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 failMsgP, 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]
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 "{-#" $whitechar* (CORE|core) { token ITcore_prag }
178 "{-#" { nested_comment }
180 -- ToDo: should only be valid inside a pragma:
181 "#-}" { token ITclose_prag}
185 -- '0' state: ordinary lexemes
186 -- 'glaexts' state: glasgow extensions (postfix '#', etc.)
191 "[:" / { ifExtension parrEnabled } { token ITopabrack }
192 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
196 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
197 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
198 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
199 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
200 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
201 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
202 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
203 "$(" / { ifExtension thEnabled } { token ITparenEscape }
207 "(|" / { ifExtension arrowsEnabled } { special IToparenbar }
208 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
212 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
213 \% @varid / { ifExtension ipEnabled } { skip_one_varid ITsplitipvarid }
217 "(#" { token IToubxparen }
218 "#)" { token ITcubxparen }
219 "{|" { token ITocurlybar }
220 "|}" { token ITccurlybar }
224 \( { special IToparen }
225 \) { special ITcparen }
226 \[ { special ITobrack }
227 \] { special ITcbrack }
228 \, { special ITcomma }
229 \; { special ITsemi }
230 \` { special ITbackquote }
237 @qual @varid { check_qvarid }
238 @qual @conid { idtoken qconid }
240 @conid { idtoken conid }
243 -- after an illegal qvarid, such as 'M.let',
244 -- we back up and try again in the bad_qvarid state:
246 @conid { pop_and (idtoken conid) }
247 @qual @conid { pop_and (idtoken qconid) }
251 @qual @varid "#"+ { idtoken qvarid }
252 @qual @conid "#"+ { idtoken qconid }
253 @varid "#"+ { varid }
254 @conid "#"+ { idtoken conid }
260 @qual @varsym { idtoken qvarsym }
261 @qual @consym { idtoken qconsym }
267 @decimal { tok_decimal }
268 0[oO] @octal { tok_octal }
269 0[xX] @hexadecimal { tok_hexadecimal }
273 @decimal \# { prim_decimal }
274 0[oO] @octal \# { prim_octal }
275 0[xX] @hexadecimal \# { prim_hexadecimal }
278 <0,glaexts> @floating_point { strtoken tok_float }
279 <glaexts> @floating_point \# { init_strtoken 1 prim_float }
280 <glaexts> @floating_point \# \# { init_strtoken 2 prim_double }
282 -- Strings and chars are lexed by hand-written code. The reason is
283 -- that even if we recognise the string or char here in the regex
284 -- lexer, we would still have to parse the string afterward in order
285 -- to convert it to a String.
288 \" { lex_string_tok }
292 -- work around bug in Alex 2.0
293 #if __GLASGOW_HASKELL__ < 503
294 unsafeAt arr i = arr ! i
297 -- -----------------------------------------------------------------------------
300 data Token = T SrcLoc{-start-} SrcLoc{-end-} Token__
303 = ITas -- Haskell keywords
327 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
329 | ITforall -- GHC extension keywords
342 | ITspecialise_prag -- Pragmas
350 | ITcore_prag -- hdaume: core annotations
353 | ITdotdot -- reserved symbols
369 | ITbiglam -- GHC-extension symbols
371 | ITocurly -- special symbols
373 | ITocurlybar -- {|, for type applications
374 | ITccurlybar -- |}, for type applications
378 | ITopabrack -- [:, for parallel arrays with -fparr
379 | ITcpabrack -- :], for parallel arrays with -fparr
