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, ExtFlags(..), 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 )
42 import Util ( maybePrefixMatch )
50 $whitechar = [\ \t\n\r\f\v]
51 $white_no_nl = $whitechar # \n
55 $digit = [$ascdigit $unidigit]
57 $special = [\(\)\,\;\[\]\`\{\}]
58 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
60 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
63 $asclarge = [A-Z \xc0-\xd6 \xd8-\xde]
64 $large = [$asclarge $unilarge]
67 $ascsmall = [a-z \xdf-\xf6 \xf8-\xff]
68 $small = [$ascsmall $unismall \_]
70 $graphic = [$small $large $symbol $digit $special \:\"\']
73 $hexit = [$digit A-F a-f]
74 $symchar = [$symbol \:]
76 $idchar = [$small $large $digit \']
78 @varid = $small $idchar*
79 @conid = $large $idchar*
81 @varsym = $symbol $symchar*
82 @consym = \: $symchar*
86 @hexadecimal = $hexit+
87 @exponent = [eE] [\-\+]? @decimal
89 -- we support the hierarchical module name extension:
92 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
96 -- everywhere: skip whitespace and comments
99 -- Everywhere: deal with nested comments. We explicitly rule out
100 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
101 -- (this can happen even though pragmas will normally take precedence due to
102 -- longest-match, because pragmas aren't valid in every state, but comments
104 "{-" / { notFollowedBy '#' } { nested_comment }
106 -- Single-line comments are a bit tricky. Haskell 98 says that two or
107 -- more dashes followed by a symbol should be parsed as a varsym, so we
108 -- have to exclude those.
109 -- The regex says: "munch all the characters after the dashes, as long as
110 -- the first one is not a symbol".
111 "--"\-* ([^$symbol] .*)? ;
113 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
114 -- blank lines) until we find a non-whitespace character, then do layout
117 -- One slight wibble here: what if the line begins with {-#? In
118 -- theory, we have to lex the pragma to see if it's one we recognise,
119 -- and if it is, then we backtrack and do_bol, otherwise we treat it
120 -- as a nested comment. We don't bother with this: if the line begins
121 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
124 ^\# (line)? { begin line_prag1 }
128 -- after a layout keyword (let, where, do, of), we begin a new layout
129 -- context if the curly brace is missing.
130 -- Careful! This stuff is quite delicate.
131 <layout, layout_do> {
132 \{ / { notFollowedBy '-' } { pop_and open_brace }
133 -- we might encounter {-# here, but {- has been handled already
135 ^\# (line)? { begin line_prag1 }
138 -- do is treated in a subtly different way, see new_layout_context
139 <layout> () { new_layout_context True }
140 <layout_do> () { new_layout_context False }
142 -- after a new layout context which was found to be to the left of the
143 -- previous context, we have generated a '{' token, and we now need to
144 -- generate a matching '}' token.
145 <layout_left> () { do_layout_left }
147 <0,glaexts> \n { begin bol }
149 "{-#" $whitechar* (line|LINE) { begin line_prag2 }
151 -- single-line line pragmas, of the form
152 -- # <line> "<file>" <extra-stuff> \n
153 <line_prag1> $digit+ { set_line line_prag1a }
154 <line_prag1a> \" [$graphic \ ]* \" { set_file line_prag1b }
155 <line_prag1b> .* { pop }
157 -- Haskell-style line pragmas, of the form
158 -- {-# LINE <line> "<file>" #-}
159 <line_prag2> $digit+ { set_line line_prag2a }
160 <line_prag2a> \" $graphic* \" { set_file line_prag2b }
161 <line_prag2b> "#-}" { pop }
164 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
165 { token ITspecialise_prag }
166 "{-#" $whitechar* (SOURCE|source) { token ITsource_prag }
167 "{-#" $whitechar* (INLINE|inline) { token ITinline_prag }
168 "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
169 { token ITnoinline_prag }
170 "{-#" $whitechar* (RULES|rules) { token ITrules_prag }
171 "{-#" $whitechar* (DEPRECATED|deprecated)
172 { token ITdeprecated_prag }
173 "{-#" $whitechar* (SCC|scc) { token ITscc_prag }
174 "{-#" $whitechar* (CORE|core) { token ITcore_prag }
176 "{-#" { nested_comment }
178 -- ToDo: should only be valid inside a pragma:
179 "#-}" { token ITclose_prag}
183 -- '0' state: ordinary lexemes
184 -- 'glaexts' state: glasgow extensions (postfix '#', etc.)
