1 -----------------------------------------------------------------------------
2 -- (c) The University of Glasgow, 2006
6 -- This is a combination of an Alex-generated lexer from a regex
7 -- definition, with some hand-coded bits.
9 -- Completely accurate information about token-spans within the source
10 -- file is maintained. Every token has a start and end SrcLoc attached to it.
12 -----------------------------------------------------------------------------
16 -- - parsing integers is a bit slow
17 -- - readRational is a bit slow
19 -- Known bugs, that were also in the previous version:
20 -- - M... should be 3 tokens, not 1.
21 -- - pragma-end should be only valid in a pragma
25 Token(..), lexer, pragState, mkPState, PState(..),
26 P(..), ParseResult(..), getSrcLoc,
27 failLocMsgP, failSpanMsgP, srcParseFail,
28 popContext, pushCurrentContext, setLastToken, setSrcLoc,
29 getLexState, popLexState, pushLexState,
30 extension, bangPatEnabled
33 #include "HsVersions.h"
35 import ErrUtils ( Message )
44 import Util ( maybePrefixMatch, readRational )
51 #if __GLASGOW_HASKELL__ >= 605
52 import Data.Char ( GeneralCategory(..), generalCategory )
54 import Compat.Unicode ( GeneralCategory(..), generalCategory )
59 $whitechar = [\ \t\n\r\f\v\xa0 $unispace]
60 $white_no_nl = $whitechar # \n
64 $decdigit = $ascdigit -- for now, should really be $digit (ToDo)
65 $digit = [$ascdigit $unidigit]
67 $special = [\(\)\,\;\[\]\`\{\}]
68 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
70 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
73 $asclarge = [A-Z \xc0-\xd6 \xd8-\xde]
74 $large = [$asclarge $unilarge]
77 $ascsmall = [a-z \xdf-\xf6 \xf8-\xff]
78 $small = [$ascsmall $unismall \_]
81 $graphic = [$small $large $symbol $digit $special $unigraphic \:\"\']
84 $hexit = [$decdigit A-F a-f]
85 $symchar = [$symbol \:]
87 $idchar = [$small $large $digit \']
89 @varid = $small $idchar*
90 @conid = $large $idchar*
92 @varsym = $symbol $symchar*
93 @consym = \: $symchar*
97 @hexadecimal = $hexit+
98 @exponent = [eE] [\-\+]? @decimal
100 -- we support the hierarchical module name extension:
103 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
107 -- everywhere: skip whitespace and comments
110 -- Everywhere: deal with nested comments. We explicitly rule out
111 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
112 -- (this can happen even though pragmas will normally take precedence due to
113 -- longest-match, because pragmas aren't valid in every state, but comments
115 "{-" / { notFollowedBy '#' } { nested_comment }
117 -- Single-line comments are a bit tricky. Haskell 98 says that two or
118 -- more dashes followed by a symbol should be parsed as a varsym, so we
119 -- have to exclude those.
120 -- The regex says: "munch all the characters after the dashes, as long as
121 -- the first one is not a symbol".
122 "--"\-* [^$symbol :] .* ;
123 "--"\-* / { atEOL } ;
125 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
126 -- blank lines) until we find a non-whitespace character, then do layout
129 -- One slight wibble here: what if the line begins with {-#? In
130 -- theory, we have to lex the pragma to see if it's one we recognise,
131 -- and if it is, then we backtrack and do_bol, otherwise we treat it
132 -- as a nested comment. We don't bother with this: if the line begins
133 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
136 ^\# (line)? { begin line_prag1 }
137 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
138 ^\# \! .* \n ; -- #!, for scripts
142 -- after a layout keyword (let, where, do, of), we begin a new layout
143 -- context if the curly brace is missing.
144 -- Careful! This stuff is quite delicate.
145 <layout, layout_do> {
146 \{ / { notFollowedBy '-' } { pop_and open_brace }
147 -- we might encounter {-# here, but {- has been handled already
149 ^\# (line)? { begin line_prag1 }
152 -- do is treated in a subtly different way, see new_layout_context
153 <layout> () { new_layout_context True }
154 <layout_do> () { new_layout_context False }
156 -- after a new layout context which was found to be to the left of the
157 -- previous context, we have generated a '{' token, and we now need to
158 -- generate a matching '}' token.
159 <layout_left> () { do_layout_left }
161 <0,option_prags,glaexts> \n { begin bol }
163 "{-#" $whitechar* (line|LINE) { begin line_prag2 }
165 -- single-line line pragmas, of the form
166 -- # <line> "<file>" <extra-stuff> \n
167 <line_prag1> $decdigit+ { setLine line_prag1a }
168 <line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
169 <line_prag1b> .* { pop }
171 -- Haskell-style line pragmas, of the form
172 -- {-# LINE <line> "<file>" #-}
173 <line_prag2> $decdigit+ { setLine line_prag2a }
174 <line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
175 <line_prag2b> "#-}"|"-}" { pop }
176 -- NOTE: accept -} at the end of a LINE pragma, for compatibility
177 -- with older versions of GHC which generated these.