390 | ITvarid FastString -- identifiers
392 | ITvarsym FastString
393 | ITconsym FastString
394 | ITqvarid (FastString,FastString)
395 | ITqconid (FastString,FastString)
396 | ITqvarsym (FastString,FastString)
397 | ITqconsym (FastString,FastString)
399 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
400 | ITsplitipvarid FastString -- GHC extension: implicit param: %x
402 | ITpragma StringBuffer
405 | ITstring FastString
407 | ITrational Rational
410 | ITprimstring FastString
412 | ITprimfloat Rational
413 | ITprimdouble Rational
415 -- MetaHaskell extension tokens
416 | ITopenExpQuote -- [| or [e|
417 | ITopenPatQuote -- [p|
418 | ITopenDecQuote -- [d|
419 | ITopenTypQuote -- [t|
421 | ITidEscape FastString -- $x
422 | ITparenEscape -- $(
427 -- Arrow notation extension
434 | ITLarrowtail -- -<<
435 | ITRarrowtail -- >>-
437 | ITunknown String -- Used when the lexer can't make sense of it
438 | ITeof -- end of file token
440 deriving Show -- debugging
443 isSpecial :: Token__ -> Bool
444 -- If we see M.x, where x is a keyword, but
445 -- is special, we treat is as just plain M.x,
447 isSpecial ITas = True
448 isSpecial IThiding = True
449 isSpecial ITqualified = True
450 isSpecial ITforall = True
451 isSpecial ITexport = True
452 isSpecial ITlabel = True
453 isSpecial ITdynamic = True
454 isSpecial ITsafe = True
455 isSpecial ITthreadsafe = True
456 isSpecial ITunsafe = True
457 isSpecial ITccallconv = True
458 isSpecial ITstdcallconv = True
459 isSpecial ITmdo = True
462 -- the bitmap provided as the third component indicates whether the
463 -- corresponding extension keyword is valid under the extension options
464 -- provided to the compiler; if the extension corresponding to *any* of the
465 -- bits set in the bitmap is enabled, the keyword is valid (this setup
466 -- facilitates using a keyword in two different extensions that can be
467 -- activated independently)
469 reservedWordsFM = listToUFM $
470 map (\(x, y, z) -> (mkFastString x, (y, z)))
471 [( "_", ITunderscore, 0 ),
473 ( "case", ITcase, 0 ),
474 ( "class", ITclass, 0 ),
475 ( "data", ITdata, 0 ),
476 ( "default", ITdefault, 0 ),
477 ( "deriving", ITderiving, 0 ),
479 ( "else", ITelse, 0 ),
480 ( "hiding", IThiding, 0 ),
482 ( "import", ITimport, 0 ),
484 ( "infix", ITinfix, 0 ),
485 ( "infixl", ITinfixl, 0 ),
486 ( "infixr", ITinfixr, 0 ),
487 ( "instance", ITinstance, 0 ),
489 ( "module", ITmodule, 0 ),
490 ( "newtype", ITnewtype, 0 ),
492 ( "qualified", ITqualified, 0 ),
493 ( "then", ITthen, 0 ),
494 ( "type", ITtype, 0 ),
495 ( "where", ITwhere, 0 ),
496 ( "_scc_", ITscc, 0 ), -- ToDo: remove
498 ( "forall", ITforall, bit glaExtsBit),
499 ( "mdo", ITmdo, bit glaExtsBit),
500 ( "reifyDecl", ITreifyDecl, bit thBit),
501 ( "reifyType", ITreifyType, bit thBit),
502 ( "reifyFixity",ITreifyFixity, bit thBit),
504 ( "foreign", ITforeign, bit ffiBit),
505 ( "export", ITexport, bit ffiBit),
506 ( "label", ITlabel, bit ffiBit),
507 ( "dynamic", ITdynamic, bit ffiBit),
508 ( "safe", ITsafe, bit ffiBit),
509 ( "threadsafe", ITthreadsafe, bit ffiBit),
510 ( "unsafe", ITunsafe, bit ffiBit),
511 ( "stdcall", ITstdcallconv, bit ffiBit),
512 ( "ccall", ITccallconv, bit ffiBit),
513 ( "dotnet", ITdotnet, bit ffiBit),
515 ( "rec", ITrec, bit arrowsBit),
516 ( "proc", ITproc, bit arrowsBit)
519 reservedSymsFM = listToUFM $
520 map (\ (x,y,z) -> (mkFastString x,(y,z)))
521 [ ("..", ITdotdot, 0)
522 ,(":", ITcolon, 0) -- (:) is a reserved op,
523 -- meaning only list cons
536 ,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT
537 ,(".", ITdot, bit glaExtsBit) -- For 'forall a . t'
539 ,("-<", ITlarrowtail, bit arrowsBit)
540 ,(">-", ITrarrowtail, bit arrowsBit)
541 ,("-<<", ITLarrowtail, bit arrowsBit)
542 ,(">>-", ITRarrowtail, bit arrowsBit)
545 -- -----------------------------------------------------------------------------
548 type Action = SrcLoc -> SrcLoc -> StringBuffer -> Int -> P Token
550 special :: Token__ -> Action
551 special tok loc end _buf len = return (T loc end tok)
553 token, layout_token :: Token__ -> Action
554 token t loc end buf len = return (T loc end t)
555 layout_token t loc end buf len = pushLexState layout >> return (T loc end t)
557 idtoken :: (StringBuffer -> Int -> Token__) -> Action
558 idtoken f loc end buf len = return (T loc end $! (f buf len))
560 skip_one_varid :: (FastString -> Token__) -> Action
561 skip_one_varid f loc end buf len
562 = return (T loc end $! f (lexemeToFastString (stepOn buf) (len-1)))
564 strtoken :: (String -> Token__) -> Action
565 strtoken f loc end buf len =
566 return (T loc end $! (f $! lexemeToString buf len))
568 init_strtoken :: Int -> (String -> Token__) -> Action
569 -- like strtoken, but drops the last N character(s)
570 init_strtoken drop f loc end buf len =
571 return (T loc end $! (f $! lexemeToString buf (len-drop)))
573 begin :: Int -> Action
574 begin code _loc _end _str _len = do pushLexState code; lexToken
577 pop _loc _end _buf _len = do popLexState; lexToken
579 pop_and :: Action -> Action
580 pop_and act loc end buf len = do popLexState; act loc end buf len
582 notFollowedBy char _ _ _ (_,buf) = atEnd buf || currentChar buf /= char
584 ifExtension pred bits _ _ _ = pred bits
587 nested comments require traversing by hand, they can't be parsed
588 using regular expressions.
590 nested_comment :: Action
591 nested_comment loc _end _str _len = do
594 where go 0 input = do setInput input; lexToken
596 case alexGetChar input of
601 case alexGetChar input of
603 Just ('\125',input) -> go (n-1) input
604 Just (c,_) -> go n input
606 case alexGetChar input of
608 Just ('-',input') -> go (n+1) input'
609 Just (c,input) -> go n input
612 err input = do failLocMsgP loc (fst input) "unterminated `{-'"
614 open_brace, close_brace :: Action
615 open_brace loc end _str _len = do
617 setContext (NoLayout:ctx)
618 return (T loc end ITocurly)
619 close_brace loc end _str _len = do
621 return (T loc end ITccurly)
623 -- We have to be careful not to count M.<varid> as a qualified name
624 -- when <varid> is a keyword. We hack around this by catching
625 -- the offending tokens afterward, and re-lexing in a different state.
626 check_qvarid loc end buf len = do
627 case lookupUFM reservedWordsFM var of
629 | not (isSpecial keyword) ->
633 b <- extension (\i -> exts .&. i /= 0)
636 _other -> return token
638 (mod,var) = splitQualName buf len
639 token = T loc end (ITqvarid (mod,var))
643 pushLexState bad_qvarid
646 qvarid buf len = ITqvarid $! splitQualName buf len
647 qconid buf len = ITqconid $! splitQualName buf len
649 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
650 -- takes a StringBuffer and a length, and returns the module name
651 -- and identifier parts of a qualified name. Splits at the *last* dot,
652 -- because of hierarchical module names.
653 splitQualName orig_buf len = split orig_buf 0 0
656 | n == len = done dot_off
657 | lookAhead buf n == '.' = split2 buf n (n+1)
658 | otherwise = split buf dot_off (n+1)
660 -- careful, we might get names like M....
661 -- so, if the character after the dot is not upper-case, this is
662 -- the end of the qualifier part.