189 "(#" { token IToubxparen }
190 "#)" { token ITcubxparen }
192 "[:" { token ITopabrack }
193 ":]" { token ITcpabrack }
195 "{|" { token ITocurlybar }
196 "|}" { token ITccurlybar }
198 "[|" { token ITopenExpQuote }
199 "[e|" { token ITopenExpQuote }
200 "[p|" { token ITopenPatQuote }
201 "[d|" { token ITopenDecQuote }
202 "[t|" { token ITopenTypQuote }
203 "|]" { token ITcloseQuote }
207 "(|" / { \b _ _ _ -> arrowsEnabled b} { special IToparenbar }
208 "|)" / { \b _ _ _ -> arrowsEnabled b} { special ITcparenbar }
209 \( { special IToparen }
210 \) { special ITcparen }
211 \[ { special ITobrack }
212 \] { special ITcbrack }
213 \, { special ITcomma }
214 \; { special ITsemi }
215 \` { special ITbackquote }
222 \? @varid { skip_one_varid ITdupipvarid }
223 \% @varid { skip_one_varid ITsplitipvarid }
224 \$ @varid { skip_one_varid ITidEscape }
225 "$(" { token ITparenEscape }
229 @qual @varid { check_qvarid }
230 @qual @conid { idtoken qconid }
232 @conid { idtoken conid }
235 -- after an illegal qvarid, such as 'M.let',
236 -- we back up and try again in the bad_qvarid state:
238 @conid { pop_and (idtoken conid) }
239 @qual @conid { pop_and (idtoken qconid) }
243 @qual @varid "#"+ { idtoken qvarid }
244 @qual @conid "#"+ { idtoken qconid }
245 @varid "#"+ { varid }
246 @conid "#"+ { idtoken conid }
252 @qual @varsym { idtoken qvarsym }
253 @qual @consym { idtoken qconsym }
259 @decimal { tok_decimal }
260 0[oO] @octal { tok_octal }
261 0[xX] @hexadecimal { tok_hexadecimal }
265 @decimal \# { prim_decimal }
266 0[oO] @octal \# { prim_octal }
267 0[xX] @hexadecimal \# { prim_hexadecimal }
270 <0,glaexts> @floating_point { strtoken tok_float }
271 <glaexts> @floating_point \# { init_strtoken 1 prim_float }
272 <glaexts> @floating_point \# \# { init_strtoken 2 prim_double }
274 -- Strings and chars are lexed by hand-written code. The reason is
275 -- that even if we recognise the string or char here in the regex
276 -- lexer, we would still have to parse the string afterward in order
277 -- to convert it to a String.
