1 -----------------------------------------------------------------------------
2 -- (c) The University of Glasgow, 2003
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(..), Token__(..), lexer, mkPState, showPFailed,
26 P(..), ParseResult(..), setSrcLocFor, getSrcLoc,
27 failLocMsgP, srcParseFail,
28 popContext, pushCurrentContext,
31 #include "HsVersions.h"
33 import ForeignCall ( Safety(..) )
34 import ErrUtils ( Message )
43 import Util ( maybePrefixMatch )
51 $whitechar = [\ \t\n\r\f\v\xa0]
52 $white_no_nl = $whitechar # \n
56 $digit = [$ascdigit $unidigit]
58 $special = [\(\)\,\;\[\]\`\{\}]
59 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
61 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
64 $asclarge = [A-Z \xc0-\xd6 \xd8-\xde]
65 $large = [$asclarge $unilarge]
68 $ascsmall = [a-z \xdf-\xf6 \xf8-\xff]
69 $small = [$ascsmall $unismall \_]
71 $graphic = [$small $large $symbol $digit $special \:\"\']
74 $hexit = [$digit A-F a-f]
75 $symchar = [$symbol \:]
77 $idchar = [$small $large $digit \']
79 @varid = $small $idchar*
80 @conid = $large $idchar*
82 @varsym = $symbol $symchar*
83 @consym = \: $symchar*
87 @hexadecimal = $hexit+
88 @exponent = [eE] [\-\+]? @decimal
90 -- we support the hierarchical module name extension:
93 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
97 -- everywhere: skip whitespace and comments
100 -- Everywhere: deal with nested comments. We explicitly rule out
101 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
102 -- (this can happen even though pragmas will normally take precedence due to
103 -- longest-match, because pragmas aren't valid in every state, but comments
105 "{-" / { notFollowedBy '#' } { nested_comment }
107 -- Single-line comments are a bit tricky. Haskell 98 says that two or
108 -- more dashes followed by a symbol should be parsed as a varsym, so we
109 -- have to exclude those.
110 -- The regex says: "munch all the characters after the dashes, as long as
111 -- the first one is not a symbol".
112 "--"\-* ([^$symbol] .*)? ;
114 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
115 -- blank lines) until we find a non-whitespace character, then do layout
118 -- One slight wibble here: what if the line begins with {-#? In
119 -- theory, we have to lex the pragma to see if it's one we recognise,
120 -- and if it is, then we backtrack and do_bol, otherwise we treat it
121 -- as a nested comment. We don't bother with this: if the line begins
122 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
125 ^\# (line)? { begin line_prag1 }
126 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
130 -- after a layout keyword (let, where, do, of), we begin a new layout
131 -- context if the curly brace is missing.
132 -- Careful! This stuff is quite delicate.
133 <layout, layout_do> {
134 \{ / { notFollowedBy '-' } { pop_and open_brace }
135 -- we might encounter {-# here, but {- has been handled already
137 ^\# (line)? { begin line_prag1 }
140 -- do is treated in a subtly different way, see new_layout_context
141 <layout> () { new_layout_context True }
142 <layout_do> () { new_layout_context False }
144 -- after a new layout context which was found to be to the left of the
145 -- previous context, we have generated a '{' token, and we now need to
146 -- generate a matching '}' token.
147 <layout_left> () { do_layout_left }
149 <0,glaexts> \n { begin bol }
151 "{-#" $whitechar* (line|LINE) { begin line_prag2 }
153 -- single-line line pragmas, of the form
154 -- # <line> "<file>" <extra-stuff> \n
155 <line_prag1> $digit+ { set_line line_prag1a }
156 <line_prag1a> \" [$graphic \ ]* \" { set_file line_prag1b }
157 <line_prag1b> .* { pop }
159 -- Haskell-style line pragmas, of the form
160 -- {-# LINE <line> "<file>" #-}
161 <line_prag2> $digit+ { set_line line_prag2a }
162 <line_prag2a> \" [$graphic \ ]* \" { set_file line_prag2b }
163 <line_prag2b> "#-}" { pop }
166 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
167 { token ITspecialise_prag }
168 "{-#" $whitechar* (SOURCE|source) { token ITsource_prag }
169 "{-#" $whitechar* (INLINE|inline) { token ITinline_prag }
170 "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
171 { token ITnoinline_prag }
172 "{-#" $whitechar* (RULES|rules) { token ITrules_prag }
173 "{-#" $whitechar* (DEPRECATED|deprecated)
174 { token ITdeprecated_prag }
175 "{-#" $whitechar* (SCC|scc) { token ITscc_prag }
176 "{-#" $whitechar* (CORE|core) { token ITcore_prag }
177 "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag }
179 "{-#" { nested_comment }
181 -- ToDo: should only be valid inside a pragma:
182 "#-}" { token ITclose_prag}
186 -- '0' state: ordinary lexemes
187 -- 'glaexts' state: glasgow extensions (postfix '#', etc.)
