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 Token(..), lexer, pragState, mkPState, PState(..),
26 P(..), ParseResult(..), getSrcLoc,
27 failLocMsgP, failSpanMsgP, srcParseFail,
29 popContext, pushCurrentContext, setLastToken, setSrcLoc,
30 getLexState, popLexState, pushLexState,
31 extension, glaExtsEnabled, bangPatEnabled
34 #include "HsVersions.h"
46 import Util ( maybePrefixMatch, readRational )
50 import Data.Char ( chr, isSpace )
54 #if __GLASGOW_HASKELL__ >= 605
55 import Data.Char ( GeneralCategory(..), generalCategory, isPrint, isUpper )
57 import Compat.Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper )
62 $whitechar = [\ \n\r\f\v\xa0 $unispace]
63 $white_no_nl = $whitechar # \n
68 $decdigit = $ascdigit -- for now, should really be $digit (ToDo)
69 $digit = [$ascdigit $unidigit]
71 $special = [\(\)\,\;\[\]\`\{\}]
72 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~ \xa1-\xbf \xd7 \xf7]
74 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
77 $asclarge = [A-Z \xc0-\xd6 \xd8-\xde]
78 $large = [$asclarge $unilarge]
81 $ascsmall = [a-z \xdf-\xf6 \xf8-\xff]
82 $small = [$ascsmall $unismall \_]
85 $graphic = [$small $large $symbol $digit $special $unigraphic \:\"\']
88 $hexit = [$decdigit A-F a-f]
89 $symchar = [$symbol \:]
91 $idchar = [$small $large $digit \']
93 $docsym = [\| \^ \* \$]
95 @varid = $small $idchar*
96 @conid = $large $idchar*
98 @varsym = $symbol $symchar*
99 @consym = \: $symchar*
101 @decimal = $decdigit+
103 @hexadecimal = $hexit+
104 @exponent = [eE] [\-\+]? @decimal
106 -- we support the hierarchical module name extension:
109 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
113 -- everywhere: skip whitespace and comments
115 $tab+ { warn Opt_WarnTabs (text "Tab character") }
117 -- Everywhere: deal with nested comments. We explicitly rule out
118 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
119 -- (this can happen even though pragmas will normally take precedence due to
120 -- longest-match, because pragmas aren't valid in every state, but comments
121 -- are). We also rule out nested Haddock comments, if the -haddock flag is
124 "{-" / { isNormalComment } { nested_comment lexToken }
126 -- Single-line comments are a bit tricky. Haskell 98 says that two or
127 -- more dashes followed by a symbol should be parsed as a varsym, so we
128 -- have to exclude those.
130 -- Since Haddock comments aren't valid in every state, we need to rule them
133 -- The following two rules match comments that begin with two dashes, but
134 -- continue with a different character. The rules test that this character
135 -- is not a symbol (in which case we'd have a varsym), and that it's not a
136 -- space followed by a Haddock comment symbol (docsym) (in which case we'd
137 -- have a Haddock comment). The rules then munch the rest of the line.
140 "--" [^$symbol : \ ] .* ;
142 -- Next, match Haddock comments if no -haddock flag
144 "-- " $docsym .* / { ifExtension (not . haddockEnabled) } ;
146 -- Now, when we've matched comments that begin with 2 dashes and continue
147 -- with a different character, we need to match comments that begin with three
148 -- or more dashes (which clearly can't be Haddock comments). We only need to
149 -- make sure that the first non-dash character isn't a symbol, and munch the
152 "---"\-* [^$symbol :] .* ;
154 -- Since the previous rules all match dashes followed by at least one
155 -- character, we also need to match a whole line filled with just dashes.
157 "--"\-* / { atEOL } ;
159 -- We need this rule since none of the other single line comment rules
160 -- actually match this case.
164 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
165 -- blank lines) until we find a non-whitespace character, then do layout
168 -- One slight wibble here: what if the line begins with {-#? In
169 -- theory, we have to lex the pragma to see if it's one we recognise,
170 -- and if it is, then we backtrack and do_bol, otherwise we treat it
171 -- as a nested comment. We don't bother with this: if the line begins
172 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
175 ^\# (line)? { begin line_prag1 }
176 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
177 ^\# \! .* \n ; -- #!, for scripts
181 -- after a layout keyword (let, where, do, of), we begin a new layout
182 -- context if the curly brace is missing.
183 -- Careful! This stuff is quite delicate.
184 <layout, layout_do> {
185 \{ / { notFollowedBy '-' } { pop_and open_brace }
186 -- we might encounter {-# here, but {- has been handled already
188 ^\# (line)? { begin line_prag1 }
191 -- do is treated in a subtly different way, see new_layout_context
192 <layout> () { new_layout_context True }
193 <layout_do> () { new_layout_context False }
195 -- after a new layout context which was found to be to the left of the
196 -- previous context, we have generated a '{' token, and we now need to
197 -- generate a matching '}' token.
198 <layout_left> () { do_layout_left }
200 <0,option_prags,glaexts> \n { begin bol }
202 "{-#" $whitechar* (line|LINE) { begin line_prag2 }
204 -- single-line line pragmas, of the form
205 -- # <line> "<file>" <extra-stuff> \n
206 <line_prag1> $decdigit+ { setLine line_prag1a }
207 <line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
208 <line_prag1b> .* { pop }
210 -- Haskell-style line pragmas, of the form
211 -- {-# LINE <line> "<file>" #-}
212 <line_prag2> $decdigit+ { setLine line_prag2a }
213 <line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
214 <line_prag2b> "#-}"|"-}" { pop }
215 -- NOTE: accept -} at the end of a LINE pragma, for compatibility
216 -- with older versions of GHC which generated these.
218 -- We only want RULES pragmas to be picked up when -fglasgow-exts
219 -- is on, because the contents of the pragma is always written using
220 -- glasgow-exts syntax (using forall etc.), so if glasgow exts are not
221 -- enabled, we're sure to get a parse error.