280 \" { lex_string_tok }
283 <glaexts> "``" (([$graphic $whitechar] # \') | \' ([$graphic $whitechar] # \'))*
287 -- work around bug in Alex 2.0
288 #if __GLASGOW_HASKELL__ < 503
289 unsafeAt arr i = arr ! i
292 -- -----------------------------------------------------------------------------
295 data Token = T SrcLoc{-start-} SrcLoc{-end-} Token__
298 = ITas -- Haskell keywords
322 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
324 | ITforall -- GHC extension keywords
336 | ITccall (Bool,Bool,Safety) -- (is_dyn, is_casm, may_gc)
339 | ITspecialise_prag -- Pragmas
347 | ITcore_prag -- hdaume: core annotations
350 | ITdotdot -- reserved symbols
366 | ITbiglam -- GHC-extension symbols
368 | ITocurly -- special symbols
370 | ITocurlybar -- {|, for type applications
371 | ITccurlybar -- |}, for type applications
375 | ITopabrack -- [:, for parallel arrays with -fparr
376 | ITcpabrack -- :], for parallel arrays with -fparr
387 | ITvarid FastString -- identifiers
389 | ITvarsym FastString
390 | ITconsym FastString
391 | ITqvarid (FastString,FastString)
392 | ITqconid (FastString,FastString)
393 | ITqvarsym (FastString,FastString)
394 | ITqconsym (FastString,FastString)
396 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
397 | ITsplitipvarid FastString -- GHC extension: implicit param: %x
399 | ITpragma StringBuffer
402 | ITstring FastString
404 | ITrational Rational
407 | ITprimstring FastString
409 | ITprimfloat Rational
410 | ITprimdouble Rational
411 | ITlitlit FastString
413 -- MetaHaskell extension tokens
414 | ITopenExpQuote -- [| or [e|
415 | ITopenPatQuote -- [p|
416 | ITopenDecQuote -- [d|
417 | ITopenTypQuote -- [t|
419 | ITidEscape FastString -- $x
420 | ITparenEscape -- $(
425 -- Arrow notation extension
432 | ITLarrowtail -- -<<
433 | ITRarrowtail -- >>-
435 | ITunknown String -- Used when the lexer can't make sense of it
436 | ITeof -- end of file token
438 deriving Show -- debugging
441 isSpecial :: Token__ -> Bool
442 -- If we see M.x, where x is a keyword, but
443 -- is special, we treat is as just plain M.x,
445 isSpecial ITas = True
446 isSpecial IThiding = True
447 isSpecial ITqualified = True
448 isSpecial ITforall = True
449 isSpecial ITexport = True
450 isSpecial ITlabel = True
451 isSpecial ITdynamic = True
452 isSpecial ITsafe = True
453 isSpecial ITthreadsafe = True
454 isSpecial ITunsafe = True
455 isSpecial ITwith = True
456 isSpecial ITccallconv = True
457 isSpecial ITstdcallconv = True
458 isSpecial ITmdo = True
461 -- the bitmap provided as the third component indicates whether the
462 -- corresponding extension keyword is valid under the extension options
463 -- provided to the compiler; if the extension corresponding to *any* of the
464 -- bits set in the bitmap is enabled, the keyword is valid (this setup
465 -- facilitates using a keyword in two different extensions that can be
466 -- activated independently)
468 reservedWordsFM = listToUFM $
469 map (\(x, y, z) -> (mkFastString x, (y, z)))
470 [( "_", ITunderscore, 0 ),
472 ( "case", ITcase, 0 ),
473 ( "class", ITclass, 0 ),
474 ( "data", ITdata, 0 ),
475 ( "default", ITdefault, 0 ),
476 ( "deriving", ITderiving, 0 ),
478 ( "else", ITelse, 0 ),
479 ( "hiding", IThiding, 0 ),
481 ( "import", ITimport, 0 ),
483 ( "infix", ITinfix, 0 ),
484 ( "infixl", ITinfixl, 0 ),
485 ( "infixr", ITinfixr, 0 ),
486 ( "instance", ITinstance, 0 ),
488 ( "module", ITmodule, 0 ),
489 ( "newtype", ITnewtype, 0 ),
491 ( "qualified", ITqualified, 0 ),
492 ( "then", ITthen, 0 ),
493 ( "type", ITtype, 0 ),
494 ( "where", ITwhere, 0 ),
495 ( "_scc_", ITscc, 0 ), -- ToDo: remove
497 ( "forall", ITforall, bit glaExtsBit),
498 ( "mdo", ITmdo, bit glaExtsBit),
499 ( "reifyDecl", ITreifyDecl, bit glaExtsBit),
500 ( "reifyType", ITreifyType, bit glaExtsBit),
501 ( "reifyFixity",ITreifyFixity, 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 ( "with", ITwith, bit withBit),
516 ( "rec", ITrec, bit arrowsBit),