192 "[:" / { ifExtension parrEnabled } { token ITopabrack }
193 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
197 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
198 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
199 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
200 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
201 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
202 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
203 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
204 "$(" / { ifExtension thEnabled } { token ITparenEscape }
208 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
209 { special IToparenbar }
210 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
214 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
215 \% @varid / { ifExtension ipEnabled } { skip_one_varid ITsplitipvarid }
219 "(#" / { notFollowedBySymbol } { token IToubxparen }
220 "#)" { token ITcubxparen }
221 "{|" { token ITocurlybar }
222 "|}" { token ITccurlybar }
226 \( { special IToparen }
227 \) { special ITcparen }
228 \[ { special ITobrack }
229 \] { special ITcbrack }
230 \, { special ITcomma }
231 \; { special ITsemi }
232 \` { special ITbackquote }
239 @qual @varid { check_qvarid }
240 @qual @conid { idtoken qconid }
242 @conid { idtoken conid }
245 -- after an illegal qvarid, such as 'M.let',
246 -- we back up and try again in the bad_qvarid state:
248 @conid { pop_and (idtoken conid) }
249 @qual @conid { pop_and (idtoken qconid) }
253 @qual @varid "#"+ { idtoken qvarid }
254 @qual @conid "#"+ { idtoken qconid }
255 @varid "#"+ { varid }
256 @conid "#"+ { idtoken conid }
262 @qual @varsym { idtoken qvarsym }
263 @qual @consym { idtoken qconsym }
269 @decimal { tok_decimal }
270 0[oO] @octal { tok_octal }
271 0[xX] @hexadecimal { tok_hexadecimal }
275 @decimal \# { prim_decimal }
276 0[oO] @octal \# { prim_octal }
277 0[xX] @hexadecimal \# { prim_hexadecimal }
280 <0,glaexts> @floating_point { strtoken tok_float }
281 <glaexts> @floating_point \# { init_strtoken 1 prim_float }
282 <glaexts> @floating_point \# \# { init_strtoken 2 prim_double }
284 -- Strings and chars are lexed by hand-written code. The reason is
285 -- that even if we recognise the string or char here in the regex
286 -- lexer, we would still have to parse the string afterward in order
287 -- to convert it to a String.
290 \" { lex_string_tok }
294 -- work around bug in Alex 2.0
295 #if __GLASGOW_HASKELL__ < 503
296 unsafeAt arr i = arr ! i
299 -- -----------------------------------------------------------------------------
302 data Token = T SrcLoc{-start-} SrcLoc{-end-} Token__
305 = ITas -- Haskell keywords
329 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
331 | ITforall -- GHC extension keywords
344 | ITspecialise_prag -- Pragmas
352 | ITcore_prag -- hdaume: core annotations
356 | ITdotdot -- reserved symbols
372 | ITbiglam -- GHC-extension symbols
374 | ITocurly -- special symbols
376 | ITocurlybar -- {|, for type applications
377 | ITccurlybar -- |}, for type applications
381 | ITopabrack -- [:, for parallel arrays with -fparr
382 | ITcpabrack -- :], for parallel arrays with -fparr
393 | ITvarid FastString -- identifiers
395 | ITvarsym FastString
396 | ITconsym FastString
397 | ITqvarid (FastString,FastString)
398 | ITqconid (FastString,FastString)
399 | ITqvarsym (FastString,FastString)
400 | ITqconsym (FastString,FastString)