179 -- We only want RULES pragmas to be picked up when -fglasgow-exts
180 -- is on, because the contents of the pragma is always written using
181 -- glasgow-exts syntax (using forall etc.), so if glasgow exts are not
182 -- enabled, we're sure to get a parse error.
183 -- (ToDo: we should really emit a warning when ignoring pragmas)
185 "{-#" $whitechar* (RULES|rules) { token ITrules_prag }
187 <0,option_prags,glaexts> {
188 "{-#" $whitechar* (INLINE|inline) { token (ITinline_prag True) }
189 "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline)
190 { token (ITinline_prag False) }
191 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
192 { token ITspec_prag }
193 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
194 $whitechar* (INLINE|inline) { token (ITspec_inline_prag True) }
195 "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
196 $whitechar* (NO(T?)INLINE|no(t?)inline)
197 { token (ITspec_inline_prag False) }
198 "{-#" $whitechar* (SOURCE|source) { token ITsource_prag }
199 "{-#" $whitechar* (DEPRECATED|deprecated)
200 { token ITdeprecated_prag }
201 "{-#" $whitechar* (SCC|scc) { token ITscc_prag }
202 "{-#" $whitechar* (CORE|core) { token ITcore_prag }
203 "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag }
205 "{-#" { nested_comment }
207 -- ToDo: should only be valid inside a pragma:
208 "#-}" { token ITclose_prag}
212 "{-#" $whitechar* (OPTIONS|options) { lex_string_prag IToptions_prag }
213 "{-#" $whitechar* (OPTIONS_GHC|options_ghc)
214 { lex_string_prag IToptions_prag }
215 "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag }
216 "{-#" $whitechar* (INCLUDE|include) { lex_string_prag ITinclude_prag }
219 -- '0' state: ordinary lexemes
220 -- 'glaexts' state: glasgow extensions (postfix '#', etc.)
225 "[:" / { ifExtension parrEnabled } { token ITopabrack }
226 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
230 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
231 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
232 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
233 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
234 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
235 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
236 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
237 "$(" / { ifExtension thEnabled } { token ITparenEscape }
241 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
242 { special IToparenbar }
243 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
247 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
248 \% @varid / { ifExtension ipEnabled } { skip_one_varid ITsplitipvarid }
252 "(#" / { notFollowedBySymbol } { token IToubxparen }
253 "#)" { token ITcubxparen }
254 "{|" { token ITocurlybar }
255 "|}" { token ITccurlybar }
258 <0,option_prags,glaexts> {
259 \( { special IToparen }
260 \) { special ITcparen }
261 \[ { special ITobrack }
262 \] { special ITcbrack }
263 \, { special ITcomma }
264 \; { special ITsemi }
265 \` { special ITbackquote }
271 <0,option_prags,glaexts> {
272 @qual @varid { check_qvarid }
273 @qual @conid { idtoken qconid }
275 @conid { idtoken conid }
278 -- after an illegal qvarid, such as 'M.let',
279 -- we back up and try again in the bad_qvarid state:
281 @conid { pop_and (idtoken conid) }
282 @qual @conid { pop_and (idtoken qconid) }
286 @qual @varid "#"+ { idtoken qvarid }
287 @qual @conid "#"+ { idtoken qconid }
288 @varid "#"+ { varid }
289 @conid "#"+ { idtoken conid }
295 @qual @varsym { idtoken qvarsym }
296 @qual @consym { idtoken qconsym }
302 @decimal { tok_decimal }
303 0[oO] @octal { tok_octal }
304 0[xX] @hexadecimal { tok_hexadecimal }
308 @decimal \# { prim_decimal }
309 0[oO] @octal \# { prim_octal }
310 0[xX] @hexadecimal \# { prim_hexadecimal }
313 <0,glaexts> @floating_point { strtoken tok_float }
314 <glaexts> @floating_point \# { init_strtoken 1 prim_float }
315 <glaexts> @floating_point \# \# { init_strtoken 2 prim_double }
317 -- Strings and chars are lexed by hand-written code. The reason is
318 -- that even if we recognise the string or char here in the regex
319 -- lexer, we would still have to parse the string afterward in order
320 -- to convert it to a String.