664 | isUpper (lookAhead buf n) = split buf dot_off (n+1)
665 | otherwise = done dot_off
668 (lexemeToFastString orig_buf dot_off,
669 lexemeToFastString (stepOnBy (dot_off+1) orig_buf) (len - dot_off -1))
671 varid loc end buf len =
672 case lookupUFM reservedWordsFM fs of
673 Just (keyword,0) -> do
675 return (T loc end keyword)
676 Just (keyword,exts) -> do
677 b <- extension (\i -> exts .&. i /= 0)
678 if b then do maybe_layout keyword
679 return (T loc end keyword)
680 else return (T loc end (ITvarid fs))
681 _other -> return (T loc end (ITvarid fs))
683 fs = lexemeToFastString buf len
685 conid buf len = ITconid fs
686 where fs = lexemeToFastString buf len
688 qvarsym buf len = ITqvarsym $! splitQualName buf len
689 qconsym buf len = ITqconsym $! splitQualName buf len
691 varsym = sym ITvarsym
692 consym = sym ITconsym
694 sym con loc end buf len =
695 case lookupUFM reservedSymsFM fs of
696 Just (keyword,0) -> return (T loc end keyword)
697 Just (keyword,exts) -> do
698 b <- extension (\i -> exts .&. i /= 0)
699 if b then return (T loc end keyword)
700 else return (T loc end $! con fs)
701 _other -> return (T loc end $! con fs)
703 fs = lexemeToFastString buf len
705 tok_decimal loc end buf len
706 = return (T loc end (ITinteger $! parseInteger buf len 10 oct_or_dec))
708 tok_octal loc end buf len
709 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 8 oct_or_dec))
711 tok_hexadecimal loc end buf len
712 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 16 hex))
714 prim_decimal loc end buf len
715 = return (T loc end (ITprimint $! parseInteger buf (len-1) 10 oct_or_dec))
717 prim_octal loc end buf len
718 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 8 oct_or_dec))
720 prim_hexadecimal loc end buf len
721 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 16 hex))
723 tok_float str = ITrational $! readRational__ str
724 prim_float str = ITprimfloat $! readRational__ str
725 prim_double str = ITprimdouble $! readRational__ str
727 parseInteger :: StringBuffer -> Int -> Integer -> (Char->Int) -> Integer
728 parseInteger buf len radix to_int
730 where go i x | i == len = x
731 | otherwise = go (i+1) (x * radix + toInteger (to_int (lookAhead buf i)))
733 -- -----------------------------------------------------------------------------
736 -- we're at the first token on a line, insert layout tokens if necessary
738 do_bol loc end _str _len = do
739 pos <- getOffside end
742 --trace "layout: inserting '}'" $ do
744 -- do NOT pop the lex state, we might have a ';' to insert
745 return (T loc end ITvccurly)
747 --trace "layout: inserting ';'" $ do
749 return (T loc end ITsemi)
754 -- certain keywords put us in the "layout" state, where we might
755 -- add an opening curly brace.
756 maybe_layout ITdo = pushLexState layout_do
757 maybe_layout ITmdo = pushLexState layout_do
758 maybe_layout ITof = pushLexState layout
759 maybe_layout ITlet = pushLexState layout
760 maybe_layout ITwhere = pushLexState layout
761 maybe_layout ITrec = pushLexState layout
762 maybe_layout _ = return ()
764 -- Pushing a new implicit layout context. If the indentation of the
765 -- next token is not greater than the previous layout context, then
766 -- Haskell 98 says that the new layout context should be empty; that is
767 -- the lexer must generate {}.
769 -- We are slightly more lenient than this: when the new context is started
770 -- by a 'do', then we allow the new context to be at the same indentation as
771 -- the previous context. This is what the 'strict' argument is for.
773 new_layout_context strict loc end _buf _len = do
775 let offset = srcLocCol loc
778 Layout prev_off : _ |
779 (strict && prev_off >= offset ||
780 not strict && prev_off > offset) -> do
781 -- token is indented to the left of the previous context.
782 -- we must generate a {} sequence now.