222 -- (ToDo: we should really emit a warning when ignoring pragmas)
224 "{-#" $whitechar* (RULES|rules) { token ITrules_prag }
226 <0,option_prags,glaexts> {
227 "{-#" $whitechar* (INLINE|inline) { token (ITinline_prag True) }
228 "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
229 { token (ITinline_prag False) }
230 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
231 { token ITspec_prag }
232 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
233 $whitechar* (INLINE|inline) { token (ITspec_inline_prag True) }
234 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
235 $whitechar* (NO(T?)INLINE|no(t?)inline)
236 { token (ITspec_inline_prag False) }
237 "{-#" $whitechar* (SOURCE|source) { token ITsource_prag }
238 "{-#" $whitechar* (DEPRECATED|deprecated)
239 { token ITdeprecated_prag }
240 "{-#" $whitechar* (SCC|scc) { token ITscc_prag }
241 "{-#" $whitechar* (GENERATED|generated)
242 { token ITgenerated_prag }
243 "{-#" $whitechar* (CORE|core) { token ITcore_prag }
244 "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag }
246 "{-#" $whitechar* (DOCOPTIONS|docoptions)
247 / { ifExtension haddockEnabled } { lex_string_prag ITdocOptions }
249 "{-#" { nested_comment lexToken }
251 -- ToDo: should only be valid inside a pragma:
252 "#-}" { token ITclose_prag}
256 "{-#" $whitechar* (OPTIONS|options) { lex_string_prag IToptions_prag }
257 "{-#" $whitechar* (OPTIONS_GHC|options_ghc)
258 { lex_string_prag IToptions_prag }
259 "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
260 "{-#" $whitechar* (INCLUDE|include) { lex_string_prag ITinclude_prag }
263 <0,option_prags,glaexts> {
264 -- This is to catch things like {-# OPTIONS OPTIONS_HUGS ...
265 "{-#" $whitechar* $idchar+ { nested_comment lexToken }
268 -- '0' state: ordinary lexemes
269 -- 'glaexts' state: glasgow extensions (postfix '#', etc.)
274 "-- " $docsym / { ifExtension haddockEnabled } { multiline_doc_comment }
275 "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
281 "[:" / { ifExtension parrEnabled } { token ITopabrack }
282 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
286 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
287 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
288 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
289 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
290 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
291 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
292 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
293 "$(" / { ifExtension thEnabled } { token ITparenEscape }
297 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
298 { special IToparenbar }
299 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
303 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
307 "(#" / { notFollowedBySymbol } { token IToubxparen }
308 "#)" { token ITcubxparen }
309 "{|" { token ITocurlybar }
310 "|}" { token ITccurlybar }
313 <0,option_prags,glaexts> {
314 \( { special IToparen }
315 \) { special ITcparen }
316 \[ { special ITobrack }
317 \] { special ITcbrack }
318 \, { special ITcomma }
319 \; { special ITsemi }
320 \` { special ITbackquote }
326 <0,option_prags,glaexts> {
327 @qual @varid { check_qvarid }
328 @qual @conid { idtoken qconid }
330 @conid { idtoken conid }
333 -- after an illegal qvarid, such as 'M.let',
334 -- we back up and try again in the bad_qvarid state:
336 @conid { pop_and (idtoken conid) }
337 @qual @conid { pop_and (idtoken qconid) }
341 @qual @varid "#"+ { idtoken qvarid }
342 @qual @conid "#"+ { idtoken qconid }
343 @varid "#"+ { varid }
344 @conid "#"+ { idtoken conid }
350 @qual @varsym { idtoken qvarsym }
351 @qual @consym { idtoken qconsym }
357 @decimal { tok_decimal }
358 0[oO] @octal { tok_octal }
359 0[xX] @hexadecimal { tok_hexadecimal }
363 @decimal \# { prim_decimal }
364 0[oO] @octal \# { prim_octal }
365 0[xX] @hexadecimal \# { prim_hexadecimal }
368 <0,glaexts> @floating_point { strtoken tok_float }
369 <glaexts> @floating_point \# { init_strtoken 1 prim_float }
370 <glaexts> @floating_point \# \# { init_strtoken 2 prim_double }
372 -- Strings and chars are lexed by hand-written code. The reason is
373 -- that even if we recognise the string or char here in the regex
374 -- lexer, we would still have to parse the string afterward in order
375 -- to convert it to a String.