323 \" { lex_string_tok }
327 -- work around bug in Alex 2.0
328 #if __GLASGOW_HASKELL__ < 503
329 unsafeAt arr i = arr ! i
332 -- -----------------------------------------------------------------------------
336 = ITas -- Haskell keywords
360 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
362 | ITforall -- GHC extension keywords
376 | ITinline_prag Bool -- True <=> INLINE, False <=> NOINLINE
377 | ITspec_prag -- SPECIALISE
378 | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE)
384 | ITcore_prag -- hdaume: core annotations
387 | IToptions_prag String
388 | ITinclude_prag String
391 | ITdotdot -- reserved symbols
407 | ITbiglam -- GHC-extension symbols
409 | ITocurly -- special symbols
411 | ITocurlybar -- {|, for type applications
412 | ITccurlybar -- |}, for type applications
416 | ITopabrack -- [:, for parallel arrays with -fparr
417 | ITcpabrack -- :], for parallel arrays with -fparr
428 | ITvarid FastString -- identifiers
430 | ITvarsym FastString
431 | ITconsym FastString
432 | ITqvarid (FastString,FastString)
433 | ITqconid (FastString,FastString)
434 | ITqvarsym (FastString,FastString)
435 | ITqconsym (FastString,FastString)
437 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
438 | ITsplitipvarid FastString -- GHC extension: implicit param: %x
440 | ITpragma StringBuffer
443 | ITstring FastString
445 | ITrational Rational
448 | ITprimstring FastString
450 | ITprimfloat Rational
451 | ITprimdouble Rational
453 -- MetaHaskell extension tokens
454 | ITopenExpQuote -- [| or [e|
455 | ITopenPatQuote -- [p|
456 | ITopenDecQuote -- [d|
457 | ITopenTypQuote -- [t|
459 | ITidEscape FastString -- $x
460 | ITparenEscape -- $(
464 -- Arrow notation extension
471 | ITLarrowtail -- -<<
472 | ITRarrowtail -- >>-
474 | ITunknown String -- Used when the lexer can't make sense of it
475 | ITeof -- end of file token
477 deriving Show -- debugging
480 isSpecial :: Token -> Bool
481 -- If we see M.x, where x is a keyword, but
482 -- is special, we treat is as just plain M.x,
484 isSpecial ITas = True
485 isSpecial IThiding = True
486 isSpecial ITqualified = True
487 isSpecial ITforall = True
488 isSpecial ITexport = True
489 isSpecial ITlabel = True
490 isSpecial ITdynamic = True
491 isSpecial ITsafe = True
492 isSpecial ITthreadsafe = True
493 isSpecial ITunsafe = True
494 isSpecial ITccallconv = True
495 isSpecial ITstdcallconv = True
496 isSpecial ITmdo = True
499 -- the bitmap provided as the third component indicates whether the
500 -- corresponding extension keyword is valid under the extension options
501 -- provided to the compiler; if the extension corresponding to *any* of the
502 -- bits set in the bitmap is enabled, the keyword is valid (this setup
503 -- facilitates using a keyword in two different extensions that can be
504 -- activated independently)
506 reservedWordsFM = listToUFM $
507 map (\(x, y, z) -> (mkFastString x, (y, z)))
508 [( "_", ITunderscore, 0 ),
510 ( "case", ITcase, 0 ),
511 ( "class", ITclass, 0 ),
512 ( "data", ITdata, 0 ),
513 ( "default", ITdefault, 0 ),
514 ( "deriving", ITderiving, 0 ),
516 ( "else", ITelse, 0 ),
517 ( "hiding", IThiding, 0 ),
519 ( "import", ITimport, 0 ),
521 ( "infix", ITinfix, 0 ),
522 ( "infixl", ITinfixl, 0 ),
523 ( "infixr", ITinfixr, 0 ),
524 ( "instance", ITinstance, 0 ),
526 ( "module", ITmodule, 0 ),
527 ( "newtype", ITnewtype, 0 ),
529 ( "qualified", ITqualified, 0 ),
530 ( "then", ITthen, 0 ),
531 ( "type", ITtype, 0 ),
532 ( "where", ITwhere, 0 ),
533 ( "_scc_", ITscc, 0 ), -- ToDo: remove
535 ( "forall", ITforall, bit tvBit),
536 ( "mdo", ITmdo, bit glaExtsBit),
538 ( "foreign", ITforeign, bit ffiBit),
539 ( "export", ITexport, bit ffiBit),
540 ( "label", ITlabel, bit ffiBit),
541 ( "dynamic", ITdynamic, bit ffiBit),
542 ( "safe", ITsafe, bit ffiBit),
543 ( "threadsafe", ITthreadsafe, bit ffiBit),
544 ( "unsafe", ITunsafe, bit ffiBit),
545 ( "stdcall", ITstdcallconv, bit ffiBit),
546 ( "ccall", ITccallconv, bit ffiBit),
547 ( "dotnet", ITdotnet, bit ffiBit),
549 ( "rec", ITrec, bit arrowsBit),
550 ( "proc", ITproc, bit arrowsBit)
553 reservedSymsFM = listToUFM $
554 map (\ (x,y,z) -> (mkFastString x,(y,z)))
555 [ ("..", ITdotdot, 0)
556 ,(":", ITcolon, 0) -- (:) is a reserved op,
