P AlexInput
1560 getInput = P $ \s@PState{ loc=l, last_offs=o, buffer=b } -> POk s (AI l o b)
1562 setInput :: AlexInput -> P ()
1563 setInput (AI l o b) = P $ \s -> POk s{ loc=l, last_offs=o, buffer=b } ()
1565 pushLexState :: Int -> P ()
1566 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1568 popLexState :: P Int
1569 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1571 getLexState :: P Int
1572 getLexState = P $ \s@PState{ lex_state=ls:_ } -> POk s ls
1574 -- for reasons of efficiency, flags indicating language extensions (eg,
1575 -- -fglasgow-exts or -XParr) are represented by a bitmap stored in an unboxed
1578 genericsBit, ffiBit, parrBit :: Int
1579 genericsBit = 0 -- {| and |}
1585 explicitForallBit = 7 -- the 'forall' keyword and '.' symbol
1586 bangPatBit = 8 -- Tells the parser to understand bang-patterns
1587 -- (doesn't affect the lexer)
1588 tyFamBit = 9 -- indexed type families: 'family' keyword and kind sigs
1589 haddockBit = 10 -- Lex and parse Haddock comments
1590 magicHashBit = 11 -- "#" in both functions and operators
1591 kindSigsBit = 12 -- Kind signatures on type variables
1592 recursiveDoBit = 13 -- mdo
1593 unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc
1594 unboxedTuplesBit = 15 -- (# and #)
1595 standaloneDerivingBit = 16 -- standalone instance deriving declarations
1596 transformComprehensionsBit = 17
1597 qqBit = 18 -- enable quasiquoting
1600 genericsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1602 genericsEnabled flags = testBit flags genericsBit
1603 ffiEnabled flags = testBit flags ffiBit
1604 parrEnabled flags = testBit flags parrBit
1605 arrowsEnabled flags = testBit flags arrowsBit
1606 thEnabled flags = testBit flags thBit
1607 ipEnabled flags = testBit flags ipBit
1608 explicitForallEnabled flags = testBit flags explicitForallBit
1609 bangPatEnabled flags = testBit flags bangPatBit
1610 tyFamEnabled flags = testBit flags tyFamBit
1611 haddockEnabled flags = testBit flags haddockBit
1612 magicHashEnabled flags = testBit flags magicHashBit
1613 kindSigsEnabled flags = testBit flags kindSigsBit
1614 recursiveDoEnabled flags = testBit flags recursiveDoBit
1615 unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit
1616 unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit
1617 standaloneDerivingEnabled flags = testBit flags standaloneDerivingBit
1618 transformComprehensionsEnabled flags = testBit flags transformComprehensionsBit
1619 qqEnabled flags = testBit flags qqBit
1620 inRulePrag flags = testBit flags inRulePragBit
1622 -- PState for parsing options pragmas
1624 pragState :: DynFlags -> StringBuffer -> SrcLoc -> PState
1625 pragState dynflags buf loc =
1628 messages = emptyMessages,
1630 last_loc = mkSrcSpan loc loc,
1637 lex_state = [bol, option_prags, 0]
1641 -- create a parse state
1643 mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
1644 mkPState buf loc flags =
1648 messages = emptyMessages,
1649 last_loc = mkSrcSpan loc loc,
1654 extsBitmap = fromIntegral bitmap,
1656 lex_state = [bol, 0]
1657 -- we begin in the layout state if toplev_layout is set
1660 bitmap = genericsBit `setBitIf` dopt Opt_Generics flags
1661 .|. ffiBit `setBitIf` dopt Opt_ForeignFunctionInterface flags
1662 .|. parrBit `setBitIf` dopt Opt_PArr flags
1663 .|. arrowsBit `setBitIf` dopt Opt_Arrows flags
1664 .|. thBit `setBitIf` dopt Opt_TemplateHaskell flags
1665 .|. qqBit `setBitIf` dopt Opt_QuasiQuotes flags
1666 .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
1667 .|. explicitForallBit `setBitIf` dopt Opt_ScopedTypeVariables flags
1668 .|. explicitForallBit `setBitIf` dopt Opt_LiberalTypeSynonyms flags
1669 .|. explicitForallBit `setBitIf` dopt Opt_PolymorphicComponents flags
1670 .|. explicitForallBit `setBitIf` dopt Opt_ExistentialQuantification flags
1671 .|. explicitForallBit `setBitIf` dopt Opt_Rank2Types flags
1672 .|. explicitForallBit `setBitIf` dopt Opt_RankNTypes flags
1673 .|. bangPatBit `setBitIf` dopt Opt_BangPatterns flags
1674 .|. tyFamBit `setBitIf` dopt Opt_TypeFamilies flags
1675 .|. haddockBit `setBitIf` dopt Opt_Haddock flags