378 \" { lex_string_tok }
382 -- -----------------------------------------------------------------------------
386 = ITas -- Haskell keywords
411 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
413 | ITforall -- GHC extension keywords
428 | ITinline_prag Bool -- True <=> INLINE, False <=> NOINLINE
429 | ITspec_prag -- SPECIALISE
430 | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE)
437 | ITcore_prag -- hdaume: core annotations
440 | IToptions_prag String
441 | ITinclude_prag String
444 | ITdotdot -- reserved symbols
460 | ITbiglam -- GHC-extension symbols
462 | ITocurly -- special symbols
464 | ITocurlybar -- {|, for type applications
465 | ITccurlybar -- |}, for type applications
469 | ITopabrack -- [:, for parallel arrays with -fparr
470 | ITcpabrack -- :], for parallel arrays with -fparr
481 | ITvarid FastString -- identifiers
483 | ITvarsym FastString
484 | ITconsym FastString
485 | ITqvarid (FastString,FastString)
486 | ITqconid (FastString,FastString)
487 | ITqvarsym (FastString,FastString)
488 | ITqconsym (FastString,FastString)
490 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
492 | ITpragma StringBuffer
495 | ITstring FastString
497 | ITrational Rational
500 | ITprimstring FastString
502 | ITprimfloat Rational
503 | ITprimdouble Rational
505 -- MetaHaskell extension tokens
506 | ITopenExpQuote -- [| or [e|
507 | ITopenPatQuote -- [p|
508 | ITopenDecQuote -- [d|
509 | ITopenTypQuote -- [t|
511 | ITidEscape FastString -- $x
512 | ITparenEscape -- $(
516 -- Arrow notation extension
523 | ITLarrowtail -- -<<
524 | ITRarrowtail -- >>-
526 | ITunknown String -- Used when the lexer can't make sense of it
527 | ITeof -- end of file token
529 -- Documentation annotations
530 | ITdocCommentNext String -- something beginning '-- |'
531 | ITdocCommentPrev String -- something beginning '-- ^'
532 | ITdocCommentNamed String -- something beginning '-- $'
533 | ITdocSection Int String -- a section heading
534 | ITdocOptions String -- doc options (prune, ignore-exports, etc)
537 deriving Show -- debugging
540 isSpecial :: Token -> Bool
541 -- If we see M.x, where x is a keyword, but
542 -- is special, we treat is as just plain M.x,
544 isSpecial ITas = True
545 isSpecial IThiding = True
546 isSpecial ITderive = True
547 isSpecial ITqualified = True
548 isSpecial ITforall = True
549 isSpecial ITexport = True
550 isSpecial ITlabel = True
551 isSpecial ITdynamic = True
552 isSpecial ITsafe = True
553 isSpecial ITthreadsafe = True
554 isSpecial ITunsafe = True
555 isSpecial ITccallconv = True
556 isSpecial ITstdcallconv = True
557 isSpecial ITmdo = True
558 isSpecial ITfamily = True
561 -- the bitmap provided as the third component indicates whether the
562 -- corresponding extension keyword is valid under the extension options
563 -- provided to the compiler; if the extension corresponding to *any* of the
564 -- bits set in the bitmap is enabled, the keyword is valid (this setup
565 -- facilitates using a keyword in two different extensions that can be
566 -- activated independently)
568 reservedWordsFM = listToUFM $
569 map (\(x, y, z) -> (mkFastString x, (y, z)))
570 [( "_", ITunderscore, 0 ),
572 ( "case", ITcase, 0 ),
573 ( "class", ITclass, 0 ),
574 ( "data", ITdata, 0 ),
575 ( "default", ITdefault, 0 ),
576 ( "deriving", ITderiving, 0 ),
577 ( "derive", ITderive, 0 ),
579 ( "else", ITelse, 0 ),
580 ( "hiding", IThiding, 0 ),
582 ( "import", ITimport, 0 ),
584 ( "infix", ITinfix, 0 ),
585 ( "infixl", ITinfixl, 0 ),
586 ( "infixr", ITinfixr, 0 ),
587 ( "instance", ITinstance, 0 ),
589 ( "module", ITmodule, 0 ),
590 ( "newtype", ITnewtype, 0 ),
592 ( "qualified", ITqualified, 0 ),
593 ( "then", ITthen, 0 ),
594 ( "type", ITtype, 0 ),
595 ( "where", ITwhere, 0 ),
596 ( "_scc_", ITscc, 0 ), -- ToDo: remove
598 ( "forall", ITforall, bit tvBit),
599 ( "mdo", ITmdo, bit glaExtsBit),
600 ( "family", ITfamily, bit idxTysBit),
602 ( "foreign", ITforeign, bit ffiBit),
603 ( "export", ITexport, bit ffiBit),
604 ( "label", ITlabel, bit ffiBit),
605 ( "dynamic", ITdynamic, bit ffiBit),
606 ( "safe", ITsafe, bit ffiBit),
607 ( "threadsafe", ITthreadsafe, bit ffiBit),
608 ( "unsafe", ITunsafe, bit ffiBit),
609 ( "stdcall", ITstdcallconv, bit ffiBit),
610 ( "ccall", ITccallconv, bit ffiBit),
611 ( "dotnet", ITdotnet, bit ffiBit),
613 ( "rec", ITrec, bit arrowsBit),
614 ( "proc", ITproc, bit arrowsBit)
617 reservedSymsFM = listToUFM $
618 map (\ (x,y,z) -> (mkFastString x,(y,z)))
619 [ ("..", ITdotdot, 0)
620 ,(":", ITcolon, 0) -- (:) is a reserved op,
621 -- meaning only list cons
634 ,("*", ITstar, bit glaExtsBit .|.