517 ( "proc", ITproc, bit arrowsBit),
520 ("_ccall_", ITccall (False, False, PlayRisky),
522 ("_ccall_GC_", ITccall (False, False, PlaySafe False),
524 ("_casm_", ITccall (False, True, PlayRisky),
526 ("_casm_GC_", ITccall (False, True, PlaySafe False),
530 reservedSymsFM = listToUFM $
531 map (\ (x,y,z) -> (mkFastString x,(y,z)))
532 [ ("..", ITdotdot, 0)
533 ,(":", ITcolon, 0) -- (:) is a reserved op,
534 -- meaning only list cons
547 ,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT
548 ,(".", ITdot, bit glaExtsBit) -- For 'forall a . t'
550 ,("-<", ITlarrowtail, bit arrowsBit)
551 ,(">-", ITrarrowtail, bit arrowsBit)
552 ,("-<<", ITLarrowtail, bit arrowsBit)
553 ,(">>-", ITRarrowtail, bit arrowsBit)
556 -- -----------------------------------------------------------------------------
559 type Action = SrcLoc -> SrcLoc -> StringBuffer -> Int -> P Token
561 special :: Token__ -> Action
562 special tok loc end _buf len = return (T loc end tok)
564 token :: Token__ -> Action
565 token t loc end buf len = return (T loc end t)
567 idtoken :: (StringBuffer -> Int -> Token__) -> Action
568 idtoken f loc end buf len = return (T loc end $! (f buf len))
570 skip_one_varid :: (FastString -> Token__) -> Action
571 skip_one_varid f loc end buf len
572 = return (T loc end $! f (lexemeToFastString (stepOn buf) (len-1)))
574 strtoken :: (String -> Token__) -> Action
575 strtoken f loc end buf len =
576 return (T loc end $! (f $! lexemeToString buf len))
578 init_strtoken :: Int -> (String -> Token__) -> Action
579 -- like strtoken, but drops the last N character(s)
580 init_strtoken drop f loc end buf len =
581 return (T loc end $! (f $! lexemeToString buf (len-drop)))
583 begin :: Int -> Action
584 begin code _loc _end _str _len = do pushLexState code; lexToken
587 pop _loc _end _buf _len = do popLexState; lexToken
589 pop_and :: Action -> Action
590 pop_and act loc end buf len = do popLexState; act loc end buf len
592 notFollowedBy char _ _ _ (_,buf) = atEnd buf || currentChar buf /= char
595 nested comments require traversing by hand, they can't be parsed
596 using regular expressions.
598 nested_comment :: Action
599 nested_comment loc _end _str _len = do
602 where go 0 input = do setInput input; lexToken
604 case alexGetChar input of
609 case alexGetChar input of
611 Just ('\125',input) -> go (n-1) input
612 Just (c,_) -> go n input
614 case alexGetChar input of
616 Just ('-',input') -> go (n+1) input'
617 Just (c,input) -> go n input
620 err input = do failLocMsgP loc (fst input) "unterminated `{-'"
622 open_brace, close_brace :: Action
623 open_brace loc end _str _len = do
625 setContext (NoLayout:ctx)
626 return (T loc end ITocurly)
627 close_brace loc end _str _len = do
629 return (T loc end ITccurly)
631 -- We have to be careful not to count M.<varid> as a qualified name
632 -- when <varid> is a keyword. We hack around this by catching
633 -- the offending tokens afterward, and re-lexing in a different state.
634 check_qvarid loc end buf len = do
635 case lookupUFM reservedWordsFM var of
637 | not (isSpecial keyword) ->
641 b <- extension (\i -> exts .&. i /= 0)
644 _other -> return token
646 (mod,var) = splitQualName buf len
647 token = T loc end (ITqvarid (mod,var))
651 pushLexState bad_qvarid
654 qvarid buf len = ITqvarid $! splitQualName buf len
655 qconid buf len = ITqconid $! splitQualName buf len
657 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
658 -- takes a StringBuffer and a length, and returns the module name
659 -- and identifier parts of a qualified name. Splits at the *last* dot,
660 -- because of hierarchical module names.
661 splitQualName orig_buf len = split orig_buf 0 0
664 | n == len = done dot_off
665 | lookAhead buf n == '.' = split2 buf n (n+1)
666 | otherwise = split buf dot_off (n+1)
668 -- careful, we might get names like M....
669 -- so, if the character after the dot is not upper-case, this is
670 -- the end of the qualifier part.