557 -- meaning only list cons
570 ,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT
571 ,(".", ITdot, bit tvBit) -- For 'forall a . t'
573 ,("-<", ITlarrowtail, bit arrowsBit)
574 ,(">-", ITrarrowtail, bit arrowsBit)
575 ,("-<<", ITLarrowtail, bit arrowsBit)
576 ,(">>-", ITRarrowtail, bit arrowsBit)
578 #if __GLASGOW_HASKELL__ >= 605
579 ,("λ", ITlam, bit glaExtsBit)
580 ,("∷", ITdcolon, bit glaExtsBit)
581 ,("⇒", ITdarrow, bit glaExtsBit)
582 ,("∀", ITforall, bit glaExtsBit)
583 ,("→", ITrarrow, bit glaExtsBit)
584 ,("←", ITlarrow, bit glaExtsBit)
585 ,("⋯", ITdotdot, bit glaExtsBit)
589 -- -----------------------------------------------------------------------------
592 type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
594 special :: Token -> Action
595 special tok span _buf len = return (L span tok)
597 token, layout_token :: Token -> Action
598 token t span buf len = return (L span t)
599 layout_token t span buf len = pushLexState layout >> return (L span t)
601 idtoken :: (StringBuffer -> Int -> Token) -> Action
602 idtoken f span buf len = return (L span $! (f buf len))
604 skip_one_varid :: (FastString -> Token) -> Action
605 skip_one_varid f span buf len
606 = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
608 strtoken :: (String -> Token) -> Action
609 strtoken f span buf len =
610 return (L span $! (f $! lexemeToString buf len))
612 init_strtoken :: Int -> (String -> Token) -> Action
613 -- like strtoken, but drops the last N character(s)
614 init_strtoken drop f span buf len =
615 return (L span $! (f $! lexemeToString buf (len-drop)))
617 begin :: Int -> Action
618 begin code _span _str _len = do pushLexState code; lexToken
621 pop _span _buf _len = do popLexState; lexToken
623 pop_and :: Action -> Action
624 pop_and act span buf len = do popLexState; act span buf len
626 notFollowedBy char _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf /= char
628 notFollowedBySymbol _ _ _ (AI _ _ buf)
629 = atEnd buf || currentChar buf `notElem` "!#$%&*+./<=>?@\\^|-~"
631 atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
633 ifExtension pred bits _ _ _ = pred bits
636 nested comments require traversing by hand, they can't be parsed
637 using regular expressions.
639 nested_comment :: Action
640 nested_comment span _str _len = do
643 where go 0 input = do setInput input; lexToken
645 case alexGetChar input of
650 case alexGetChar input of
652 Just ('\125',input) -> go (n-1) input
653 Just (c,_) -> go n input
655 case alexGetChar input of
657 Just ('-',input') -> go (n+1) input'
658 Just (c,input) -> go n input
661 err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
663 open_brace, close_brace :: Action
664 open_brace span _str _len = do
666 setContext (NoLayout:ctx)
667 return (L span ITocurly)
668 close_brace span _str _len = do
670 return (L span ITccurly)
672 -- We have to be careful not to count M.<varid> as a qualified name
673 -- when <varid> is a keyword. We hack around this by catching
674 -- the offending tokens afterward, and re-lexing in a different state.
675 check_qvarid span buf len = do
676 case lookupUFM reservedWordsFM var of
678 | not (isSpecial keyword) ->
682 b <- extension (\i -> exts .&. i /= 0)
685 _other -> return token
687 (mod,var) = splitQualName buf len
688 token = L span (ITqvarid (mod,var))
691 (AI _ offs _) <- getInput
692 setInput (AI (srcSpanStart span) (offs-len) buf)
693 pushLexState bad_qvarid
696 qvarid buf len = ITqvarid $! splitQualName buf len
697 qconid buf len = ITqconid $! splitQualName buf len
699 splitQualName :: StringBuffer -> Int -> (FastString,FastString)
700 -- takes a StringBuffer and a length, and returns the module name
701 -- and identifier parts of a qualified name. Splits at the *last* dot,
702 -- because of hierarchical module names.
703 splitQualName orig_buf len = split orig_buf orig_buf
706 | orig_buf `byteDiff` buf >= len = done dot_buf
707 | c == '.' = found_dot buf'
708 | otherwise = split buf' dot_buf
710 (c,buf') = nextChar buf
712 -- careful, we might get names like M....
713 -- so, if the character after the dot is not upper-case, this is
714 -- the end of the qualifier part.
715 found_dot buf -- buf points after the '.'
