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 }
291 <glaexts> "``" (([$graphic $whitechar] # \') | \' ([$graphic $whitechar] # \'))*
295 -- work around bug in Alex 2.0
296 #if __GLASGOW_HASKELL__ < 503
297 unsafeAt arr i = arr ! i
300 -- -----------------------------------------------------------------------------
303 data Token = T SrcLoc{-start-} SrcLoc{-end-} Token__
306 = ITas -- Haskell keywords
330 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
332 | ITforall -- GHC extension keywords
344 | ITccall (Bool,Bool,Safety) -- (is_dyn, is_casm, may_gc)
347 | ITspecialise_prag -- Pragmas
355 | ITcore_prag -- hdaume: core annotations
358 | ITdotdot -- reserved symbols
374 | ITbiglam -- GHC-extension symbols
376 | ITocurly -- special symbols
378 | ITocurlybar -- {|, for type applications
379 | ITccurlybar -- |}, for type applications
383 | ITopabrack -- [:, for parallel arrays with -fparr
384 | ITcpabrack -- :], for parallel arrays with -fparr
395 | ITvarid FastString -- identifiers
397 | ITvarsym FastString
398 | ITconsym FastString
399 | ITqvarid (FastString,FastString)
400 | ITqconid (FastString,FastString)
401 | ITqvarsym (FastString,FastString)
402 | ITqconsym (FastString,FastString)
404 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
405 | ITsplitipvarid FastString -- GHC extension: implicit param: %x
407 | ITpragma StringBuffer
410 | ITstring FastString
412 | ITrational Rational
415 | ITprimstring FastString
417 | ITprimfloat Rational
418 | ITprimdouble Rational
419 | ITlitlit FastString
421 -- MetaHaskell extension tokens
422 | ITopenExpQuote -- [| or [e|
423 | ITopenPatQuote -- [p|
424 | ITopenDecQuote -- [d|
425 | ITopenTypQuote -- [t|
427 | ITidEscape FastString -- $x
428 | ITparenEscape -- $(
433 -- Arrow notation extension
440 | ITLarrowtail -- -<<
441 | ITRarrowtail -- >>-
443 | ITunknown String -- Used when the lexer can't make sense of it
444 | ITeof -- end of file token
446 deriving Show -- debugging
449 isSpecial :: Token__ -> Bool
450 -- If we see M.x, where x is a keyword, but
451 -- is special, we treat is as just plain M.x,
453 isSpecial ITas = True
454 isSpecial IThiding = True
455 isSpecial ITqualified = True
456 isSpecial ITforall = True
457 isSpecial ITexport = True
458 isSpecial ITlabel = True
459 isSpecial ITdynamic = True
460 isSpecial ITsafe = True
461 isSpecial ITthreadsafe = True
462 isSpecial ITunsafe = True
463 isSpecial ITwith = True
464 isSpecial ITccallconv = True
465 isSpecial ITstdcallconv = True
466 isSpecial ITmdo = True
469 -- the bitmap provided as the third component indicates whether the
470 -- corresponding extension keyword is valid under the extension options
471 -- provided to the compiler; if the extension corresponding to *any* of the
472 -- bits set in the bitmap is enabled, the keyword is valid (this setup
473 -- facilitates using a keyword in two different extensions that can be
474 -- activated independently)
476 reservedWordsFM = listToUFM $
477 map (\(x, y, z) -> (mkFastString x, (y, z)))
478 [( "_", ITunderscore, 0 ),
480 ( "case", ITcase, 0 ),
481 ( "class", ITclass, 0 ),
482 ( "data", ITdata, 0 ),
483 ( "default", ITdefault, 0 ),
484 ( "deriving", ITderiving, 0 ),
486 ( "else", ITelse, 0 ),
487 ( "hiding", IThiding, 0 ),
489 ( "import", ITimport, 0 ),
491 ( "infix", ITinfix, 0 ),
492 ( "infixl", ITinfixl, 0 ),
493 ( "infixr", ITinfixr, 0 ),
494 ( "instance", ITinstance, 0 ),
496 ( "module", ITmodule, 0 ),
497 ( "newtype", ITnewtype, 0 ),
499 ( "qualified", ITqualified, 0 ),
500 ( "then", ITthen, 0 ),
501 ( "type", ITtype, 0 ),
502 ( "where", ITwhere, 0 ),
503 ( "_scc_", ITscc, 0 ), -- ToDo: remove
505 ( "forall", ITforall, bit glaExtsBit),
506 ( "mdo", ITmdo, bit glaExtsBit),