672 | isUpper (lookAhead buf n) = split buf dot_off (n+1)
673 | otherwise = done dot_off
676 (lexemeToFastString orig_buf dot_off,
677 lexemeToFastString (stepOnBy (dot_off+1) orig_buf) (len - dot_off -1))
679 varid loc end buf len =
680 case lookupUFM reservedWordsFM fs of
681 Just (keyword,0) -> do
683 return (T loc end keyword)
684 Just (keyword,exts) -> do
685 b <- extension (\i -> exts .&. i /= 0)
686 if b then do maybe_layout keyword
687 return (T loc end keyword)
688 else return (T loc end (ITvarid fs))
689 _other -> return (T loc end (ITvarid fs))
691 fs = lexemeToFastString buf len
693 conid buf len = ITconid fs
694 where fs = lexemeToFastString buf len
696 qvarsym buf len = ITqvarsym $! splitQualName buf len
697 qconsym buf len = ITqconsym $! splitQualName buf len
699 varsym = sym ITvarsym
700 consym = sym ITconsym
702 sym con loc end buf len =
703 case lookupUFM reservedSymsFM fs of
704 Just (keyword,0) -> return (T loc end keyword)
705 Just (keyword,exts) -> do
706 b <- extension (\i -> exts .&. i /= 0)
707 if b then return (T loc end keyword)
708 else return (T loc end $! con fs)
709 _other -> return (T loc end $! con fs)
711 fs = lexemeToFastString buf len
713 tok_decimal loc end buf len
714 = return (T loc end (ITinteger $! parseInteger buf len 10 oct_or_dec))
716 tok_octal loc end buf len
717 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 8 oct_or_dec))
719 tok_hexadecimal loc end buf len
720 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 16 hex))
722 prim_decimal loc end buf len
723 = return (T loc end (ITprimint $! parseInteger buf (len-1) 10 oct_or_dec))
725 prim_octal loc end buf len
726 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 8 oct_or_dec))
728 prim_hexadecimal loc end buf len
729 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 16 hex))
731 tok_float str = ITrational $! readRational__ str
732 prim_float str = ITprimfloat $! readRational__ str
733 prim_double str = ITprimdouble $! readRational__ str
735 parseInteger :: StringBuffer -> Int -> Integer -> (Char->Int) -> Integer
736 parseInteger buf len radix to_int
738 where go i x | i == len = x
739 | otherwise = go (i+1) (x * radix + toInteger (to_int (lookAhead buf i)))
742 clitlit loc end buf len =
743 return (T loc end (ITlitlit $! lexemeToFastString (stepOnBy 2 buf) (len-4)))
745 -- -----------------------------------------------------------------------------
748 -- we're at the first token on a line, insert layout tokens if necessary
750 do_bol loc end _str _len = do
751 pos <- getOffside end
754 --trace "layout: inserting '}'" $ do
756 -- do NOT pop the lex state, we might have a ';' to insert
757 return (T loc end ITvccurly)
759 --trace "layout: inserting ';'" $ do
761 return (T loc end ITsemi)
766 -- certain keywords put us in the "layout" state, where we might
767 -- add an opening curly brace.
768 maybe_layout ITdo = pushLexState layout_do
769 maybe_layout ITmdo = pushLexState layout_do
770 maybe_layout ITof = pushLexState layout
771 maybe_layout ITlet = pushLexState layout
772 maybe_layout ITwhere = pushLexState layout
773 maybe_layout _ = return ()
775 -- Pushing a new implicit layout context. If the indentation of the
776 -- next token is not greater than the previous layout context, then
777 -- Haskell 98 says that the new layout context should be empty; that is
778 -- the lexer must generate {}.
780 -- We are slightly more lenient than this: when the new context is started
781 -- by a 'do', then we allow the new context to be at the same indentation as
782 -- the previous context. This is what the 'strict' argument is for.
784 new_layout_context strict loc end _buf _len = do
786 let offset = srcLocCol loc
789 Layout prev_off : _ |
790 (strict && prev_off >= offset ||
791 not strict && prev_off > offset) -> do
792 -- token is indented to the left of the previous context.
793 -- we must generate a {} sequence now.