783 pushLexState layout_left
784 return (T loc end ITvocurly)
786 setContext (Layout offset : ctx)
787 return (T loc end ITvocurly)
789 do_layout_left loc end _buf _len = do
791 pushLexState bol -- we must be at the start of a line
792 return (T loc end ITvccurly)
794 -- -----------------------------------------------------------------------------
797 set_line :: Int -> Action
798 set_line code loc end buf len = do
799 let line = parseInteger buf len 10 oct_or_dec
800 setSrcLoc (mkSrcLoc (srcLocFile end) (fromIntegral line - 1) 0)
801 -- subtract one: the line number refers to the *following* line
806 set_file :: Int -> Action
807 set_file code loc end buf len = do
808 let file = lexemeToFastString (stepOn buf) (len-2)
809 setSrcLoc (mkSrcLoc file (srcLocLine end) (srcLocCol end))
814 -- -----------------------------------------------------------------------------
817 -- This stuff is horrible. I hates it.
819 lex_string_tok :: Action
820 lex_string_tok loc end buf len = do
823 return (T loc end tok)
825 lex_string :: String -> P Token__
828 case alexGetChar i of
833 glaexts <- extension glaExtsEnabled
837 case alexGetChar i of
841 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
842 else let s' = mkFastStringNarrow (reverse s) in
843 -- always a narrow string/byte array
844 return (ITprimstring s')
846 return (ITstring (mkFastString (reverse s)))
848 return (ITstring (mkFastString (reverse s)))
851 | Just ('&',i) <- next -> do
852 setInput i; lex_string s
853 | Just (c,i) <- next, is_space c -> do
854 setInput i; lex_stringgap s
855 where next = alexGetChar i
866 c | is_space c -> lex_stringgap s
870 lex_char_tok :: Action
871 lex_char_tok loc _end buf len = do
876 glaexts <- extension glaExtsEnabled
879 i@(end,_) <- getInput
880 case alexGetChar i of
881 Just ('#',i@(end,_)) -> do
883 return (T loc end (ITprimchar c))
885 return (T loc end (ITchar c))
888 return (T loc end (ITchar c))
897 c | is_any c -> return c
914 '^' -> do c <- getCharOrFail
915 if c >= '@' && c <= '_'
916 then return (chr (ord c - ord '@'))
919 'x' -> readNum is_hexdigit 16 hex
920 'o' -> readNum is_octdigit 8 oct_or_dec
921 x | is_digit x -> readNum2 is_digit 10 oct_or_dec (oct_or_dec x)
925 case alexGetChar i of
928 case alexGetChar i2 of
931 let str = [c1,c2,c3] in
932 case [ (c,rest) | (p,c) <- silly_escape_chars,
933 Just rest <- [maybePrefixMatch p str] ] of
934 (escape_char,[]):_ -> do
937 (escape_char,_:_):_ -> do
942 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
943 readNum is_digit base conv = do
946 then readNum2 is_digit base conv (conv c)
949 readNum2 is_digit base conv i = do
952 where read i input = do
953 case alexGetChar input of
954 Just (c,input') | is_digit c -> do
955 read (i*base + conv c) input'
958 if i >= 0 && i <= 0x10FFFF
964 || (c >= 'a' && c <= 'f')
965 || (c >= 'A' && c <= 'F')
967 hex c | is_digit c = ord c - ord '0'
968 | otherwise = ord (to_lower c) - ord 'a' + 10
970 oct_or_dec c = ord c - ord '0'
972 is_octdigit c = c >= '0' && c <= '7'
975 | c >= 'A' && c <= 'Z' = chr (ord c - (ord 'A' - ord 'a'))
978 silly_escape_chars = [