716 | isUpper c = split buf' buf
717 | otherwise = done buf
719 (c,buf') = nextChar buf
722 (lexemeToFastString orig_buf (qual_size - 1),
723 lexemeToFastString dot_buf (len - qual_size))
725 qual_size = orig_buf `byteDiff` dot_buf
728 case lookupUFM reservedWordsFM fs of
729 Just (keyword,0) -> do
731 return (L span keyword)
732 Just (keyword,exts) -> do
733 b <- extension (\i -> exts .&. i /= 0)
734 if b then do maybe_layout keyword
735 return (L span keyword)
736 else return (L span (ITvarid fs))
737 _other -> return (L span (ITvarid fs))
739 fs = lexemeToFastString buf len
741 conid buf len = ITconid fs
742 where fs = lexemeToFastString buf len
744 qvarsym buf len = ITqvarsym $! splitQualName buf len
745 qconsym buf len = ITqconsym $! splitQualName buf len
747 varsym = sym ITvarsym
748 consym = sym ITconsym
750 sym con span buf len =
751 case lookupUFM reservedSymsFM fs of
752 Just (keyword,0) -> return (L span keyword)
753 Just (keyword,exts) -> do
754 b <- extension (\i -> exts .&. i /= 0)
755 if b then return (L span keyword)
756 else return (L span $! con fs)
757 _other -> return (L span $! con fs)
759 fs = lexemeToFastString buf len
761 tok_decimal span buf len
762 = return (L span (ITinteger $! parseInteger buf len 10 octDecDigit))
764 tok_octal span buf len
765 = return (L span (ITinteger $! parseInteger (offsetBytes 2 buf) (len-2) 8 octDecDigit))
767 tok_hexadecimal span buf len
768 = return (L span (ITinteger $! parseInteger (offsetBytes 2 buf) (len-2) 16 hexDigit))
770 prim_decimal span buf len
771 = return (L span (ITprimint $! parseInteger buf (len-1) 10 octDecDigit))
773 prim_octal span buf len
774 = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 8 octDecDigit))
776 prim_hexadecimal span buf len
777 = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 16 hexDigit))
779 tok_float str = ITrational $! readRational str
780 prim_float str = ITprimfloat $! readRational str
781 prim_double str = ITprimdouble $! readRational str
783 -- -----------------------------------------------------------------------------
786 -- we're at the first token on a line, insert layout tokens if necessary
788 do_bol span _str _len = do
792 --trace "layout: inserting '}'" $ do
794 -- do NOT pop the lex state, we might have a ';' to insert
795 return (L span ITvccurly)
797 --trace "layout: inserting ';'" $ do
799 return (L span ITsemi)
804 -- certain keywords put us in the "layout" state, where we might
805 -- add an opening curly brace.
806 maybe_layout ITdo = pushLexState layout_do
807 maybe_layout ITmdo = pushLexState layout_do
808 maybe_layout ITof = pushLexState layout
809 maybe_layout ITlet = pushLexState layout
810 maybe_layout ITwhere = pushLexState layout
811 maybe_layout ITrec = pushLexState layout
812 maybe_layout _ = return ()
814 -- Pushing a new implicit layout context. If the indentation of the
815 -- next token is not greater than the previous layout context, then
816 -- Haskell 98 says that the new layout context should be empty; that is
817 -- the lexer must generate {}.
819 -- We are slightly more lenient than this: when the new context is started
820 -- by a 'do', then we allow the new context to be at the same indentation as
821 -- the previous context. This is what the 'strict' argument is for.
823 new_layout_context strict span _buf _len = do
825 (AI _ offset _) <- getInput
828 Layout prev_off : _ |
829 (strict && prev_off >= offset ||
830 not strict && prev_off > offset) -> do
831 -- token is indented to the left of the previous context.
832 -- we must generate a {} sequence now.
833 pushLexState layout_left
834 return (L span ITvocurly)
836 setContext (Layout offset : ctx)
837 return (L span ITvocurly)
839 do_layout_left span _buf _len = do
841 pushLexState bol -- we must be at the start of a line
842 return (L span ITvccurly)
844 -- -----------------------------------------------------------------------------
847 setLine :: Int -> Action
848 setLine code span buf len = do
849 let line = parseInteger buf len 10 octDecDigit
850 setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0)
851 -- subtract one: the line number refers to the *following* line
856 setFile :: Int -> Action
857 setFile code span buf len = do
858 let file = lexemeToFastString (stepOn buf) (len-2)
859 setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
865 -- -----------------------------------------------------------------------------
866 -- Options, includes and language pragmas.
868 lex_string_prag :: (String -> Token) -> Action
869 lex_string_prag mkTok span buf len
870 = do input <- getInput
874 return (L (mkSrcSpan start end) tok)
876 = if isString input "#-}"
877 then do setInput input
878 return (mkTok (reverse acc))
879 else case alexGetChar input of
880 Just (c,i) -> go (c:acc) i
884 = case alexGetChar i of
885 Just (c,i') | c == x -> isString i' xs
887 err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"
890 -- -----------------------------------------------------------------------------
893 -- This stuff is horrible. I hates it.
895 lex_string_tok :: Action
896 lex_string_tok span buf len = do
899 return (L (mkSrcSpan (srcSpanStart span) end) tok)
901 lex_string :: String -> P Token
904 case alexGetChar' i of
909 glaexts <- extension glaExtsEnabled
913 case alexGetChar' i of
917 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
918 else let s' = mkZFastString (reverse s) in
919 return (ITprimstring s')
920 -- mkZFastString is a hack to avoid encoding the
921 -- string in UTF-8. We just want the exact bytes.