402 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
403 | ITsplitipvarid FastString -- GHC extension: implicit param: %x
405 | ITpragma StringBuffer
408 | ITstring FastString
410 | ITrational Rational
413 | ITprimstring FastString
415 | ITprimfloat Rational
416 | ITprimdouble Rational
418 -- MetaHaskell extension tokens
419 | ITopenExpQuote -- [| or [e|
420 | ITopenPatQuote -- [p|
421 | ITopenDecQuote -- [d|
422 | ITopenTypQuote -- [t|
424 | ITidEscape FastString -- $x
425 | ITparenEscape -- $(
429 -- Arrow notation extension
436 | ITLarrowtail -- -<<
437 | ITRarrowtail -- >>-
439 | ITunknown String -- Used when the lexer can't make sense of it
440 | ITeof -- end of file token
442 deriving Show -- debugging
445 isSpecial :: Token__ -> Bool
446 -- If we see M.x, where x is a keyword, but
447 -- is special, we treat is as just plain M.x,
449 isSpecial ITas = True
450 isSpecial IThiding = True
451 isSpecial ITqualified = True
452 isSpecial ITforall = True
453 isSpecial ITexport = True
454 isSpecial ITlabel = True
455 isSpecial ITdynamic = True
456 isSpecial ITsafe = True
457 isSpecial ITthreadsafe = True
458 isSpecial ITunsafe = True
459 isSpecial ITccallconv = True
460 isSpecial ITstdcallconv = True
461 isSpecial ITmdo = True
464 -- the bitmap provided as the third component indicates whether the
465 -- corresponding extension keyword is valid under the extension options
466 -- provided to the compiler; if the extension corresponding to *any* of the
467 -- bits set in the bitmap is enabled, the keyword is valid (this setup
468 -- facilitates using a keyword in two different extensions that can be
469 -- activated independently)
471 reservedWordsFM = listToUFM $
472 map (\(x, y, z) -> (mkFastString x, (y, z)))
473 [( "_", ITunderscore, 0 ),
475 ( "case", ITcase, 0 ),
476 ( "class", ITclass, 0 ),
477 ( "data", ITdata, 0 ),
478 ( "default", ITdefault, 0 ),
479 ( "deriving", ITderiving, 0 ),
481 ( "else", ITelse, 0 ),
482 ( "hiding", IThiding, 0 ),
484 ( "import", ITimport, 0 ),
486 ( "infix", ITinfix, 0 ),
487 ( "infixl", ITinfixl, 0 ),
488 ( "infixr", ITinfixr, 0 ),
489 ( "instance", ITinstance, 0 ),
491 ( "module", ITmodule, 0 ),
492 ( "newtype", ITnewtype, 0 ),
494 ( "qualified", ITqualified, 0 ),
495 ( "then", ITthen, 0 ),
496 ( "type", ITtype, 0 ),
497 ( "where", ITwhere, 0 ),
498 ( "_scc_", ITscc, 0 ), -- ToDo: remove
500 ( "forall", ITforall, bit glaExtsBit),
501 ( "mdo", ITmdo, bit glaExtsBit),
503 ( "foreign", ITforeign, bit ffiBit),
504 ( "export", ITexport, bit ffiBit),
505 ( "label", ITlabel, bit ffiBit),
506 ( "dynamic", ITdynamic, bit ffiBit),
507 ( "safe", ITsafe, bit ffiBit),
508 ( "threadsafe", ITthreadsafe, bit ffiBit),
509 ( "unsafe", ITunsafe, bit ffiBit),
510 ( "stdcall", ITstdcallconv, bit ffiBit),
511 ( "ccall", ITccallconv, bit ffiBit),
512 ( "dotnet", ITdotnet, bit ffiBit),
514 ( "rec", ITrec, bit arrowsBit),
515 ( "proc", ITproc, bit arrowsBit)
518 reservedSymsFM = listToUFM $
519 map (\ (x,y,z) -> (mkFastString x,(y,z)))
520 [ ("..", ITdotdot, 0)
521 ,(":", ITcolon, 0) -- (:) is a reserved op,
522 -- meaning only list cons
535 ,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT
536 ,(".", ITdot, bit glaExtsBit) -- For 'forall a . t'
538 ,("-<", ITlarrowtail, bit arrowsBit)
539 ,(">-", ITrarrowtail, bit arrowsBit)
540 ,("-<<", ITLarrowtail, bit arrowsBit)
541 ,(">>-", ITRarrowtail, bit arrowsBit)
544 -- -----------------------------------------------------------------------------