1015 lit_error = lexError "lexical error in string/character literal"
1017 getCharOrFail :: P Char
1020 case alexGetChar i of
1021 Nothing -> lexError "unexpected end-of-file in string/character literal"
1022 Just (c,i) -> do setInput i; return c
1024 -- -----------------------------------------------------------------------------
1027 readRational :: ReadS Rational -- NB: doesn't handle leading "-"
1029 (n,d,s) <- readFix r
1031 return ((n%1)*10^^(k-d), t)
1034 (ds,s) <- lexDecDigits r
1035 (ds',t) <- lexDotDigits s
1036 return (read (ds++ds'), length ds', t)
1038 readExp (e:s) | e `elem` "eE" = readExp' s
1039 readExp s = return (0,s)
1041 readExp' ('+':s) = readDec s
1042 readExp' ('-':s) = do
1045 readExp' s = readDec s
1048 (ds,r) <- nonnull isDigit s
1049 return (foldl1 (\n d -> n * 10 + d) [ ord d - ord '0' | d <- ds ],
1052 lexDecDigits = nonnull isDigit
1054 lexDotDigits ('.':s) = return (span isDigit s)
1055 lexDotDigits s = return ("",s)
1057 nonnull p s = do (cs@(_:_),t) <- return (span p s)
1060 readRational__ :: String -> Rational -- NB: *does* handle a leading "-"
1061 readRational__ top_s
1063 '-' : xs -> - (read_me xs)
1067 = case (do { (x,"") <- readRational s ; return x }) of
1069 [] -> error ("readRational__: no parse:" ++ top_s)
1070 _ -> error ("readRational__: ambiguous parse:" ++ top_s)
1072 -- -----------------------------------------------------------------------------
1082 SrcLoc SrcLoc -- The start and end of the text span related to
1083 -- the error. Might be used in environments which can
1084 -- show this span, e.g. by highlighting it.
1085 Message -- The error message
1087 showPFailed loc1 loc2 err = hcat [ppr loc1, text ": ", err]
1089 data PState = PState {
1090 buffer :: StringBuffer,
1091 last_loc :: SrcLoc, -- pos of previous token
1092 last_len :: !Int, -- len of previous token
1093 loc :: SrcLoc, -- current loc (end of prev token + 1)
1094 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1095 context :: [LayoutContext],
1098 -- last_loc and last_len are used when generating error messages,
1099 -- and in pushCurrentContext only.
1101 newtype P a = P { unP :: PState -> ParseResult a }
1103 instance Monad P where
1109 returnP a = P $ \s -> POk s a
1111 thenP :: P a -> (a -> P b) -> P b
1112 (P m) `thenP` k = P $ \ s ->
1114 POk s1 a -> (unP (k a)) s1
1115 PFailed l1 l2 err -> PFailed l1 l2 err
1117 failP :: String -> P a
1118 failP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1120 failMsgP :: String -> P a
1121 failMsgP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1123 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1124 failLocMsgP loc1 loc2 str = P $ \s -> PFailed loc1 loc2 (text str)
1126 extension :: (Int -> Bool) -> P Bool
1127 extension p = P $ \s -> POk s (p $! extsBitmap s)
1130 getExts = P $ \s -> POk s (extsBitmap s)
1132 setSrcLoc :: SrcLoc -> P ()
1133 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1135 -- tmp, for supporting stuff in RdrHsSyn. The scope better not include
1136 -- any calls to the lexer, because it assumes things about the SrcLoc.