923 return (ITstring (mkFastString (reverse s)))
925 return (ITstring (mkFastString (reverse s)))
928 | Just ('&',i) <- next -> do
929 setInput i; lex_string s
930 | Just (c,i) <- next, is_space c -> do
931 setInput i; lex_stringgap s
932 where next = alexGetChar' i
942 c | is_space c -> lex_stringgap s
946 lex_char_tok :: Action
947 -- Here we are basically parsing character literals, such as 'x' or '\n'
948 -- but, when Template Haskell is on, we additionally spot
949 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
950 -- but WIHTOUT CONSUMING the x or T part (the parser does that).
951 -- So we have to do two characters of lookahead: when we see 'x we need to
952 -- see if there's a trailing quote
953 lex_char_tok span buf len = do -- We've seen '
954 i1 <- getInput -- Look ahead to first character
955 let loc = srcSpanStart span
956 case alexGetChar' i1 of
959 Just ('\'', i2@(AI end2 _ _)) -> do -- We've seen ''
960 th_exts <- extension thEnabled
963 return (L (mkSrcSpan loc end2) ITtyQuote)
966 Just ('\\', i2@(AI end2 _ _)) -> do -- We've seen 'backslash
969 mc <- getCharOrFail -- Trailing quote
970 if mc == '\'' then finish_char_tok loc lit_ch
971 else do setInput i2; lit_error
973 Just (c, i2@(AI end2 _ _))
974 | not (isAny c) -> lit_error
977 -- We've seen 'x, where x is a valid character
978 -- (i.e. not newline etc) but not a quote or backslash
979 case alexGetChar' i2 of -- Look ahead one more character
981 Just ('\'', i3) -> do -- We've seen 'x'
983 finish_char_tok loc c
984 _other -> do -- We've seen 'x not followed by quote
985 -- If TH is on, just parse the quote only
986 th_exts <- extension thEnabled
987 let (AI end _ _) = i1
988 if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
989 else do setInput i2; lit_error
991 finish_char_tok :: SrcLoc -> Char -> P (Located Token)
992 finish_char_tok loc ch -- We've already seen the closing quote
993 -- Just need to check for trailing #
994 = do glaexts <- extension glaExtsEnabled
995 i@(AI end _ _) <- getInput
997 case alexGetChar' i of
998 Just ('#',i@(AI end _ _)) -> do
1000 return (L (mkSrcSpan loc end) (ITprimchar ch))
1002 return (L (mkSrcSpan loc end) (ITchar ch))
1004 return (L (mkSrcSpan loc end) (ITchar ch))
1006 lex_char :: Char -> AlexInput -> P Char
1009 '\\' -> do setInput inp; lex_escape
1010 c | isAny c -> do setInput inp; return c
1013 isAny c | c > '\xff' = isPrint c
1014 | otherwise = is_any c
1016 lex_escape :: P Char
1030 '^' -> do c <- getCharOrFail
1031 if c >= '@' && c <= '_'
1032 then return (chr (ord c - ord '@'))
1035 'x' -> readNum is_hexdigit 16 hexDigit
1036 'o' -> readNum is_octdigit 8 octDecDigit
1037 x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x)
1041 case alexGetChar' i of
1042 Nothing -> lit_error
1044 case alexGetChar' i2 of
1045 Nothing -> do setInput i2; lit_error
1047 let str = [c1,c2,c3] in
1048 case [ (c,rest) | (p,c) <- silly_escape_chars,
1049 Just rest <- [maybePrefixMatch p str] ] of
1050 (escape_char,[]):_ -> do
1053 (escape_char,_:_):_ -> do
1058 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
1059 readNum is_digit base conv = do
1063 then readNum2 is_digit base conv (conv c)
1064 else do setInput i; lit_error
1066 readNum2 is_digit base conv i = do
1069 where read i input = do
1070 case alexGetChar' input of
1071 Just (c,input') | is_digit c -> do
1072 read (i*base + conv c) input'
1074 if i >= 0 && i <= 0x10FFFF
1075 then do setInput input; return (chr i)
1078 silly_escape_chars = [
1115 -- before calling lit_error, ensure that the current input is pointing to
1116 -- the position of the error in the buffer. This is so that we can report
1117 -- a correct location to the user, but also so we can detect UTF-8 decoding
1118 -- errors if they occur.
1119 lit_error = lexError "lexical error in string/character literal"
1121 getCharOrFail :: P Char
1124 case alexGetChar' i of
1125 Nothing -> lexError "unexpected end-of-file in string/character literal"
1126 Just (c,i) -> do setInput i; return c
1128 -- -----------------------------------------------------------------------------
1138 SrcSpan -- The start and end of the text span related to
1139 -- the error. Might be used in environments which can
1140 -- show this span, e.g. by highlighting it.
1141 Message -- The error message
1143 data PState = PState {
1144 buffer :: StringBuffer,
1145 last_loc :: SrcSpan, -- pos of previous token
1146 last_offs :: !Int, -- offset of the previous token from the
1147 -- beginning of the current line.