507 ( "reifyDecl", ITreifyDecl, bit thBit),
508 ( "reifyType", ITreifyType, bit thBit),
509 ( "reifyFixity",ITreifyFixity, bit thBit),
511 ( "foreign", ITforeign, bit ffiBit),
512 ( "export", ITexport, bit ffiBit),
513 ( "label", ITlabel, bit ffiBit),
514 ( "dynamic", ITdynamic, bit ffiBit),
515 ( "safe", ITsafe, bit ffiBit),
516 ( "threadsafe", ITthreadsafe, bit ffiBit),
517 ( "unsafe", ITunsafe, bit ffiBit),
518 ( "stdcall", ITstdcallconv, bit ffiBit),
519 ( "ccall", ITccallconv, bit ffiBit),
520 ( "dotnet", ITdotnet, bit ffiBit),
522 ( "with", ITwith, bit withBit),
524 ( "rec", ITrec, bit arrowsBit),
525 ( "proc", ITproc, bit arrowsBit),
528 ("_ccall_", ITccall (False, False, PlayRisky),
530 ("_ccall_GC_", ITccall (False, False, PlaySafe False),
532 ("_casm_", ITccall (False, True, PlayRisky),
534 ("_casm_GC_", ITccall (False, True, PlaySafe False),
538 reservedSymsFM = listToUFM $
539 map (\ (x,y,z) -> (mkFastString x,(y,z)))
540 [ ("..", ITdotdot, 0)
541 ,(":", ITcolon, 0) -- (:) is a reserved op,
542 -- meaning only list cons
555 ,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT
556 ,(".", ITdot, bit glaExtsBit) -- For 'forall a . t'
558 ,("-<", ITlarrowtail, bit arrowsBit)
559 ,(">-", ITrarrowtail, bit arrowsBit)
560 ,("-<<", ITLarrowtail, bit arrowsBit)
561 ,(">>-", ITRarrowtail, bit arrowsBit)
564 -- -----------------------------------------------------------------------------
567 type Action = SrcLoc -> SrcLoc -> StringBuffer -> Int -> P Token
569 special :: Token__ -> Action
570 special tok loc end _buf len = return (T loc end tok)
572 token, layout_token :: Token__ -> Action
573 token t loc end buf len = return (T loc end t)
574 layout_token t loc end buf len = pushLexState layout >> return (T loc end t)
576 idtoken :: (StringBuffer -> Int -> Token__) -> Action
577 idtoken f loc end buf len = return (T loc end $! (f buf len))
579 skip_one_varid :: (FastString -> Token__) -> Action
580 skip_one_varid f loc end buf len
581 = return (T loc end $! f (lexemeToFastString (stepOn buf) (len-1)))
583 strtoken :: (String -> Token__) -> Action
584 strtoken f loc end buf len =
585 return (T loc end $! (f $! lexemeToString buf len))
587 init_strtoken :: Int -> (String -> Token__) -> Action
588 -- like strtoken, but drops the last N character(s)
589 init_strtoken drop f loc end buf len =
590 return (T loc end $! (f $! lexemeToString buf (len-drop)))
592 begin :: Int -> Action
593 begin code _loc _end _str _len = do pushLexState code; lexToken
596 pop _loc _end _buf _len = do popLexState; lexToken
598 pop_and :: Action -> Action
599 pop_and act loc end buf len = do popLexState; act loc end buf len
601 notFollowedBy char _ _ _ (_,buf) = atEnd buf || currentChar buf /= char
603 ifExtension pred bits _ _ _ = pred bits
606 nested comments require traversing by hand, they can't be parsed
607 using regular expressions.
609 nested_comment :: Action
610 nested_comment loc _end _str _len = do
613 where go 0 input = do setInput input; lexToken
615 case alexGetChar input of
620 case alexGetChar input of
622 Just ('\125',input) -> go (n-1) input
623 Just (c,_) -> go n input
625 case alexGetChar input of
627 Just ('-',input') -> go (n+1) input'
628 Just (c,input) -> go n input
631 err input = do failLocMsgP loc (fst input) "unterminated `{-'"
633 open_brace, close_brace :: Action
634 open_brace loc end _str _len = do
636 setContext (NoLayout:ctx)
637 return (T loc end ITocurly)
638 close_brace loc end _str _len = do
640 return (T loc end ITccurly)
642 -- We have to be careful not to count M.<varid> as a qualified name
643 -- when <varid> is a keyword. We hack around this by catching
644 -- the offending tokens afterward, and re-lexing in a different state.