794 pushLexState layout_left
795 return (T loc end ITvocurly)
797 setContext (Layout offset : ctx)
798 return (T loc end ITvocurly)
800 do_layout_left loc end _buf _len = do
802 pushLexState bol -- we must be at the start of a line
803 return (T loc end ITvccurly)
805 -- -----------------------------------------------------------------------------
808 set_line :: Int -> Action
809 set_line code loc end buf len = do
810 let line = parseInteger buf len 10 oct_or_dec
811 setSrcLoc (mkSrcLoc (srcLocFile end) (fromIntegral line - 1) 0)
812 -- subtract one: the line number refers to the *following* line
817 set_file :: Int -> Action
818 set_file code loc end buf len = do
819 let file = lexemeToFastString (stepOn buf) (len-2)
820 setSrcLoc (mkSrcLoc file (srcLocLine end) (srcLocCol end))
825 -- -----------------------------------------------------------------------------
828 -- This stuff is horrible. I hates it.
830 lex_string_tok :: Action
831 lex_string_tok loc end buf len = do
834 return (T loc end tok)
836 lex_string :: String -> P Token__
839 case alexGetChar i of
844 glaexts <- extension glaExtsEnabled
848 case alexGetChar i of
852 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
853 else let s' = mkFastStringNarrow (reverse s) in
854 -- always a narrow string/byte array
855 return (ITprimstring s')
857 return (ITstring (mkFastString (reverse s)))
859 return (ITstring (mkFastString (reverse s)))
862 | Just ('&',i) <- next -> do
863 setInput i; lex_string s
864 | Just (c,i) <- next, is_space c -> do
865 setInput i; lex_stringgap s
866 where next = alexGetChar i
877 c | is_space c -> lex_stringgap s
881 lex_char_tok :: Action
882 lex_char_tok loc _end buf len = do
887 glaexts <- extension glaExtsEnabled
890 i@(end,_) <- getInput
891 case alexGetChar i of
892 Just ('#',i@(end,_)) -> do
894 return (T loc end (ITprimchar c))
896 return (T loc end (ITchar c))
899 return (T loc end (ITchar c))
908 c | is_any c -> return c
925 '^' -> do c <- getCharOrFail
926 if c >= '@' && c <= '_'
927 then return (chr (ord c - ord '@'))
930 'x' -> readNum is_hexdigit 16 hex
931 'o' -> readNum is_octdigit 8 oct_or_dec
932 x | is_digit x -> readNum2 is_digit 10 oct_or_dec (oct_or_dec x)
936 case alexGetChar i of
939 case alexGetChar i2 of
942 let str = [c1,c2,c3] in
943 case [ (c,rest) | (p,c) <- silly_escape_chars,
944 Just rest <- [maybePrefixMatch p str] ] of
945 (escape_char,[]):_ -> do
948 (escape_char,_:_):_ -> do
953 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
954 readNum is_digit base conv = do
957 then readNum2 is_digit base conv (conv c)
960 readNum2 is_digit base conv i = do
963 where read i input = do
964 case alexGetChar input of
965 Just (c,input') | is_digit c -> do
966 read (i*base + conv c) input'
969 if i >= 0 && i <= 0x10FFFF
975 || (c >= 'a' && c <= 'f')
976 || (c >= 'A' && c <= 'F')
978 hex c | is_digit c = ord c - ord '0'
979 | otherwise = ord (to_lower c) - ord 'a' + 10
981 oct_or_dec c = ord c - ord '0'
983 is_octdigit c = c >= '0' && c <= '7'
986 | c >= 'A' && c <= 'Z' = chr (ord c - (ord 'A' - ord 'a'))
989 silly_escape_chars = [
1026 lit_error = lexError "lexical error in string/character literal"
1028 getCharOrFail :: P Char
1031 case alexGetChar i of
1032 Nothing -> lexError "unexpected end-of-file in string/character literal"
1033 Just (c,i) -> do setInput i; return c
1035 -- -----------------------------------------------------------------------------
1038 readRational :: ReadS Rational -- NB: doesn't handle leading "-"
1040 (n,d,s) <- readFix r
1042 return ((n%1)*10^^(k-d), t)
1045 (ds,s) <- lexDecDigits r
1046 (ds',t) <- lexDotDigits s
1047 return (read (ds++ds'), length ds', t)
1049 readExp (e:s) | e `elem` "eE" = readExp' s
1050 readExp s = return (0,s)
1052 readExp' ('+':s) = readDec s