1148 -- \t is equal to 8 spaces.
1149 last_len :: !Int, -- len of previous token
1150 loc :: SrcLoc, -- current loc (end of prev token + 1)
1151 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1152 context :: [LayoutContext],
1155 -- last_loc and last_len are used when generating error messages,
1156 -- and in pushCurrentContext only. Sigh, if only Happy passed the
1157 -- current token to happyError, we could at least get rid of last_len.
1158 -- Getting rid of last_loc would require finding another way to
1159 -- implement pushCurrentContext (which is only called from one place).
1161 newtype P a = P { unP :: PState -> ParseResult a }
1163 instance Monad P where
1169 returnP a = P $ \s -> POk s a
1171 thenP :: P a -> (a -> P b) -> P b
1172 (P m) `thenP` k = P $ \ s ->
1174 POk s1 a -> (unP (k a)) s1
1175 PFailed span err -> PFailed span err
1177 failP :: String -> P a
1178 failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1180 failMsgP :: String -> P a
1181 failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1183 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1184 failLocMsgP loc1 loc2 str = P $ \s -> PFailed (mkSrcSpan loc1 loc2) (text str)
1186 failSpanMsgP :: SrcSpan -> String -> P a
1187 failSpanMsgP span msg = P $ \s -> PFailed span (text msg)
1189 extension :: (Int -> Bool) -> P Bool
1190 extension p = P $ \s -> POk s (p $! extsBitmap s)
1193 getExts = P $ \s -> POk s (extsBitmap s)
1195 setSrcLoc :: SrcLoc -> P ()
1196 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1198 getSrcLoc :: P SrcLoc
1199 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1201 setLastToken :: SrcSpan -> Int -> P ()
1202 setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } ()
1204 data AlexInput = AI SrcLoc {-#UNPACK#-}!Int StringBuffer
1206 alexInputPrevChar :: AlexInput -> Char
1207 alexInputPrevChar (AI _ _ buf) = prevChar buf '\n'
1209 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1210 alexGetChar (AI loc ofs s)
1212 | otherwise = adj_c `seq` loc' `seq` ofs' `seq` s' `seq`
1213 Just (adj_c, (AI loc' ofs' s'))
1214 where (c,s') = nextChar s
1215 loc' = advanceSrcLoc loc c
1216 ofs' = advanceOffs c ofs
1224 other_graphic = '\x6'
1227 | c <= '\x06' = non_graphic
1230 case generalCategory c of
1231 UppercaseLetter -> upper
1232 LowercaseLetter -> lower
1233 TitlecaseLetter -> upper
1234 ModifierLetter -> other_graphic
1235 OtherLetter -> other_graphic
1236 NonSpacingMark -> other_graphic
1237 SpacingCombiningMark -> other_graphic
1238 EnclosingMark -> other_graphic
1239 DecimalNumber -> digit
1240 LetterNumber -> other_graphic
1241 OtherNumber -> other_graphic
1242 ConnectorPunctuation -> other_graphic
1243 DashPunctuation -> other_graphic
1244 OpenPunctuation -> other_graphic
1245 ClosePunctuation -> other_graphic
1246 InitialQuote -> other_graphic
1247 FinalQuote -> other_graphic
1248 OtherPunctuation -> other_graphic
1249 MathSymbol -> symbol
1250 CurrencySymbol -> symbol
1251 ModifierSymbol -> symbol
1252 OtherSymbol -> symbol
1254 _other -> non_graphic
1256 -- This version does not squash unicode characters, it is used when
1258 alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
1259 alexGetChar' (AI loc ofs s)
1261 | otherwise = c `seq` loc' `seq` ofs' `seq` s' `seq`
1262 Just (c, (AI loc' ofs' s'))
1263 where (c,s') = nextChar s
1264 loc' = advanceSrcLoc loc c
1265 ofs' = advanceOffs c ofs
1267 advanceOffs :: Char -> Int -> Int
1268 advanceOffs '\n' offs = 0
1269 advanceOffs '\t' offs = (offs `quot` 8 + 1) * 8
1270 advanceOffs _ offs = offs + 1
1272 getInput :: P AlexInput
1273 getInput = P $ \s@PState{ loc=l, last_offs=o, buffer=b } -> POk s (AI l o b)
1275 setInput :: AlexInput -> P ()
1276 setInput (AI l o b) = P $ \s -> POk s{ loc=l, last_offs=o, buffer=b } ()
1278 pushLexState :: Int -> P ()
1279 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1281 popLexState :: P Int
1282 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1284 getLexState :: P Int
1285 getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls
1287 -- for reasons of efficiency, flags indicating language extensions (eg,
1288 -- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed
1291 glaExtsBit, ffiBit, parrBit :: Int
1298 tvBit = 7 -- Scoped type variables enables 'forall' keyword
1299 bangPatBit = 8 -- Tells the parser to understand bang-patterns
1300 -- (doesn't affect the lexer)
1302 glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
1303 glaExtsEnabled flags = testBit flags glaExtsBit