1137 setSrcLocFor :: SrcLoc -> P a -> P a
1138 setSrcLocFor new_loc scope = P $ \s@PState{ loc = old_loc } ->
1139 case unP scope s{loc=new_loc} of
1140 PFailed l1 l2 msg -> PFailed l1 l2 msg
1143 getSrcLoc :: P SrcLoc
1144 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1146 setLastToken :: SrcLoc -> Int -> P ()
1147 setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } ()
1149 type AlexInput = (SrcLoc,StringBuffer)
1151 alexInputPrevChar :: AlexInput -> Char
1152 alexInputPrevChar (_,s) = prevChar s '\n'
1154 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1157 | otherwise = c `seq` loc' `seq` s' `seq` Just (c, (loc', s'))
1158 where c = currentChar s
1159 loc' = advanceSrcLoc loc c
1162 getInput :: P AlexInput
1163 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (l,b)
1165 setInput :: AlexInput -> P ()
1166 setInput (l,b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1168 pushLexState :: Int -> P ()
1169 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1171 popLexState :: P Int
1172 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1174 getLexState :: P Int
1175 getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls
1177 -- for reasons of efficiency, flags indicating language extensions (eg,
1178 -- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
1181 glaExtsBit, ffiBit, parrBit :: Int
1189 glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1190 glaExtsEnabled flags = testBit flags glaExtsBit
1191 ffiEnabled flags = testBit flags ffiBit
1192 parrEnabled flags = testBit flags parrBit
1193 arrowsEnabled flags = testBit flags arrowsBit
1194 thEnabled flags = testBit flags thBit
1195 ipEnabled flags = testBit flags ipBit
1197 -- create a parse state
1199 mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
1200 mkPState buf loc flags =
1206 extsBitmap = fromIntegral bitmap,
1208 lex_state = [bol, if glaExtsEnabled bitmap then glaexts else 0]
1209 -- we begin in the layout state if toplev_layout is set
1212 bitmap = glaExtsBit `setBitIf` dopt Opt_GlasgowExts flags
1213 .|. ffiBit `setBitIf` dopt Opt_FFI flags
1214 .|. parrBit `setBitIf` dopt Opt_PArr flags
1215 .|. arrowsBit `setBitIf` dopt Opt_Arrows flags
1216 .|. thBit `setBitIf` dopt Opt_TH flags
1217 .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
1219 setBitIf :: Int -> Bool -> Int
1220 b `setBitIf` cond | cond = bit b
1223 getContext :: P [LayoutContext]
1224 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1226 setContext :: [LayoutContext] -> P ()
1227 setContext ctx = P $ \s -> POk s{context=ctx} ()
1230 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1231 loc = loc, last_len = len, last_loc = last_loc }) ->
1233 (_:tl) -> POk s{ context = tl } ()
1234 [] -> PFailed last_loc loc (srcParseErr buf len)
1236 -- Push a new layout context at the indentation of the last token read.
1237 -- This is only used at the outer level of a module when the 'module'
1238 -- keyword is missing.
1239 pushCurrentContext :: P ()
1240 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1241 POk s{ context = Layout (srcLocCol loc) : ctx} ()
1243 getOffside :: SrcLoc -> P Ordering
1244 getOffside loc = P $ \s@PState{context=stk} ->
1245 let ord = case stk of
1246 (Layout n:_) -> compare (srcLocCol loc) n
1250 -- ---------------------------------------------------------------------------
1251 -- Construct a parse error
1254 :: StringBuffer -- current buffer (placed just after the last token)
1255 -> Int -- length of the previous token
1258 = hcat [ if null token
1259 then ptext SLIT("parse error (possibly incorrect indentation)")
1260 else hcat [ptext SLIT("parse error on input "),
1261 char '`', text token, char '\'']
1263 where token = lexemeToString (stepOnBy (-len) buf) len
1265 -- Report a parse failure, giving the span of the previous token as
1266 -- the location of the error. This is the entry point for errors
1267 -- detected during parsing.
1269 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1270 last_loc = last_loc, loc = loc } ->
1271 PFailed last_loc loc (srcParseErr buf len)
1273 -- A lexical error is reported at a particular position in the source file,
1274 -- not over a token range. TODO: this is slightly wrong, because we record
1275 -- the error at the character position following the one which caused the
1276 -- error. We should somehow back up by one character.
1277 lexError :: String -> P a
1280 failLocMsgP loc loc str
1282 -- -----------------------------------------------------------------------------
1283 -- This is the top-level function: called from the parser each time a
1284 -- new token is to be read from the input.
1286 lexer :: (Token -> P a) -> P a
1288 tok@(T _ _ tok__) <- lexToken
1289 --trace ("token: " ++ show tok__) $ do
1294 inp@(loc1,buf) <- getInput
1297 case alexScanUser exts inp sc of
1298 AlexEOF -> do setLastToken loc1 0
1299 return (T loc1 loc1 ITeof)
1300 AlexError (loc2,_) -> do failLocMsgP loc1 loc2 "lexical error"
1301 AlexSkip inp2 _ -> do
1304 AlexToken inp2@(end,buf2) len t -> do
1306 setLastToken loc1 len