635 bit idxTysBit) -- For data T (a::*) = MkT
636 ,(".", ITdot, bit tvBit) -- For 'forall a . t'
638 ,("-<", ITlarrowtail, bit arrowsBit)
639 ,(">-", ITrarrowtail, bit arrowsBit)
640 ,("-<<", ITLarrowtail, bit arrowsBit)
641 ,(">>-", ITRarrowtail, bit arrowsBit)
643 #if __GLASGOW_HASKELL__ >= 605
644 ,("∷", ITdcolon, bit glaExtsBit)
645 ,("⇒", ITdarrow, bit glaExtsBit)
646 ,("∀", ITforall, bit glaExtsBit)
647 ,("→", ITrarrow, bit glaExtsBit)
648 ,("←", ITlarrow, bit glaExtsBit)
649 ,("⋯", ITdotdot, bit glaExtsBit)
650 -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
651 -- form part of a large operator. This would let us have a better
652 -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
656 -- -----------------------------------------------------------------------------
659 type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
661 special :: Token -> Action
662 special tok span _buf len = return (L span tok)
664 token, layout_token :: Token -> Action
665 token t span buf len = return (L span t)
666 layout_token t span buf len = pushLexState layout >> return (L span t)
668 idtoken :: (StringBuffer -> Int -> Token) -> Action
669 idtoken f span buf len = return (L span $! (f buf len))
671 skip_one_varid :: (FastString -> Token) -> Action
672 skip_one_varid f span buf len
673 = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
675 strtoken :: (String -> Token) -> Action
676 strtoken f span buf len =
677 return (L span $! (f $! lexemeToString buf len))
679 init_strtoken :: Int -> (String -> Token) -> Action
680 -- like strtoken, but drops the last N character(s)
681 init_strtoken drop f span buf len =
682 return (L span $! (f $! lexemeToString buf (len-drop)))
684 begin :: Int -> Action
685 begin code _span _str _len = do pushLexState code; lexToken
688 pop _span _buf _len = do popLexState; lexToken
690 pop_and :: Action -> Action
691 pop_and act span buf len = do popLexState; act span buf len
693 {-# INLINE nextCharIs #-}
694 nextCharIs buf p = not (atEnd buf) && p (currentChar buf)
696 notFollowedBy char _ _ _ (AI _ _ buf)
697 = nextCharIs buf (/=char)
699 notFollowedBySymbol _ _ _ (AI _ _ buf)
700 = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
702 isNormalComment bits _ _ (AI _ _ buf)
703 = nextCharIs buf (/='#')
705 haddockDisabledAnd p bits _ _ (AI _ _ buf)
706 = if haddockEnabled bits then False else (p buf)
708 atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
710 ifExtension pred bits _ _ _ = pred bits
712 multiline_doc_comment :: Action
713 multiline_doc_comment span buf _len = withLexedDocType (worker "")
715 worker commentAcc input docType oneLine = case alexGetChar input of
717 | oneLine -> docCommentEnd input commentAcc docType buf span
718 | otherwise -> case checkIfCommentLine input' of
719 Just input -> worker ('\n':commentAcc) input docType False
720 Nothing -> docCommentEnd input commentAcc docType buf span
721 Just (c, input) -> worker (c:commentAcc) input docType oneLine
722 Nothing -> docCommentEnd input commentAcc docType buf span
724 checkIfCommentLine input = check (dropNonNewlineSpace input)
726 check input = case alexGetChar input of
727 Just ('-', input) -> case alexGetChar input of
728 Just ('-', input) -> case alexGetChar input of
729 Just (c, _) | c /= '-' -> Just input
734 dropNonNewlineSpace input = case alexGetChar input of
736 | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
741 nested comments require traversing by hand, they can't be parsed
742 using regular expressions.
744 nested_comment :: P (Located Token) -> Action
745 nested_comment cont span _str _len = do
749 go 0 input = do setInput input; cont
750 go n input = case alexGetChar input of
751 Nothing -> errBrace input span
752 Just ('-',input) -> case alexGetChar input of
753 Nothing -> errBrace input span
754 Just ('\125',input) -> go (n-1) input
755 Just (c,_) -> go n input
756 Just ('\123',input) -> case alexGetChar input of
757 Nothing -> errBrace input span
758 Just ('-',input) -> go (n+1) input
759 Just (c,_) -> go n input
760 Just (c,input) -> go n input
762 nested_doc_comment :: Action
763 nested_doc_comment span buf _len = withLexedDocType (go "")
765 go commentAcc input docType _ = case alexGetChar input of
766 Nothing -> errBrace input span
767 Just ('-',input) -> case alexGetChar input of
768 Nothing -> errBrace input span
769 Just ('\125',input@(AI end _ buf2)) ->
770 docCommentEnd input commentAcc docType buf span
771 Just (c,_) -> go ('-':commentAcc) input docType False
772 Just ('\123', input) -> case alexGetChar input of
773 Nothing -> errBrace input span
774 Just ('-',input) -> do
776 let cont = do input <- getInput; go commentAcc input docType False
777 nested_comment cont span buf _len
778 Just (c,_) -> go ('\123':commentAcc) input docType False
779 Just (c,input) -> go (c:commentAcc) input docType False
781 withLexedDocType lexDocComment = do
782 input@(AI _ _ buf) <- getInput
783 case prevChar buf ' ' of
784 '|' -> lexDocComment input ITdocCommentNext False
785 '^' -> lexDocComment input ITdocCommentPrev False
786 '$' -> lexDocComment input ITdocCommentNamed False
787 '*' -> lexDocSection 1 input
789 lexDocSection n input = case alexGetChar input of
790 Just ('*', input) -> lexDocSection (n+1) input
791 Just (c, _) -> lexDocComment input (ITdocSection n) True
792 Nothing -> do setInput input; lexToken -- eof reached, lex it normally
795 -------------------------------------------------------------------------------
796 -- This function is quite tricky. We can't just return a new token, we also
797 -- need to update the state of the parser. Why? Because the token is longer
798 -- than what was lexed by Alex, and the lexToken function doesn't know this, so
799 -- it writes the wrong token length to the parser state. This function is
800 -- called afterwards, so it can just update the state.
802 -- This is complicated by the fact that Haddock tokens can span multiple lines,
803 -- which is something that the original lexer didn't account for.
804 -- I have added last_line_len in the parser state which represents the length
805 -- of the part of the token that is on the last line. It is now used for layout
806 -- calculation in pushCurrentContext instead of last_len. last_len is, like it
807 -- was before, the full length of the token, and it is now only used for error
810 docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
811 SrcSpan -> P (Located Token)
812 docCommentEnd input commentAcc docType buf span = do
814 let (AI loc last_offs nextBuf) = input
815 comment = reverse commentAcc
816 span' = mkSrcSpan (srcSpanStart span) loc
817 last_len = byteDiff buf nextBuf
819 last_line_len = if (last_offs - last_len < 0)
823 span `seq` setLastToken span' last_len last_line_len
824 return (L span' (docType comment))
826 errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
828 open_brace, close_brace :: Action
829 open_brace span _str _len = do
831 setContext (NoLayout:ctx)
832 return (L span ITocurly)
833 close_brace span _str _len = do
835 return (L span ITccurly)
837 -- We have to be careful not to count M.<varid> as a qualified name
838 -- when <varid> is a keyword. We hack around this by catching
839 -- the offending tokens afterward, and re-lexing in a different state.