1304 ffiEnabled flags = testBit flags ffiBit
1305 parrEnabled flags = testBit flags parrBit
1306 arrowsEnabled flags = testBit flags arrowsBit
1307 thEnabled flags = testBit flags thBit
1308 ipEnabled flags = testBit flags ipBit
1309 tvEnabled flags = testBit flags tvBit
1310 bangPatEnabled flags = testBit flags bangPatBit
1312 -- PState for parsing options pragmas
1314 pragState :: StringBuffer -> SrcLoc -> PState
1318 last_loc = mkSrcSpan loc loc,
1324 lex_state = [bol, option_prags, 0]
1328 -- create a parse state
1330 mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState
1331 mkPState buf loc flags =
1334 last_loc = mkSrcSpan loc loc,
1338 extsBitmap = fromIntegral bitmap,
1340 lex_state = [bol, if glaExtsEnabled bitmap then glaexts else 0]
1341 -- we begin in the layout state if toplev_layout is set
1344 bitmap = glaExtsBit `setBitIf` dopt Opt_GlasgowExts flags
1345 .|. ffiBit `setBitIf` dopt Opt_FFI flags
1346 .|. parrBit `setBitIf` dopt Opt_PArr flags
1347 .|. arrowsBit `setBitIf` dopt Opt_Arrows flags
1348 .|. thBit `setBitIf` dopt Opt_TH flags
1349 .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
1350 .|. tvBit `setBitIf` dopt Opt_ScopedTypeVariables flags
1351 .|. bangPatBit `setBitIf` dopt Opt_BangPatterns flags
1353 setBitIf :: Int -> Bool -> Int
1354 b `setBitIf` cond | cond = bit b
1357 getContext :: P [LayoutContext]
1358 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1360 setContext :: [LayoutContext] -> P ()
1361 setContext ctx = P $ \s -> POk s{context=ctx} ()
1364 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1365 loc = loc, last_len = len, last_loc = last_loc }) ->
1367 (_:tl) -> POk s{ context = tl } ()
1368 [] -> PFailed last_loc (srcParseErr buf len)
1370 -- Push a new layout context at the indentation of the last token read.
1371 -- This is only used at the outer level of a module when the 'module'
1372 -- keyword is missing.
1373 pushCurrentContext :: P ()
1374 pushCurrentContext = P $ \ s@PState{ last_offs=offs, last_len=len, context=ctx } ->
1375 POk s{context = Layout (offs-len) : ctx} ()
1377 getOffside :: P Ordering
1378 getOffside = P $ \s@PState{last_offs=offs, context=stk} ->
1379 let ord = case stk of
1380 (Layout n:_) -> compare offs n
1384 -- ---------------------------------------------------------------------------
1385 -- Construct a parse error
1388 :: StringBuffer -- current buffer (placed just after the last token)
1389 -> Int -- length of the previous token
1392 = hcat [ if null token
1393 then ptext SLIT("parse error (possibly incorrect indentation)")
1394 else hcat [ptext SLIT("parse error on input "),
1395 char '`', text token, char '\'']
1397 where token = lexemeToString (offsetBytes (-len) buf) len
1399 -- Report a parse failure, giving the span of the previous token as
1400 -- the location of the error. This is the entry point for errors
1401 -- detected during parsing.
1403 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1404 last_loc = last_loc } ->
1405 PFailed last_loc (srcParseErr buf len)
1407 -- A lexical error is reported at a particular position in the source file,
1408 -- not over a token range.
1409 lexError :: String -> P a
1412 i@(AI end _ buf) <- getInput
1413 reportLexError loc end buf str
1415 -- -----------------------------------------------------------------------------
1416 -- This is the top-level function: called from the parser each time a
1417 -- new token is to be read from the input.
1419 lexer :: (Located Token -> P a) -> P a
1421 tok@(L _ tok__) <- lexToken
1422 --trace ("token: " ++ show tok__) $ do
1425 lexToken :: P (Located Token)
1427 inp@(AI loc1 _ buf) <- getInput
1430 case alexScanUser exts inp sc of
1431 AlexEOF -> do let span = mkSrcSpan loc1 loc1
1433 return (L span ITeof)
1434 AlexError (AI loc2 _ buf) -> do
1435 reportLexError loc1 loc2 buf "lexical error"
1436 AlexSkip inp2 _ -> do
1439 AlexToken inp2@(AI end _ buf2) len t -> do
1441 let span = mkSrcSpan loc1 end
1442 let bytes = byteDiff buf buf2
1443 span `seq` setLastToken span bytes
1446 -- ToDo: Alex reports the buffer at the start of the erroneous lexeme,
1447 -- but it would be more informative to report the location where the
1448 -- error was actually discovered, especially if this is a decoding
1450 reportLexError loc1 loc2 buf str =
1452 c = fst (nextChar buf)
1454 if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar#
1455 then failLocMsgP loc2 loc2 "UTF-8 decoding error"
1456 else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c)