645 check_qvarid loc end buf len = do
646 case lookupUFM reservedWordsFM var of
648 | not (isSpecial keyword) ->
652 b <- extension (\i -> exts .&. i /= 0)
655 _other -> return token
657 (mod,var) = splitQualName buf len
658 token = T loc end (ITqvarid (mod,var))
662 pushLexState bad_qvarid
665 qvarid buf len = ITqvarid $! splitQualName buf len
666 qconid buf len = ITqconid $! splitQualName buf len
668 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
669 -- takes a StringBuffer and a length, and returns the module name
670 -- and identifier parts of a qualified name. Splits at the *last* dot,
671 -- because of hierarchical module names.
672 splitQualName orig_buf len = split orig_buf 0 0
675 | n == len = done dot_off
676 | lookAhead buf n == '.' = split2 buf n (n+1)
677 | otherwise = split buf dot_off (n+1)
679 -- careful, we might get names like M....
680 -- so, if the character after the dot is not upper-case, this is
681 -- the end of the qualifier part.
683 | isUpper (lookAhead buf n) = split buf dot_off (n+1)
684 | otherwise = done dot_off
687 (lexemeToFastString orig_buf dot_off,
688 lexemeToFastString (stepOnBy (dot_off+1) orig_buf) (len - dot_off -1))
690 varid loc end buf len =
691 case lookupUFM reservedWordsFM fs of
692 Just (keyword,0) -> do
694 return (T loc end keyword)
695 Just (keyword,exts) -> do
696 b <- extension (\i -> exts .&. i /= 0)
697 if b then do maybe_layout keyword
698 return (T loc end keyword)
699 else return (T loc end (ITvarid fs))
700 _other -> return (T loc end (ITvarid fs))
702 fs = lexemeToFastString buf len
704 conid buf len = ITconid fs
705 where fs = lexemeToFastString buf len
707 qvarsym buf len = ITqvarsym $! splitQualName buf len
708 qconsym buf len = ITqconsym $! splitQualName buf len
710 varsym = sym ITvarsym
711 consym = sym ITconsym
713 sym con loc end buf len =
714 case lookupUFM reservedSymsFM fs of
715 Just (keyword,0) -> return (T loc end keyword)
716 Just (keyword,exts) -> do
717 b <- extension (\i -> exts .&. i /= 0)
718 if b then return (T loc end keyword)
719 else return (T loc end $! con fs)
720 _other -> return (T loc end $! con fs)
722 fs = lexemeToFastString buf len
724 tok_decimal loc end buf len
725 = return (T loc end (ITinteger $! parseInteger buf len 10 oct_or_dec))
727 tok_octal loc end buf len
728 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 8 oct_or_dec))
730 tok_hexadecimal loc end buf len
731 = return (T loc end (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 16 hex))
733 prim_decimal loc end buf len
734 = return (T loc end (ITprimint $! parseInteger buf (len-1) 10 oct_or_dec))
736 prim_octal loc end buf len
737 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 8 oct_or_dec))
739 prim_hexadecimal loc end buf len
740 = return (T loc end (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 16 hex))
742 tok_float str = ITrational $! readRational__ str
743 prim_float str = ITprimfloat $! readRational__ str
744 prim_double str = ITprimdouble $! readRational__ str
746 parseInteger :: StringBuffer -> Int -> Integer -> (Char->Int) -> Integer
747 parseInteger buf len radix to_int
749 where go i x | i == len = x
750 | otherwise = go (i+1) (x * radix + toInteger (to_int (lookAhead buf i)))
753 clitlit loc end buf len =
754 return (T loc end (ITlitlit $! lexemeToFastString (stepOnBy 2 buf) (len-4)))
756 -- -----------------------------------------------------------------------------
759 -- we're at the first token on a line, insert layout tokens if necessary
761 do_bol loc end _str _len = do
762 pos <- getOffside end
765 --trace "layout: inserting '}'" $ do
767 -- do NOT pop the lex state, we might have a ';' to insert
768 return (T loc end ITvccurly)
770 --trace "layout: inserting ';'" $ do
772 return (T loc end ITsemi)
777 -- certain keywords put us in the "layout" state, where we might
778 -- add an opening curly brace.