547 type Action = SrcLoc -> SrcLoc -> StringBuffer -> Int -> P Token
549 special :: Token__ -> Action
550 special tok loc end _buf len = return (T loc end tok)
552 token, layout_token :: Token__ -> Action
553 token t loc end buf len = return (T loc end t)
554 layout_token t loc end buf len = pushLexState layout >> return (T loc end t)
556 idtoken :: (StringBuffer -> Int -> Token__) -> Action
557 idtoken f loc end buf len = return (T loc end $! (f buf len))
559 skip_one_varid :: (FastString -> Token__) -> Action
560 skip_one_varid f loc end buf len
561 = return (T loc end $! f (lexemeToFastString (stepOn buf) (len-1)))
563 strtoken :: (String -> Token__) -> Action
564 strtoken f loc end buf len =
565 return (T loc end $! (f $! lexemeToString buf len))
567 init_strtoken :: Int -> (String -> Token__) -> Action
568 -- like strtoken, but drops the last N character(s)
569 init_strtoken drop f loc end buf len =
570 return (T loc end $! (f $! lexemeToString buf (len-drop)))
572 begin :: Int -> Action
573 begin code _loc _end _str _len = do pushLexState code; lexToken
576 pop _loc _end _buf _len = do popLexState; lexToken
578 pop_and :: Action -> Action
579 pop_and act loc end buf len = do popLexState; act loc end buf len
581 notFollowedBy char _ _ _ (_,buf) = atEnd buf || currentChar buf /= char
583 notFollowedBySymbol _ _ _ (_,buf)
584 = atEnd buf || currentChar buf `notElem` "!#$%&*+./<=>?@\\^|-~"
586 ifExtension pred bits _ _ _ = pred bits
589 nested comments require traversing by hand, they can't be parsed
590 using regular expressions.
592 nested_comment :: Action
593 nested_comment loc _end _str _len = do
596 where go 0 input = do setInput input; lexToken
598 case alexGetChar input of
603 case alexGetChar input of
605 Just ('\125',input) -> go (n-1) input
606 Just (c,_) -> go n input
608 case alexGetChar input of
610 Just ('-',input') -> go (n+1) input'
611 Just (c,input) -> go n input
614 err input = do failLocMsgP loc (fst input) "unterminated `{-'"
616 open_brace, close_brace :: Action
617 open_brace loc end _str _len = do
619 setContext (NoLayout:ctx)
620 return (T loc end ITocurly)
621 close_brace loc end _str _len = do
623 return (T loc end ITccurly)
625 -- We have to be careful not to count M.<varid> as a qualified name
626 -- when <varid> is a keyword. We hack around this by catching
627 -- the offending tokens afterward, and re-lexing in a different state.
628 check_qvarid loc end buf len = do
629 case lookupUFM reservedWordsFM var of
631 | not (isSpecial keyword) ->
635 b <- extension (\i -> exts .&. i /= 0)
638 _other -> return token
640 (mod,var) = splitQualName buf len
641 token = T loc end (ITqvarid (mod,var))
645 pushLexState bad_qvarid
648 qvarid buf len = ITqvarid $! splitQualName buf len
649 qconid buf len = ITqconid $! splitQualName buf len
651 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
652 -- takes a StringBuffer and a length, and returns the module name
653 -- and identifier parts of a qualified name. Splits at the *last* dot,
654 -- because of hierarchical module names.
655 splitQualName orig_buf len = split orig_buf 0 0
658 | n == len = done dot_off
659 | lookAhead buf n == '.' = split2 buf n (n+1)
660 | otherwise = split buf dot_off (n+1)
662 -- careful, we might get names like M....
663 -- so, if the character after the dot is not upper-case, this is
664 -- the end of the qualifier part.