840 check_qvarid span buf len = do
841 case lookupUFM reservedWordsFM var of
843 | not (isSpecial keyword) ->
847 b <- extension (\i -> exts .&. i /= 0)
850 _other -> return token
852 (mod,var) = splitQualName buf len
853 token = L span (ITqvarid (mod,var))
856 (AI _ offs _) <- getInput
857 setInput (AI (srcSpanStart span) (offs-len) buf)
858 pushLexState bad_qvarid
861 qvarid buf len = ITqvarid $! splitQualName buf len
862 qconid buf len = ITqconid $! splitQualName buf len
864 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
865 -- takes a StringBuffer and a length, and returns the module name
866 -- and identifier parts of a qualified name. Splits at the *last* dot,
867 -- because of hierarchical module names.
868 splitQualName orig_buf len = split orig_buf orig_buf
871 | orig_buf `byteDiff` buf >= len = done dot_buf
872 | c == '.' = found_dot buf'
873 | otherwise = split buf' dot_buf
875 (c,buf') = nextChar buf
877 -- careful, we might get names like M....
878 -- so, if the character after the dot is not upper-case, this is
879 -- the end of the qualifier part.
880 found_dot buf -- buf points after the '.'
881 | isUpper c = split buf' buf
882 | otherwise = done buf
884 (c,buf') = nextChar buf
887 (lexemeToFastString orig_buf (qual_size - 1),
888 lexemeToFastString dot_buf (len - qual_size))
890 qual_size = orig_buf `byteDiff` dot_buf
893 case lookupUFM reservedWordsFM fs of
894 Just (keyword,0) -> do
896 return (L span keyword)
897 Just (keyword,exts) -> do
898 b <- extension (\i -> exts .&. i /= 0)
899 if b then do maybe_layout keyword
900 return (L span keyword)
901 else return (L span (ITvarid fs))
902 _other -> return (L span (ITvarid fs))
904 fs = lexemeToFastString buf len
906 conid buf len = ITconid fs
907 where fs = lexemeToFastString buf len
909 qvarsym buf len = ITqvarsym $! splitQualName buf len
910 qconsym buf len = ITqconsym $! splitQualName buf len
912 varsym = sym ITvarsym
913 consym = sym ITconsym
915 sym con span buf len =
916 case lookupUFM reservedSymsFM fs of
917 Just (keyword,0) -> return (L span keyword)
918 Just (keyword,exts) -> do
919 b <- extension (\i -> exts .&. i /= 0)
920 if b then return (L span keyword)
921 else return (L span $! con fs)
922 _other -> return (L span $! con fs)
924 fs = lexemeToFastString buf len
926 tok_decimal span buf len
927 = return (L span (ITinteger $! parseInteger buf len 10 octDecDigit))
929 tok_octal span buf len
930 = return (L span (ITinteger $! parseInteger (offsetBytes 2 buf) (len-2) 8 octDecDigit))
932 tok_hexadecimal span buf len
933 = return (L span (ITinteger $! parseInteger (offsetBytes 2 buf) (len-2) 16 hexDigit))
935 prim_decimal span buf len
936 = return (L span (ITprimint $! parseInteger buf (len-1) 10 octDecDigit))
938 prim_octal span buf len
939 = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 8 octDecDigit))
941 prim_hexadecimal span buf len
942 = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 16 hexDigit))
944 tok_float str = ITrational $! readRational str
945 prim_float str = ITprimfloat $! readRational str
946 prim_double str = ITprimdouble $! readRational str
948 -- -----------------------------------------------------------------------------
951 -- we're at the first token on a line, insert layout tokens if necessary
953 do_bol span _str _len = do
957 --trace "layout: inserting '}'" $ do
959 -- do NOT pop the lex state, we might have a ';' to insert
960 return (L span ITvccurly)
962 --trace "layout: inserting ';'" $ do
964 return (L span ITsemi)
969 -- certain keywords put us in the "layout" state, where we might
970 -- add an opening curly brace.
971 maybe_layout ITdo = pushLexState layout_do
972 maybe_layout ITmdo = pushLexState layout_do
973 maybe_layout ITof = pushLexState layout
974 maybe_layout ITlet = pushLexState layout
975 maybe_layout ITwhere = pushLexState layout
976 maybe_layout ITrec = pushLexState layout
977 maybe_layout _ = return ()
979 -- Pushing a new implicit layout context. If the indentation of the
980 -- next token is not greater than the previous layout context, then
981 -- Haskell 98 says that the new layout context should be empty; that is
982 -- the lexer must generate {}.
984 -- We are slightly more lenient than this: when the new context is started
985 -- by a 'do', then we allow the new context to be at the same indentation as
986 -- the previous context. This is what the 'strict' argument is for.
988 new_layout_context strict span _buf _len = do
990 (AI _ offset _) <- getInput
993 Layout prev_off : _ |
994 (strict && prev_off >= offset ||
995 not strict && prev_off > offset) -> do
996 -- token is indented to the left of the previous context.
997 -- we must generate a {} sequence now.
998 pushLexState layout_left
999 return (L span ITvocurly)
1001 setContext (Layout offset : ctx)
1002 return (L span ITvocurly)
1004 do_layout_left span _buf _len = do
1006 pushLexState bol -- we must be at the start of a line
1007 return (L span ITvccurly)
1009 -- -----------------------------------------------------------------------------
1012 setLine :: Int -> Action
1013 setLine code span buf len = do
1014 let line = parseInteger buf len 10 octDecDigit
1015 setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
1016 -- subtract one: the line number refers to the *following* line
1021 setFile :: Int -> Action
1022 setFile code span buf len = do
1023 let file = lexemeToFastString (stepOn buf) (len-2)
1024 setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1030 -- -----------------------------------------------------------------------------
1031 -- Options, includes and language pragmas.