779 maybe_layout ITdo = pushLexState layout_do
780 maybe_layout ITmdo = pushLexState layout_do
781 maybe_layout ITof = pushLexState layout
782 maybe_layout ITlet = pushLexState layout
783 maybe_layout ITwhere = pushLexState layout
784 maybe_layout ITrec = pushLexState layout
785 maybe_layout _ = return ()
787 -- Pushing a new implicit layout context. If the indentation of the
788 -- next token is not greater than the previous layout context, then
789 -- Haskell 98 says that the new layout context should be empty; that is
790 -- the lexer must generate {}.
792 -- We are slightly more lenient than this: when the new context is started
793 -- by a 'do', then we allow the new context to be at the same indentation as
794 -- the previous context. This is what the 'strict' argument is for.
796 new_layout_context strict loc end _buf _len = do
798 let offset = srcLocCol loc
801 Layout prev_off : _ |
802 (strict && prev_off >= offset ||
803 not strict && prev_off > offset) -> do
804 -- token is indented to the left of the previous context.
805 -- we must generate a {} sequence now.
806 pushLexState layout_left
807 return (T loc end ITvocurly)
809 setContext (Layout offset : ctx)
810 return (T loc end ITvocurly)
812 do_layout_left loc end _buf _len = do
814 pushLexState bol -- we must be at the start of a line
815 return (T loc end ITvccurly)
817 -- -----------------------------------------------------------------------------
820 set_line :: Int -> Action
821 set_line code loc end buf len = do
822 let line = parseInteger buf len 10 oct_or_dec
823 setSrcLoc (mkSrcLoc (srcLocFile end) (fromIntegral line - 1) 0)
824 -- subtract one: the line number refers to the *following* line
829 set_file :: Int -> Action
830 set_file code loc end buf len = do
831 let file = lexemeToFastString (stepOn buf) (len-2)
832 setSrcLoc (mkSrcLoc file (srcLocLine end) (srcLocCol end))
837 -- -----------------------------------------------------------------------------
840 -- This stuff is horrible. I hates it.
842 lex_string_tok :: Action
843 lex_string_tok loc end buf len = do
846 return (T loc end tok)
848 lex_string :: String -> P Token__
851 case alexGetChar i of
856 glaexts <- extension glaExtsEnabled
860 case alexGetChar i of
864 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
865 else let s' = mkFastStringNarrow (reverse s) in
866 -- always a narrow string/byte array
867 return (ITprimstring s')
869 return (ITstring (mkFastString (reverse s)))
871 return (ITstring (mkFastString (reverse s)))
874 | Just ('&',i) <- next -> do
875 setInput i; lex_string s
876 | Just (c,i) <- next, is_space c -> do
877 setInput i; lex_stringgap s
878 where next = alexGetChar i
889 c | is_space c -> lex_stringgap s
893 lex_char_tok :: Action
894 lex_char_tok loc _end buf len = do
899 glaexts <- extension glaExtsEnabled
902 i@(end,_) <- getInput
903 case alexGetChar i of
904 Just ('#',i@(end,_)) -> do
906 return (T loc end (ITprimchar c))
908 return (T loc end (ITchar c))
911 return (T loc end (ITchar c))
920 c | is_any c -> return c
937 '^' -> do c <- getCharOrFail
938 if c >= '@' && c <= '_'
939 then return (chr (ord c - ord '@'))
942 'x' -> readNum is_hexdigit 16 hex
943 'o' -> readNum is_octdigit 8 oct_or_dec
944 x | is_digit x -> readNum2 is_digit 10 oct_or_dec (oct_or_dec x)
948 case alexGetChar i of
951 case alexGetChar i2 of
954 let str = [c1,c2,c3] in
955 case [ (c,rest) | (p,c) <- silly_escape_chars,
956 Just rest <- [maybePrefixMatch p str] ] of
957 (escape_char,[]):_ -> do
960 (escape_char,_:_):_ -> do
965 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
966 readNum is_digit base conv = do