666 | isUpper (lookAhead buf n) = split buf dot_off (n+1)
667 | otherwise = done dot_off
670 (lexemeToFastString orig_buf dot_off,
671 lexemeToFastString (stepOnBy (dot_off+1) orig_buf) (len - dot_off -1))
673 varid loc end buf len =
674 case lookupUFM reservedWordsFM fs of
675 Just (keyword,0) -> do
677 return (T loc end keyword)
678 Just (keyword,exts) -> do
679 b <- extension (\i -> exts .&. i /= 0)
680 if b then do maybe_layout keyword
681 return (T loc end keyword)
682 else return (T loc end (ITvarid fs))
683 _other -> return (T loc end (ITvarid fs))
685 fs = lexemeToFastString buf len
687 conid buf len = ITconid fs
688 where fs = lexemeToFastString buf len
690 qvarsym buf len = ITqvarsym $! splitQualName buf len
691 qconsym buf len = ITqconsym $! splitQualName buf len
693 varsym = sym ITvarsym
694 consym = sym ITconsym
696 sym con loc end buf len =
697 case lookupUFM reservedSymsFM fs of
698 Just (keyword,0) -> return (T loc end keyword)
699 Just (keyword,exts) -> do
700 b <- extension (\i -> exts .&. i /= 0)
701 if b then return (T loc end keyword)
702 else return (T loc end $! con fs)
703 _other -> return (T loc end $! con fs)
705 fs = lexemeToFastString buf len
707 tok_decimal loc end buf len
708 = return (T loc end (ITinteger $! parseInteger buf len 10 oct_or_dec))
710 tok_octal loc end buf len
711 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 8 oct_or_dec))
713 tok_hexadecimal loc end buf len
714 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 16 hex))
716 prim_decimal loc end buf len
717 = return (T loc end (ITprimint $! parseInteger buf (len-1) 10 oct_or_dec))
719 prim_octal loc end buf len
720 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 8 oct_or_dec))
722 prim_hexadecimal loc end buf len
723 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 16 hex))
725 tok_float str = ITrational $! readRational__ str
726 prim_float str = ITprimfloat $! readRational__ str
727 prim_double str = ITprimdouble $! readRational__ str
729 parseInteger :: StringBuffer -> Int -> Integer -> (Char->Int) -> Integer
730 parseInteger buf len radix to_int
732 where go i x | i == len = x
733 | otherwise = go (i+1) (x * radix + toInteger (to_int (lookAhead buf i)))
735 -- -----------------------------------------------------------------------------
738 -- we're at the first token on a line, insert layout tokens if necessary
740 do_bol loc end _str _len = do
741 pos <- getOffside end
744 --trace "layout: inserting '}'" $ do
746 -- do NOT pop the lex state, we might have a ';' to insert
747 return (T loc end ITvccurly)
749 --trace "layout: inserting ';'" $ do
751 return (T loc end ITsemi)
756 -- certain keywords put us in the "layout" state, where we might
757 -- add an opening curly brace.
758 maybe_layout ITdo = pushLexState layout_do
759 maybe_layout ITmdo = pushLexState layout_do
760 maybe_layout ITof = pushLexState layout
761 maybe_layout ITlet = pushLexState layout
762 maybe_layout ITwhere = pushLexState layout
763 maybe_layout ITrec = pushLexState layout
764 maybe_layout _ = return ()
766 -- Pushing a new implicit layout context. If the indentation of the
767 -- next token is not greater than the previous layout context, then
768 -- Haskell 98 says that the new layout context should be empty; that is
769 -- the lexer must generate {}.
771 -- We are slightly more lenient than this: when the new context is started
772 -- by a 'do', then we allow the new context to be at the same indentation as
773 -- the previous context. This is what the 'strict' argument is for.
775 new_layout_context strict loc end _buf _len = do
777 let offset = srcLocCol loc
780 Layout prev_off : _ |
781 (strict && prev_off >= offset ||
782 not strict && prev_off > offset) -> do
783 -- token is indented to the left of the previous context.
784 -- we must generate a {} sequence now.
785 pushLexState layout_left
786 return (T loc end ITvocurly)
788 setContext (Layout offset : ctx)
789 return (T loc end ITvocurly)
791 do_layout_left loc end _buf _len = do
793 pushLexState bol -- we must be at the start of a line
794 return (T loc end ITvccurly)
796 -- -----------------------------------------------------------------------------
799 set_line :: Int -> Action
800 set_line code loc end buf len = do
801 let line = parseInteger buf len 10 oct_or_dec
802 setSrcLoc (mkSrcLoc (srcLocFile end) (fromIntegral line - 1) 0)
803 -- subtract one: the line number refers to the *following* line
808 set_file :: Int -> Action
809 set_file code loc end buf len = do
810 let file = lexemeToFastString (stepOn buf) (len-2)
811 setSrcLoc (mkSrcLoc file (srcLocLine end) (srcLocCol end))
816 -- -----------------------------------------------------------------------------
819 -- This stuff is horrible. I hates it.