1033 lex_string_prag :: (String -> Token) -> Action
1034 lex_string_prag mkTok span buf len
1035 = do input <- getInput
1039 return (L (mkSrcSpan start end) tok)
1041 = if isString input "#-}"
1042 then do setInput input
1043 return (mkTok (reverse acc))
1044 else case alexGetChar input of
1045 Just (c,i) -> go (c:acc) i
1046 Nothing -> err input
1047 isString i [] = True
1049 = case alexGetChar i of
1050 Just (c,i') | c == x -> isString i' xs
1052 err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"
1055 -- -----------------------------------------------------------------------------
1058 -- This stuff is horrible. I hates it.
1060 lex_string_tok :: Action
1061 lex_string_tok span buf len = do
1062 tok <- lex_string ""
1064 return (L (mkSrcSpan (srcSpanStart span) end) tok)
1066 lex_string :: String -> P Token
1069 case alexGetChar' i of
1070 Nothing -> lit_error
1074 glaexts <- extension glaExtsEnabled
1078 case alexGetChar' i of
1082 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1083 else let s' = mkZFastString (reverse s) in
1084 return (ITprimstring s')
1085 -- mkZFastString is a hack to avoid encoding the
1086 -- string in UTF-8. We just want the exact bytes.
1088 return (ITstring (mkFastString (reverse s)))
1090 return (ITstring (mkFastString (reverse s)))
1093 | Just ('&',i) <- next -> do
1094 setInput i; lex_string s
1095 | Just (c,i) <- next, is_space c -> do
1096 setInput i; lex_stringgap s
1097 where next = alexGetChar' i
1103 lex_stringgap s = do
1106 '\\' -> lex_string s
1107 c | is_space c -> lex_stringgap s
1111 lex_char_tok :: Action
1112 -- Here we are basically parsing character literals, such as 'x' or '\n'
1113 -- but, when Template Haskell is on, we additionally spot
1114 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
1115 -- but WIHTOUT CONSUMING the x or T part (the parser does that).
1116 -- So we have to do two characters of lookahead: when we see 'x we need to
1117 -- see if there's a trailing quote
1118 lex_char_tok span buf len = do -- We've seen '
1119 i1 <- getInput -- Look ahead to first character
1120 let loc = srcSpanStart span
1121 case alexGetChar' i1 of
1122 Nothing -> lit_error
1124 Just ('\'', i2@(AI end2 _ _)) -> do -- We've seen ''
1125 th_exts <- extension thEnabled
1128 return (L (mkSrcSpan loc end2) ITtyQuote)
1131 Just ('\\', i2@(AI end2 _ _)) -> do -- We've seen 'backslash
1133 lit_ch <- lex_escape
1134 mc <- getCharOrFail -- Trailing quote
1135 if mc == '\'' then finish_char_tok loc lit_ch
1136 else do setInput i2; lit_error
1138 Just (c, i2@(AI end2 _ _))
1139 | not (isAny c) -> lit_error
1142 -- We've seen 'x, where x is a valid character
1143 -- (i.e. not newline etc) but not a quote or backslash
1144 case alexGetChar' i2 of -- Look ahead one more character
1145 Nothing -> lit_error
1146 Just ('\'', i3) -> do -- We've seen 'x'
1148 finish_char_tok loc c
1149 _other -> do -- We've seen 'x not followed by quote
1150 -- If TH is on, just parse the quote only
1151 th_exts <- extension thEnabled
1152 let (AI end _ _) = i1
1153 if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1154 else do setInput i2; lit_error
1156 finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1157 finish_char_tok loc ch -- We've already seen the closing quote
1158 -- Just need to check for trailing #
1159 = do glaexts <- extension glaExtsEnabled
1160 i@(AI end _ _) <- getInput
1162 case alexGetChar' i of
1163 Just ('#',i@(AI end _ _)) -> do
1165 return (L (mkSrcSpan loc end) (ITprimchar ch))
1167 return (L (mkSrcSpan loc end) (ITchar ch))
1169 return (L (mkSrcSpan loc end) (ITchar ch))
1171 lex_char :: Char -> AlexInput -> P Char
1174 '\\' -> do setInput inp; lex_escape
1175 c | isAny c -> do setInput inp; return c
1178 isAny c | c > '\xff' = isPrint c
1179 | otherwise = is_any c
1181 lex_escape :: P Char
1195 '^' -> do c <- getCharOrFail
1196 if c >= '@' && c <= '_'
1197 then return (chr (ord c - ord '@'))
1200 'x' -> readNum is_hexdigit 16 hexDigit
1201 'o' -> readNum is_octdigit 8 octDecDigit
1202 x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x)
1206 case alexGetChar' i of
1207 Nothing -> lit_error
1209 case alexGetChar' i2 of
1210 Nothing -> do setInput i2; lit_error
1212 let str = [c1,c2,c3] in
1213 case [ (c,rest) | (p,c) <- silly_escape_chars,
1214 Just rest <- [maybePrefixMatch p str] ] of
1215 (escape_char,[]):_ -> do
1218 (escape_char,_:_):_ -> do
1223 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
1224 readNum is_digit base conv = do
1228 then readNum2 is_digit base conv (conv c)
1229 else do setInput i; lit_error
1231 readNum2 is_digit base conv i = do
1234 where read i input = do
1235 case alexGetChar' input of
1236 Just (c,input') | is_digit c -> do
1237 read (i*base + conv c) input'
1239 if i >= 0 && i <= 0x10FFFF
1240 then do setInput input; return (chr i)
1243 silly_escape_chars = [
1280 -- before calling lit_error, ensure that the current input is pointing to
1281 -- the position of the error in the buffer. This is so that we can report
1282 -- a correct location to the user, but also so we can detect UTF-8 decoding
1283 -- errors if they occur.
1284 lit_error = lexError "lexical error in string/character literal"
1286 getCharOrFail :: P Char
1289 case alexGetChar' i of
1290 Nothing -> lexError "unexpected end-of-file in string/character literal"
1291 Just (c,i) -> do setInput i; return c
1293 -- -----------------------------------------------------------------------------
1296 warn :: DynFlag -> SDoc -> Action
1297 warn option warning span _buf _len = do
1298 addWarning option (mkWarnMsg span alwaysQualify warning)
1301 -- -----------------------------------------------------------------------------
1312 SrcSpan -- The start and end of the text span related to
1313 -- the error. Might be used in environments which can
1314 -- show this span, e.g. by highlighting it.