969 then readNum2 is_digit base conv (conv c)
972 readNum2 is_digit base conv i = do
975 where read i input = do
976 case alexGetChar input of
977 Just (c,input') | is_digit c -> do
978 read (i*base + conv c) input'
981 if i >= 0 && i <= 0x10FFFF
987 || (c >= 'a' && c <= 'f')
988 || (c >= 'A' && c <= 'F')
990 hex c | is_digit c = ord c - ord '0'
991 | otherwise = ord (to_lower c) - ord 'a' + 10
993 oct_or_dec c = ord c - ord '0'
995 is_octdigit c = c >= '0' && c <= '7'
998 | c >= 'A' && c <= 'Z' = chr (ord c - (ord 'A' - ord 'a'))
1001 silly_escape_chars = [
1038 lit_error = lexError "lexical error in string/character literal"
1040 getCharOrFail :: P Char
1043 case alexGetChar i of
1044 Nothing -> lexError "unexpected end-of-file in string/character literal"
1045 Just (c,i) -> do setInput i; return c
1047 -- -----------------------------------------------------------------------------
1050 readRational :: ReadS Rational -- NB: doesn't handle leading "-"
1052 (n,d,s) <- readFix r
1054 return ((n%1)*10^^(k-d), t)
1057 (ds,s) <- lexDecDigits r
1058 (ds',t) <- lexDotDigits s
1059 return (read (ds++ds'), length ds', t)
1061 readExp (e:s) | e `elem` "eE" = readExp' s
1062 readExp s = return (0,s)
1064 readExp' ('+':s) = readDec s
1065 readExp' ('-':s) = do
1068 readExp' s = readDec s
1071 (ds,r) <- nonnull isDigit s
1072 return (foldl1 (\n d -> n * 10 + d) [ ord d - ord '0' | d <- ds ],
1075 lexDecDigits = nonnull isDigit
1077 lexDotDigits ('.':s) = return (span isDigit s)
1078 lexDotDigits s = return ("",s)
1080 nonnull p s = do (cs@(_:_),t) <- return (span p s)
1083 readRational__ :: String -> Rational -- NB: *does* handle a leading "-"
1084 readRational__ top_s
1086 '-' : xs -> - (read_me xs)
1090 = case (do { (x,"") <- readRational s ; return x }) of
1092 [] -> error ("readRational__: no parse:" ++ top_s)
1093 _ -> error ("readRational__: ambiguous parse:" ++ top_s)
1095 -- -----------------------------------------------------------------------------
1105 SrcLoc SrcLoc -- The start and end of the text span related to
1106 -- the error. Might be used in environments which can
1107 -- show this span, e.g. by highlighting it.
1108 Message -- The error message
1110 showPFailed loc1 loc2 err
1111 = showSDoc (hcat [ppr loc1, text ": ", err])
1113 data PState = PState {
1114 buffer :: StringBuffer,
1115 last_loc :: SrcLoc, -- pos of previous token
1116 last_len :: !Int, -- len of previous token
1117 loc :: SrcLoc, -- current loc (end of prev token + 1)
1118 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1119 context :: [LayoutContext],
1122 -- last_loc and last_len are used when generating error messages,
1123 -- and in pushCurrentContext only.
1125 newtype P a = P { unP :: PState -> ParseResult a }
1127 instance Monad P where
1133 returnP a = P $ \s -> POk s a
1135 thenP :: P a -> (a -> P b) -> P b
1136 (P m) `thenP` k = P $ \ s ->
1138 POk s1 a -> (unP (k a)) s1
1139 PFailed l1 l2 err -> PFailed l1 l2 err
1141 failP :: String -> P a
1142 failP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1144 failMsgP :: String -> P a
1145 failMsgP msg = P $ \s -> PFailed (last_loc s) (loc s) (text msg)
1147 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1148 failLocMsgP loc1 loc2 str = P $ \s -> PFailed loc1 loc2 (text str)
1150 extension :: (Int -> Bool) -> P Bool
1151 extension p = P $ \s -> POk s (p $! extsBitmap s)
1154 getExts = P $ \s -> POk s (extsBitmap s)
1156 setSrcLoc :: SrcLoc -> P ()
1157 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1159 -- tmp, for supporting stuff in RdrHsSyn. The scope better not include
1160 -- any calls to the lexer, because it assumes things about the SrcLoc.