821 lex_string_tok :: Action
822 lex_string_tok loc end buf len = do
825 return (T loc end tok)
827 lex_string :: String -> P Token__
830 case alexGetChar i of
835 glaexts <- extension glaExtsEnabled
839 case alexGetChar i of
843 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
844 else let s' = mkFastStringNarrow (reverse s) in
845 -- always a narrow string/byte array
846 return (ITprimstring s')
848 return (ITstring (mkFastString (reverse s)))
850 return (ITstring (mkFastString (reverse s)))
853 | Just ('&',i) <- next -> do
854 setInput i; lex_string s
855 | Just (c,i) <- next, is_space c -> do
856 setInput i; lex_stringgap s
857 where next = alexGetChar i
868 c | is_any c -> return c
875 c | is_space c -> lex_stringgap s
879 lex_char_tok :: Action
880 -- Here we are basically parsing character literals, such as 'x' or '\n'
881 -- but, when Template Haskell is on, we additionally spot
882 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
883 -- but WIHTOUT CONSUMING the x or T part (the parser does that).
884 -- So we have to do two characters of lookahead: when we see 'x we need to
885 -- see if there's a trailing quote
886 lex_char_tok loc _end buf len = do -- We've seen '
887 i1 <- getInput -- Look ahead to first character
888 case alexGetChar i1 of
891 Just ('\'', i2@(end2,_)) -> do -- We've seen ''
892 th_exts <- extension thEnabled
895 return (T loc end2 ITtyQuote)
898 Just ('\\', i2@(end2,_)) -> do -- We've seen 'backslash
901 mc <- getCharOrFail -- Trailing quote
902 if mc == '\'' then finish_char_tok loc lit_ch
905 Just (c, i2@(end2,_)) | not (is_any c) -> lit_error
908 -- We've seen 'x, where x is a valid character
909 -- (i.e. not newline etc) but not a quote or backslash
910 case alexGetChar i2 of -- Look ahead one more character
912 Just ('\'', i3) -> do -- We've seen 'x'
914 finish_char_tok loc c
915 _other -> do -- We've seen 'x not followed by quote
916 -- If TH is on, just parse the quote only
917 th_exts <- extension thEnabled
918 if th_exts then return (T loc (fst i1) ITvarQuote)
921 finish_char_tok :: SrcLoc -> Char -> P Token
922 finish_char_tok loc ch -- We've already seen the closing quote
923 -- Just need to check for trailing #
924 = do glaexts <- extension glaExtsEnabled
926 i@(end,_) <- getInput
927 case alexGetChar i of
928 Just ('#',i@(end,_)) -> do
930 return (T loc end (ITprimchar ch))
932 return (T loc end (ITchar ch))
933 else do end <- getSrcLoc
934 return (T loc end (ITchar ch))
950 '^' -> do c <- getCharOrFail
951 if c >= '@' && c <= '_'
952 then return (chr (ord c - ord '@'))
955 'x' -> readNum is_hexdigit 16 hex
956 'o' -> readNum is_octdigit 8 oct_or_dec
957 x | is_digit x -> readNum2 is_digit 10 oct_or_dec (oct_or_dec x)
961 case alexGetChar i of
964 case alexGetChar i2 of
967 let str = [c1,c2,c3] in
968 case [ (c,rest) | (p,c) <- silly_escape_chars,
969 Just rest <- [maybePrefixMatch p str] ] of
970 (escape_char,[]):_ -> do
973 (escape_char,_:_):_ -> do
978 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
979 readNum is_digit base conv = do
982 then readNum2 is_digit base conv (conv c)
985 readNum2 is_digit base conv i = do
988 where read i input = do
989 case alexGetChar input of
990 Just (c,input') | is_digit c -> do
991 read (i*base + conv c) input'
994 if i >= 0 && i <= 0x10FFFF
1000 || (c >= 'a' && c <= 'f')
1001 || (c >= 'A' && c <= 'F')