1676 .|. magicHashBit `setBitIf` dopt Opt_MagicHash flags
1677 .|. kindSigsBit `setBitIf` dopt Opt_KindSignatures flags
1678 .|. recursiveDoBit `setBitIf` dopt Opt_RecursiveDo flags
1679 .|. unicodeSyntaxBit `setBitIf` dopt Opt_UnicodeSyntax flags
1680 .|. unboxedTuplesBit `setBitIf` dopt Opt_UnboxedTuples flags
1681 .|. standaloneDerivingBit `setBitIf` dopt Opt_StandaloneDeriving flags
1682 .|. transformComprehensionsBit `setBitIf` dopt Opt_TransformListComp flags
1684 setBitIf :: Int -> Bool -> Int
1685 b `setBitIf` cond | cond = bit b
1688 addWarning :: DynFlag -> SrcSpan -> SDoc -> P ()
1689 addWarning option srcspan warning
1690 = P $ \s@PState{messages=(ws,es), dflags=d} ->
1691 let warning' = mkWarnMsg srcspan alwaysQualify warning
1692 ws' = if dopt option d then ws `snocBag` warning' else ws
1693 in POk s{messages=(ws', es)} ()
1695 getMessages :: PState -> Messages
1696 getMessages PState{messages=ms} = ms
1698 getContext :: P [LayoutContext]
1699 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1701 setContext :: [LayoutContext] -> P ()
1702 setContext ctx = P $ \s -> POk s{context=ctx} ()
1705 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1706 last_len = len, last_loc = last_loc }) ->
1708 (_:tl) -> POk s{ context = tl } ()
1709 [] -> PFailed last_loc (srcParseErr buf len)
1711 -- Push a new layout context at the indentation of the last token read.
1712 -- This is only used at the outer level of a module when the 'module'
1713 -- keyword is missing.
1714 pushCurrentContext :: P ()
1715 pushCurrentContext = P $ \ s@PState{ last_offs=offs, last_line_len=len, context=ctx } ->
1716 POk s{context = Layout (offs-len) : ctx} ()
1717 --trace ("off: " ++ show offs ++ ", len: " ++ show len) $ POk s{context = Layout (offs-len) : ctx} ()
1719 getOffside :: P Ordering
1720 getOffside = P $ \s@PState{last_offs=offs, context=stk} ->
1721 let ord = case stk of
1722 (Layout n:_) -> compare offs n
1726 -- ---------------------------------------------------------------------------
1727 -- Construct a parse error
1730 :: StringBuffer -- current buffer (placed just after the last token)
1731 -> Int -- length of the previous token
1734 = hcat [ if null token
1735 then ptext (sLit "parse error (possibly incorrect indentation)")
1736 else hcat [ptext (sLit "parse error on input "),
1737 char '`', text token, char '\'']
1739 where token = lexemeToString (offsetBytes (-len) buf) len
1741 -- Report a parse failure, giving the span of the previous token as
1742 -- the location of the error. This is the entry point for errors
1743 -- detected during parsing.
1745 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1746 last_loc = last_loc } ->
1747 PFailed last_loc (srcParseErr buf len)
1749 -- A lexical error is reported at a particular position in the source file,
1750 -- not over a token range.
1751 lexError :: String -> P a
1754 (AI end _ buf) <- getInput
1755 reportLexError loc end buf str
1757 -- -----------------------------------------------------------------------------
1758 -- This is the top-level function: called from the parser each time a
1759 -- new token is to be read from the input.
1761 lexer :: (Located Token -> P a) -> P a
1763 tok@(L _span _tok__) <- lexToken
1764 -- trace ("token: " ++ show tok__) $ do
1767 lexToken :: P (Located Token)
1769 inp@(AI loc1 _ buf) <- getInput
1772 case alexScanUser exts inp sc of
1774 let span = mkSrcSpan loc1 loc1
1775 setLastToken span 0 0
1776 return (L span ITeof)
1777 AlexError (AI loc2 _ buf) ->
1778 reportLexError loc1 loc2 buf "lexical error"
1779 AlexSkip inp2 _ -> do
1782 AlexToken inp2@(AI end _ buf2) _ t -> do
1784 let span = mkSrcSpan loc1 end
1785 let bytes = byteDiff buf buf2
1786 span `seq` setLastToken span bytes bytes
1789 reportLexError loc1 loc2 buf str
1790 | atEnd buf = failLocMsgP loc1 loc2 (str ++ " at end of input")
1793 c = fst (nextChar buf)
1795 if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar#
1796 then failLocMsgP loc2 loc2 (str ++ " (UTF-8 decoding error)")
1797 else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c)