1161 setSrcLocFor :: SrcLoc -> P a -> P a
1162 setSrcLocFor new_loc scope = P $ \s@PState{ loc = old_loc } ->
1163 case unP scope s{loc=new_loc} of
1164 PFailed l1 l2 msg -> PFailed l1 l2 msg
1167 getSrcLoc :: P SrcLoc
1168 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1170 setLastToken :: SrcLoc -> Int -> P ()
1171 setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } ()
1173 type AlexInput = (SrcLoc,StringBuffer)
1175 alexInputPrevChar :: AlexInput -> Char
1176 alexInputPrevChar (_,s) = prevChar s '\n'
1178 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1181 | otherwise = c `seq` loc' `seq` s' `seq` Just (c, (loc', s'))
1182 where c = currentChar s
1183 loc' = advanceSrcLoc loc c
1186 getInput :: P AlexInput
1187 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (l,b)
1189 setInput :: AlexInput -> P ()
1190 setInput (l,b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1192 pushLexState :: Int -> P ()
1193 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1195 popLexState :: P Int
1196 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1198 getLexState :: P Int
1199 getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls
1201 -- for reasons of efficiency, flags indicating language extensions (eg,
1202 -- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
1205 glaExtsBit, ffiBit, parrBit :: Int
1214 glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1215 glaExtsEnabled flags = testBit flags glaExtsBit
1216 ffiEnabled flags = testBit flags ffiBit
1217 withEnabled flags = testBit flags withBit
1218 parrEnabled flags = testBit flags parrBit
1219 arrowsEnabled flags = testBit flags arrowsBit
1220 thEnabled flags = testBit flags thBit
1221 ipEnabled flags = testBit flags ipBit
1223 -- create a parse state
1225 mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
1226 mkPState buf loc flags =
1232 extsBitmap = fromIntegral bitmap,
1234 lex_state = [bol, if glaExtsEnabled bitmap then glaexts else 0]
1235 -- we begin in the layout state if toplev_layout is set
1238 bitmap = glaExtsBit `setBitIf` dopt Opt_GlasgowExts flags
1239 .|. ffiBit `setBitIf` dopt Opt_FFI flags
1240 .|. withBit `setBitIf` dopt Opt_With flags
1241 .|. parrBit `setBitIf` dopt Opt_PArr flags
1242 .|. arrowsBit `setBitIf` dopt Opt_Arrows flags
1243 .|. thBit `setBitIf` dopt Opt_TH flags
1244 .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
1246 setBitIf :: Int -> Bool -> Int
1247 b `setBitIf` cond | cond = bit b
1250 getContext :: P [LayoutContext]
1251 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1253 setContext :: [LayoutContext] -> P ()
1254 setContext ctx = P $ \s -> POk s{context=ctx} ()
1257 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1258 loc = loc, last_len = len, last_loc = last_loc }) ->
1260 (_:tl) -> POk s{ context = tl } ()
1261 [] -> PFailed last_loc loc (srcParseErr buf len)
1263 -- Push a new layout context at the indentation of the last token read.
1264 -- This is only used at the outer level of a module when the 'module'
1265 -- keyword is missing.
1266 pushCurrentContext :: P ()
1267 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1268 POk s{ context = Layout (srcLocCol loc) : ctx} ()
1270 getOffside :: SrcLoc -> P Ordering
1271 getOffside loc = P $ \s@PState{context=stk} ->
1272 let ord = case stk of
1273 (Layout n:_) -> compare (srcLocCol loc) n
1277 -- ---------------------------------------------------------------------------
1278 -- Construct a parse error
1281 :: StringBuffer -- current buffer (placed just after the last token)
1282 -> Int -- length of the previous token
1285 = hcat [ if null token
1286 then ptext SLIT("parse error (possibly incorrect indentation)")
1287 else hcat [ptext SLIT("parse error on input "),
1288 char '`', text token, char '\'']
1290 where token = lexemeToString (stepOnBy (-len) buf) len
1292 -- Report a parse failure, giving the span of the previous token as
1293 -- the location of the error. This is the entry point for errors
1294 -- detected during parsing.
1296 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1297 last_loc = last_loc, loc = loc } ->
1298 PFailed last_loc loc (srcParseErr buf len)
1300 -- A lexical error is reported at a particular position in the source file,
1301 -- not over a token range. TODO: this is slightly wrong, because we record
1302 -- the error at the character position following the one which caused the
1303 -- error. We should somehow back up by one character.
1304 lexError :: String -> P a
1307 failLocMsgP loc loc str
1309 -- -----------------------------------------------------------------------------
1310 -- This is the top-level function: called from the parser each time a
1311 -- new token is to be read from the input.
1313 lexer :: (Token -> P a) -> P a
1315 tok@(T _ _ tok__) <- lexToken
1316 --trace ("token: " ++ show tok__) $ do
1321 inp@(loc1,buf) <- getInput
1324 case alexScanUser exts inp sc of
1325 AlexEOF -> do setLastToken loc1 0
1326 return (T loc1 loc1 ITeof)
1327 AlexError (loc2,_) -> do failLocMsgP loc1 loc2 "lexical error"
1328 AlexSkip inp2 _ -> do
1331 AlexToken inp2@(end,buf2) len t -> do
1333 setLastToken loc1 len