1003 hex c | is_digit c = ord c - ord '0'
1004 | otherwise = ord (to_lower c) - ord 'a' + 10
1006 oct_or_dec c = ord c - ord '0'
1008 is_octdigit c = c >= '0' && c <= '7'
1011 | c >= 'A' && c <= 'Z' = chr (ord c - (ord 'A' - ord 'a'))
1014 silly_escape_chars = [
1051 lit_error = lexError "lexical error in string/character literal"
1053 getCharOrFail :: P Char
1056 case alexGetChar i of
1057 Nothing -> lexError "unexpected end-of-file in string/character literal"
1058 Just (c,i) -> do setInput i; return c
1060 -- -----------------------------------------------------------------------------
1063 readRational :: ReadS Rational -- NB: doesn't handle leading "-"
1065 (n,d,s) <- readFix r
1067 return ((n%1)*10^^(k-d), t)
1070 (ds,s) <- lexDecDigits r
1071 (ds',t) <- lexDotDigits s
1072 return (read (ds++ds'), length ds', t)
1074 readExp (e:s) | e `elem` "eE" = readExp' s
1075 readExp s = return (0,s)
1077 readExp' ('+':s) = readDec s
1078 readExp' ('-':s) = do
1081 readExp' s = readDec s
1084 (ds,r) <- nonnull isDigit s
1085 return (foldl1 (\n d -> n * 10 + d) [ ord d - ord '0' | d <- ds ],
1088 lexDecDigits = nonnull isDigit
1090 lexDotDigits ('.':s) = return (span isDigit s)
1091 lexDotDigits s = return ("",s)
1093 nonnull p s = do (cs@(_:_),t) <- return (span p s)
1096 readRational__ :: String -> Rational -- NB: *does* handle a leading "-"
1097 readRational__ top_s
1099 '-' : xs -> - (read_me xs)
1103 = case (do { (x,"") <- readRational s ; return x }) of
1105 [] -> error ("readRational__: no parse:" ++ top_s)
1106 _ -> error ("readRational__: ambiguous parse:" ++ top_s)
1108 -- -----------------------------------------------------------------------------
1118 SrcLoc SrcLoc -- The start and end of the text span related to
1119 -- the error. Might be used in environments which can
1120 -- show this span, e.g. by highlighting it.
1121 Message -- The error message
1123 showPFailed loc1 loc2 err = hcat [ppr loc1, text ": ", err]
1125 data PState = PState {
1126 buffer :: StringBuffer,
1127 last_loc :: SrcLoc, -- pos of previous token
1128 last_len :: !Int, -- len of previous token
1129 loc :: SrcLoc, -- current loc (end of prev token + 1)
1130 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1131 context :: [LayoutContext],
1134 -- last_loc and last_len are used when generating error messages,
1135 -- and in pushCurrentContext only.
1137 newtype P a = P { unP :: PState -> ParseResult a }
1139 instance Monad P where
1145 returnP a = P $ \s -> POk s a
1147 thenP :: P a -> (a -> P b) -> P b
1148 (P m) `thenP` k = P $ \ s ->
1150 POk s1 a -> (unP (k a)) s1
1151 PFailed l1 l2 err -> PFailed l1 l2 err
1153 failP :: String -> P a
1154 failP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1156 failMsgP :: String -> P a
1157 failMsgP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1159 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1160 failLocMsgP loc1 loc2 str = P $ \s -> PFailed loc1 loc2 (text str)
1162 extension :: (Int -> Bool) -> P Bool
1163 extension p = P $ \s -> POk s (p $! extsBitmap s)
1166 getExts = P $ \s -> POk s (extsBitmap s)
1168 setSrcLoc :: SrcLoc -> P ()
1169 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1171 -- tmp, for supporting stuff in RdrHsSyn. The scope better not include
1172 -- any calls to the lexer, because it assumes things about the SrcLoc.