1053 readExp' ('-':s) = do
1056 readExp' s = readDec s
1059 (ds,r) <- nonnull isDigit s
1060 return (foldl1 (\n d -> n * 10 + d) [ ord d - ord '0' | d <- ds ],
1063 lexDecDigits = nonnull isDigit
1065 lexDotDigits ('.':s) = return (span isDigit s)
1066 lexDotDigits s = return ("",s)
1068 nonnull p s = do (cs@(_:_),t) <- return (span p s)
1071 readRational__ :: String -> Rational -- NB: *does* handle a leading "-"
1072 readRational__ top_s
1074 '-' : xs -> - (read_me xs)
1078 = case (do { (x,"") <- readRational s ; return x }) of
1080 [] -> error ("readRational__: no parse:" ++ top_s)
1081 _ -> error ("readRational__: ambiguous parse:" ++ top_s)
1083 -- -----------------------------------------------------------------------------
1093 SrcLoc SrcLoc -- The start and end of the text span related to
1094 -- the error. Might be used in environments which can
1095 -- show this span, e.g. by highlighting it.
1096 Message -- The error message
1098 showPFailed loc1 loc2 err
1099 = showSDoc (hcat [ppr loc1, text ": ", err])
1101 data PState = PState {
1102 buffer :: StringBuffer,
1103 last_loc :: SrcLoc, -- pos of previous token
1104 last_len :: !Int, -- len of previous token
1105 loc :: SrcLoc, -- current loc (end of prev token + 1)
1106 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1107 context :: [LayoutContext],
1110 -- last_loc and last_len are used when generating error messages,
1111 -- and in pushCurrentContext only.
1113 newtype P a = P { unP :: PState -> ParseResult a }
1115 instance Monad P where
1121 returnP a = P $ \s -> POk s a
1123 thenP :: P a -> (a -> P b) -> P b
1124 (P m) `thenP` k = P $ \ s ->
1126 POk s1 a -> (unP (k a)) s1
1127 PFailed l1 l2 err -> PFailed l1 l2 err
1129 failP :: String -> P a
1130 failP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1132 failMsgP :: String -> P a
1133 failMsgP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1135 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1136 failLocMsgP loc1 loc2 str = P $ \s -> PFailed loc1 loc2 (text str)
1138 extension :: (Int -> Bool) -> P Bool
1139 extension p = P $ \s -> POk s (p $! extsBitmap s)
1142 getExts = P $ \s -> POk s (extsBitmap s)
1144 setSrcLoc :: SrcLoc -> P ()
1145 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1147 -- tmp, for supporting stuff in RdrHsSyn. The scope better not include
1148 -- any calls to the lexer, because it assumes things about the SrcLoc.
1149 setSrcLocFor :: SrcLoc -> P a -> P a
1150 setSrcLocFor new_loc scope = P $ \s@PState{ loc = old_loc } ->
1151 case unP scope s{loc=new_loc} of
1152 PFailed l1 l2 msg -> PFailed l1 l2 msg
1155 getSrcLoc :: P SrcLoc
1156 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1158 setLastToken :: SrcLoc -> Int -> P ()
1159 setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } ()
1161 type AlexInput = (SrcLoc,StringBuffer)
1163 alexInputPrevChar :: AlexInput -> Char
1164 alexInputPrevChar (_,s) = prevChar s '\n'
1166 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1169 | otherwise = c `seq` loc' `seq` s' `seq` Just (c, (loc', s'))
1170 where c = currentChar s
1171 loc' = advanceSrcLoc loc c
1174 getInput :: P AlexInput
1175 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (l,b)
1177 setInput :: AlexInput -> P ()
1178 setInput (l,b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1180 pushLexState :: Int -> P ()
1181 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1183 popLexState :: P Int
1184 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1186 getLexState :: P Int
1187 getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls
1189 -- for reasons of efficiency, flags indicating language extensions (eg,
1190 -- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
1193 glaExtsBit, ffiBit, parrBit :: Int