1315 Message -- The error message
1317 data PState = PState {
1318 buffer :: StringBuffer,
1320 messages :: Messages,
1321 last_loc :: SrcSpan, -- pos of previous token
1322 last_offs :: !Int, -- offset of the previous token from the
1323 -- beginning of the current line.
1324 -- \t is equal to 8 spaces.
1325 last_len :: !Int, -- len of previous token
1326 last_line_len :: !Int,
1327 loc :: SrcLoc, -- current loc (end of prev token + 1)
1328 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1329 context :: [LayoutContext],
1332 -- last_loc and last_len are used when generating error messages,
1333 -- and in pushCurrentContext only. Sigh, if only Happy passed the
1334 -- current token to happyError, we could at least get rid of last_len.
1335 -- Getting rid of last_loc would require finding another way to
1336 -- implement pushCurrentContext (which is only called from one place).
1338 newtype P a = P { unP :: PState -> ParseResult a }
1340 instance Monad P where
1346 returnP a = P $ \s -> POk s a
1348 thenP :: P a -> (a -> P b) -> P b
1349 (P m) `thenP` k = P $ \ s ->
1351 POk s1 a -> (unP (k a)) s1
1352 PFailed span err -> PFailed span err
1354 failP :: String -> P a
1355 failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1357 failMsgP :: String -> P a
1358 failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1360 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1361 failLocMsgP loc1 loc2 str = P $ \s -> PFailed (mkSrcSpan loc1 loc2) (text str)
1363 failSpanMsgP :: SrcSpan -> String -> P a
1364 failSpanMsgP span msg = P $ \s -> PFailed span (text msg)
1366 extension :: (Int -> Bool) -> P Bool
1367 extension p = P $ \s -> POk s (p $! extsBitmap s)
1370 getExts = P $ \s -> POk s (extsBitmap s)
1372 setSrcLoc :: SrcLoc -> P ()
1373 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1375 getSrcLoc :: P SrcLoc
1376 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1378 setLastToken :: SrcSpan -> Int -> Int -> P ()
1379 setLastToken loc len line_len = P $ \s -> POk s {
1382 last_line_len=line_len
1385 data AlexInput = AI SrcLoc {-#UNPACK#-}!Int StringBuffer
1387 alexInputPrevChar :: AlexInput -> Char
1388 alexInputPrevChar (AI _ _ buf) = prevChar buf '\n'
1390 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1391 alexGetChar (AI loc ofs s)
1393 | otherwise = adj_c `seq` loc' `seq` ofs' `seq` s' `seq`
1394 --trace (show (ord c)) $
1395 Just (adj_c, (AI loc' ofs' s'))
1396 where (c,s') = nextChar s
1397 loc' = advanceSrcLoc loc c
1398 ofs' = advanceOffs c ofs
1406 other_graphic = '\x6'
1409 | c <= '\x06' = non_graphic
1412 case generalCategory c of
1413 UppercaseLetter -> upper
1414 LowercaseLetter -> lower
1415 TitlecaseLetter -> upper
1416 ModifierLetter -> other_graphic
1417 OtherLetter -> other_graphic
1418 NonSpacingMark -> other_graphic
1419 SpacingCombiningMark -> other_graphic
1420 EnclosingMark -> other_graphic
1421 DecimalNumber -> digit
1422 LetterNumber -> other_graphic
1423 OtherNumber -> other_graphic
1424 ConnectorPunctuation -> other_graphic
1425 DashPunctuation -> other_graphic
1426 OpenPunctuation -> other_graphic
1427 ClosePunctuation -> other_graphic
1428 InitialQuote -> other_graphic
1429 FinalQuote -> other_graphic
1430 OtherPunctuation -> other_graphic
1431 MathSymbol -> symbol
1432 CurrencySymbol -> symbol
1433 ModifierSymbol -> symbol
1434 OtherSymbol -> symbol
1436 _other -> non_graphic
1438 -- This version does not squash unicode characters, it is used when
1440 alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
1441 alexGetChar' (AI loc ofs s)
1443 | otherwise = c `seq` loc' `seq` ofs' `seq` s' `seq`
1444 --trace (show (ord c)) $
1445 Just (c, (AI loc' ofs' s'))
1446 where (c,s') = nextChar s
1447 loc' = advanceSrcLoc loc c
1448 ofs' = advanceOffs c ofs
1450 advanceOffs :: Char -> Int -> Int
1451 advanceOffs '\n' offs = 0
1452 advanceOffs '\t' offs = (offs `quot` 8 + 1) * 8
1453 advanceOffs _ offs = offs + 1
1455 getInput :: P AlexInput
1456 getInput = P $ \s@PState{ loc=l, last_offs=o, buffer=b } -> POk s (AI l o b)
1458 setInput :: AlexInput -> P ()
1459 setInput (AI l o b) = P $ \s -> POk s{ loc=l, last_offs=o, buffer=b } ()
1461 pushLexState :: Int -> P ()
1462 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1464 popLexState :: P Int
1465 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1467 getLexState :: P Int
1468 getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls
1470 -- for reasons of efficiency, flags indicating language extensions (eg,
1471 -- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
1474 glaExtsBit, ffiBit, parrBit :: Int
1481 tvBit = 7 -- Scoped type variables enables 'forall' keyword
1482 bangPatBit = 8 -- Tells the parser to understand bang-patterns
1483 -- (doesn't affect the lexer)
1484 idxTysBit = 9 -- indexed type families: 'family' keyword and kind sigs
1485 haddockBit = 10 -- Lex and parse Haddock comments
1487 glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1488 glaExtsEnabled flags = testBit flags glaExtsBit
1489 ffiEnabled flags = testBit flags ffiBit
1490 parrEnabled flags = testBit flags parrBit