1173 setSrcLocFor :: SrcLoc -> P a -> P a
1174 setSrcLocFor new_loc scope = P $ \s@PState{ loc = old_loc } ->
1175 case unP scope s{loc=new_loc} of
1176 PFailed l1 l2 msg -> PFailed l1 l2 msg
1179 getSrcLoc :: P SrcLoc
1180 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1182 setLastToken :: SrcLoc -> Int -> P ()
1183 setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } ()
1185 type AlexInput = (SrcLoc,StringBuffer)
1187 alexInputPrevChar :: AlexInput -> Char
1188 alexInputPrevChar (_,s) = prevChar s '\n'
1190 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1193 | otherwise = c `seq` loc' `seq` s' `seq` Just (c, (loc', s'))
1194 where c = currentChar s
1195 loc' = advanceSrcLoc loc c
1198 getInput :: P AlexInput
1199 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (l,b)
1201 setInput :: AlexInput -> P ()
1202 setInput (l,b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1204 pushLexState :: Int -> P ()
1205 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1207 popLexState :: P Int
1208 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1210 getLexState :: P Int
1211 getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls
1213 -- for reasons of efficiency, flags indicating language extensions (eg,
1214 -- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
1217 glaExtsBit, ffiBit, parrBit :: Int
1225 glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1226 glaExtsEnabled flags = testBit flags glaExtsBit
1227 ffiEnabled flags = testBit flags ffiBit
1228 parrEnabled flags = testBit flags parrBit
1229 arrowsEnabled flags = testBit flags arrowsBit
1230 thEnabled flags = testBit flags thBit
1231 ipEnabled flags = testBit flags ipBit
1233 -- create a parse state
1235 mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
1236 mkPState buf loc flags =
1242 extsBitmap = fromIntegral bitmap,
1244 lex_state = [bol, if glaExtsEnabled bitmap then glaexts else 0]
1245 -- we begin in the layout state if toplev_layout is set
1248 bitmap = glaExtsBit `setBitIf` dopt Opt_GlasgowExts flags
1249 .|. ffiBit `setBitIf` dopt Opt_FFI flags
1250 .|. parrBit `setBitIf` dopt Opt_PArr flags
1251 .|. arrowsBit `setBitIf` dopt Opt_Arrows flags
1252 .|. thBit `setBitIf` dopt Opt_TH flags
1253 .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
1255 setBitIf :: Int -> Bool -> Int
1256 b `setBitIf` cond | cond = bit b
1259 getContext :: P [LayoutContext]
1260 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1262 setContext :: [LayoutContext] -> P ()
1263 setContext ctx = P $ \s -> POk s{context=ctx} ()
1266 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1267 loc = loc, last_len = len, last_loc = last_loc }) ->
1269 (_:tl) -> POk s{ context = tl } ()
1270 [] -> PFailed last_loc loc (srcParseErr buf len)
1272 -- Push a new layout context at the indentation of the last token read.
1273 -- This is only used at the outer level of a module when the 'module'
1274 -- keyword is missing.
1275 pushCurrentContext :: P ()
1276 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1277 POk s{ context = Layout (srcLocCol loc) : ctx} ()
1279 getOffside :: SrcLoc -> P Ordering
1280 getOffside loc = P $ \s@PState{context=stk} ->
1281 let ord = case stk of
1282 (Layout n:_) -> compare (srcLocCol loc) n
1286 -- ---------------------------------------------------------------------------
1287 -- Construct a parse error
1290 :: StringBuffer -- current buffer (placed just after the last token)
1291 -> Int -- length of the previous token
1294 = hcat [ if null token
1295 then ptext SLIT("parse error (possibly incorrect indentation)")
1296 else hcat [ptext SLIT("parse error on input "),
1297 char '`', text token, char '\'']
1299 where token = lexemeToString (stepOnBy (-len) buf) len
1301 -- Report a parse failure, giving the span of the previous token as
1302 -- the location of the error. This is the entry point for errors
1303 -- detected during parsing.
1305 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1306 last_loc = last_loc, loc = loc } ->
1307 PFailed last_loc loc (srcParseErr buf len)
1309 -- A lexical error is reported at a particular position in the source file,
1310 -- not over a token range. TODO: this is slightly wrong, because we record
1311 -- the error at the character position following the one which caused the
1312 -- error. We should somehow back up by one character.
1313 lexError :: String -> P a
1316 failLocMsgP loc loc str
1318 -- -----------------------------------------------------------------------------
1319 -- This is the top-level function: called from the parser each time a
1320 -- new token is to be read from the input.
1322 lexer :: (Token -> P a) -> P a
1324 tok@(T _ _ tok__) <- lexToken
1325 --trace ("token: " ++ show tok__) $ do
1330 inp@(loc1,buf) <- getInput
1333 case alexScanUser exts inp sc of
1334 AlexEOF -> do setLastToken loc1 0
1335 return (T loc1 loc1 ITeof)
1336 AlexError (loc2,_) -> do failLocMsgP loc1 loc2 "lexical error"
1337 AlexSkip inp2 _ -> do
1340 AlexToken inp2@(end,buf2) len t -> do
1342 setLastToken loc1 len