1200 glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1201 glaExtsEnabled flags = testBit flags glaExtsBit
1202 ffiEnabled flags = testBit flags ffiBit
1203 withEnabled flags = testBit flags withBit
1204 parrEnabled flags = testBit flags parrBit
1205 arrowsEnabled flags = testBit flags arrowsBit
1207 -- convenient record-based bitmap for the interface to the rest of the world
1209 -- NB: `glasgowExtsEF' implies `ffiEF' (see `mkPState' below)
1211 data ExtFlags = ExtFlags {
1212 glasgowExtsEF :: Bool,
1219 -- create a parse state
1221 mkPState :: StringBuffer -> SrcLoc -> ExtFlags -> PState
1222 mkPState buf loc exts =
1228 extsBitmap = fromIntegral bitmap,
1230 lex_state = [bol, if glaExtsEnabled bitmap then glaexts else 0]
1231 -- we begin in the layout state if toplev_layout is set
1234 bitmap = glaExtsBit `setBitIf` glasgowExtsEF exts
1235 .|. ffiBit `setBitIf` (ffiEF exts
1236 || glasgowExtsEF exts)
1237 .|. withBit `setBitIf` withEF exts
1238 .|. parrBit `setBitIf` parrEF exts
1239 .|. arrowsBit `setBitIf` arrowsEF exts
1241 setBitIf :: Int -> Bool -> Int
1242 b `setBitIf` cond | cond = bit b
1245 getContext :: P [LayoutContext]
1246 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1248 setContext :: [LayoutContext] -> P ()
1249 setContext ctx = P $ \s -> POk s{context=ctx} ()
1252 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1253 loc = loc, last_len = len, last_loc = last_loc }) ->
1255 (_:tl) -> POk s{ context = tl } ()
1256 [] -> PFailed last_loc loc (srcParseErr buf len)
1258 -- Push a new layout context at the indentation of the last token read.
1259 -- This is only used at the outer level of a module when the 'module'
1260 -- keyword is missing.
1261 pushCurrentContext :: P ()
1262 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1263 POk s{ context = Layout (srcLocCol loc) : ctx} ()
1265 getOffside :: SrcLoc -> P Ordering
1266 getOffside loc = P $ \s@PState{context=stk} ->
1267 let ord = case stk of
1268 (Layout n:_) -> compare (srcLocCol loc) n
1272 -- ---------------------------------------------------------------------------
1273 -- Construct a parse error
1276 :: StringBuffer -- current buffer (placed just after the last token)
1277 -> Int -- length of the previous token
1280 = hcat [ if null token
1281 then ptext SLIT("parse error (possibly incorrect indentation)")
1282 else hcat [ptext SLIT("parse error on input "),
1283 char '`', text token, char '\'']
1285 where token = lexemeToString (stepOnBy (-len) buf) len
1287 -- Report a parse failure, giving the span of the previous token as
1288 -- the location of the error. This is the entry point for errors
1289 -- detected during parsing.
1291 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1292 last_loc = last_loc, loc = loc } ->
1293 PFailed last_loc loc (srcParseErr buf len)
1295 -- A lexical error is reported at a particular position in the source file,
1296 -- not over a token range. TODO: this is slightly wrong, because we record
1297 -- the error at the character position following the one which caused the
1298 -- error. We should somehow back up by one character.
1299 lexError :: String -> P a
1302 failLocMsgP loc loc str
1304 -- -----------------------------------------------------------------------------
1305 -- This is the top-level function: called from the parser each time a
1306 -- new token is to be read from the input.
1308 lexer :: (Token -> P a) -> P a
1310 tok@(T _ _ tok__) <- lexToken
1311 --trace ("token: " ++ show tok__) $ do
1316 inp@(loc1,buf) <- getInput
1319 case alexScanUser exts inp sc of
1320 AlexEOF -> do setLastToken loc1 0
1321 return (T loc1 loc1 ITeof)
1322 AlexError (loc2,_) -> do failLocMsgP loc1 loc2 "lexical error"
1323 AlexSkip inp2 _ -> do
1326 AlexToken inp2@(end,buf2) len t -> do
1328 setLastToken loc1 len