1491 arrowsEnabled flags = testBit flags arrowsBit
1492 thEnabled flags = testBit flags thBit
1493 ipEnabled flags = testBit flags ipBit
1494 tvEnabled flags = testBit flags tvBit
1495 bangPatEnabled flags = testBit flags bangPatBit
1496 idxTysEnabled flags = testBit flags idxTysBit
1497 haddockEnabled flags = testBit flags haddockBit
1499 -- PState for parsing options pragmas
1501 pragState :: StringBuffer -> SrcLoc -> PState
1505 messages = emptyMessages,
1506 -- XXX defaultDynFlags is not right, but we don't have a real
1508 dflags = defaultDynFlags,
1509 last_loc = mkSrcSpan loc loc,
1516 lex_state = [bol, option_prags, 0]
1520 -- create a parse state
1522 mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
1523 mkPState buf loc flags =
1527 messages = emptyMessages,
1528 last_loc = mkSrcSpan loc loc,
1533 extsBitmap = fromIntegral bitmap,
1535 lex_state = [bol, if glaExtsEnabled bitmap then glaexts else 0]
1536 -- we begin in the layout state if toplev_layout is set
1539 bitmap = glaExtsBit `setBitIf` dopt Opt_GlasgowExts flags
1540 .|. ffiBit `setBitIf` dopt Opt_FFI flags
1541 .|. parrBit `setBitIf` dopt Opt_PArr flags
1542 .|. arrowsBit `setBitIf` dopt Opt_Arrows flags
1543 .|. thBit `setBitIf` dopt Opt_TH flags
1544 .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
1545 .|. tvBit `setBitIf` dopt Opt_ScopedTypeVariables flags
1546 .|. bangPatBit `setBitIf` dopt Opt_BangPatterns flags
1547 .|. idxTysBit `setBitIf` dopt Opt_IndexedTypes flags
1548 .|. haddockBit `setBitIf` dopt Opt_Haddock flags
1550 setBitIf :: Int -> Bool -> Int
1551 b `setBitIf` cond | cond = bit b
1554 addWarning :: DynFlag -> WarnMsg -> P ()
1556 = P $ \s@PState{messages=(ws,es), dflags=d} ->
1557 let ws' = if dopt option d then ws `snocBag` w else ws
1558 in POk s{messages=(ws', es)} ()
1560 getMessages :: PState -> Messages
1561 getMessages PState{messages=ms} = ms
1563 getContext :: P [LayoutContext]
1564 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1566 setContext :: [LayoutContext] -> P ()
1567 setContext ctx = P $ \s -> POk s{context=ctx} ()
1570 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1571 loc = loc, last_len = len, last_loc = last_loc }) ->
1573 (_:tl) -> POk s{ context = tl } ()
1574 [] -> PFailed last_loc (srcParseErr buf len)
1576 -- Push a new layout context at the indentation of the last token read.
1577 -- This is only used at the outer level of a module when the 'module'
1578 -- keyword is missing.
1579 pushCurrentContext :: P ()
1580 pushCurrentContext = P $ \ s@PState{ last_offs=offs, last_line_len=len, context=ctx } ->
1581 POk s{context = Layout (offs-len) : ctx} ()
1582 --trace ("off: " ++ show offs ++ ", len: " ++ show len) $ POk s{context = Layout (offs-len) : ctx} ()
1584 getOffside :: P Ordering
1585 getOffside = P $ \s@PState{last_offs=offs, context=stk} ->
1586 let ord = case stk of
1587 (Layout n:_) -> compare offs n
1591 -- ---------------------------------------------------------------------------
1592 -- Construct a parse error
1595 :: StringBuffer -- current buffer (placed just after the last token)
1596 -> Int -- length of the previous token
1599 = hcat [ if null token
1600 then ptext SLIT("parse error (possibly incorrect indentation)")
1601 else hcat [ptext SLIT("parse error on input "),
1602 char '`', text token, char '\'']
1604 where token = lexemeToString (offsetBytes (-len) buf) len
1606 -- Report a parse failure, giving the span of the previous token as
1607 -- the location of the error. This is the entry point for errors
1608 -- detected during parsing.
1610 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1611 last_loc = last_loc } ->
1612 PFailed last_loc (srcParseErr buf len)
1614 -- A lexical error is reported at a particular position in the source file,
1615 -- not over a token range.
1616 lexError :: String -> P a
1619 i@(AI end _ buf) <- getInput
1620 reportLexError loc end buf str
1622 -- -----------------------------------------------------------------------------
1623 -- This is the top-level function: called from the parser each time a
1624 -- new token is to be read from the input.
1626 lexer :: (Located Token -> P a) -> P a
1628 tok@(L span tok__) <- lexToken
1629 -- trace ("token: " ++ show tok__) $ do
1632 lexToken :: P (Located Token)
1634 inp@(AI loc1 _ buf) <- getInput
1637 case alexScanUser exts inp sc of
1638 AlexEOF -> do let span = mkSrcSpan loc1 loc1
1639 setLastToken span 0 0
1640 return (L span ITeof)
1641 AlexError (AI loc2 _ buf) -> do
1642 reportLexError loc1 loc2 buf "lexical error"
1643 AlexSkip inp2 _ -> do
1646 AlexToken inp2@(AI end _ buf2) len t -> do
1648 let span = mkSrcSpan loc1 end
1649 let bytes = byteDiff buf buf2
1650 span `seq` setLastToken span bytes bytes
1653 reportLexError loc1 loc2 buf str
1654 | atEnd buf = failLocMsgP loc1 loc2 (str ++ " at end of input")
1657 c = fst (nextChar buf)
1659 if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar#
1660 then failLocMsgP loc2 loc2 (str ++ " (UTF-8 decoding error)")
1661 else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c)