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 -----------------------------------------------------------------------------
15 -- - parsing integers is a bit slow
16 -- - readRational is a bit slow
18 -- Known bugs, that were also in the previous version:
19 -- - M... should be 3 tokens, not 1.
20 -- - pragma-end should be only valid in a pragma
22 -- qualified operator NOTES.
24 -- - If M.(+) is a single lexeme, then..
25 -- - Probably (+) should be a single lexeme too, for consistency.
26 -- Otherwise ( + ) would be a prefix operator, but M.( + ) would not be.
27 -- - But we have to rule out reserved operators, otherwise (..) becomes
28 -- a different lexeme.
29 -- - Should we therefore also rule out reserved operators in the qualified
30 -- form? This is quite difficult to achieve. We don't do it for
34 -- XXX The above flags turn off warnings in the generated code:
35 {-# OPTIONS_GHC -fno-warn-unused-matches #-}
36 {-# OPTIONS_GHC -fno-warn-unused-binds #-}
37 {-# OPTIONS_GHC -fno-warn-unused-imports #-}
38 {-# OPTIONS_GHC -fno-warn-missing-signatures #-}
39 -- But alex still generates some code that causes the "lazy unlifted bindings"
40 -- warning, and old compilers don't know about it so we can't easily turn
41 -- it off, so for now we use the sledge hammer:
42 {-# OPTIONS_GHC -w #-}
44 {-# OPTIONS_GHC -funbox-strict-fields #-}
47 Token(..), lexer, pragState, mkPState, PState(..),
48 P(..), ParseResult(..), getSrcLoc,
49 getPState, getDynFlags, withThisPackage,
50 failLocMsgP, failSpanMsgP, srcParseFail,
52 popContext, pushCurrentContext, setLastToken, setSrcLoc,
53 getLexState, popLexState, pushLexState,
54 extension, bangPatEnabled, datatypeContextsEnabled,
69 import BasicTypes ( InlineSpec(..), RuleMatchInfo(..) )
70 import Util ( readRational )
78 import qualified Data.Map as Map
82 $unispace = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
83 $whitechar = [\ \n\r\f\v $unispace]
84 $white_no_nl = $whitechar # \n
88 $unidigit = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
89 $decdigit = $ascdigit -- for now, should really be $digit (ToDo)
90 $digit = [$ascdigit $unidigit]
92 $special = [\(\)\,\;\[\]\`\{\}]
93 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
94 $unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
95 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
97 $unilarge = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
99 $large = [$asclarge $unilarge]
101 $unismall = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
103 $small = [$ascsmall $unismall \_]
105 $unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
106 $graphic = [$small $large $symbol $digit $special $unigraphic \:\"\']
109 $hexit = [$decdigit A-F a-f]
110 $symchar = [$symbol \:]
112 $idchar = [$small $large $digit \']
114 $pragmachar = [$small $large $digit]
116 $docsym = [\| \^ \* \$]
118 @varid = $small $idchar*
119 @conid = $large $idchar*
121 @varsym = $symbol $symchar*
122 @consym = \: $symchar*
124 @decimal = $decdigit+
126 @hexadecimal = $hexit+
127 @exponent = [eE] [\-\+]? @decimal
129 -- we support the hierarchical module name extension:
132 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
134 -- normal signed numerical literals can only be explicitly negative,
135 -- not explicitly positive (contrast @exponent)
137 @signed = @negative ?
141 -- everywhere: skip whitespace and comments
143 $tab+ { warn Opt_WarnTabs (text "Tab character") }
145 -- Everywhere: deal with nested comments. We explicitly rule out
146 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
147 -- (this can happen even though pragmas will normally take precedence due to
148 -- longest-match, because pragmas aren't valid in every state, but comments
149 -- are). We also rule out nested Haddock comments, if the -haddock flag is
152 "{-" / { isNormalComment } { nested_comment lexToken }
154 -- Single-line comments are a bit tricky. Haskell 98 says that two or
155 -- more dashes followed by a symbol should be parsed as a varsym, so we
156 -- have to exclude those.
158 -- Since Haddock comments aren't valid in every state, we need to rule them
161 -- The following two rules match comments that begin with two dashes, but
162 -- continue with a different character. The rules test that this character
163 -- is not a symbol (in which case we'd have a varsym), and that it's not a
164 -- space followed by a Haddock comment symbol (docsym) (in which case we'd
165 -- have a Haddock comment). The rules then munch the rest of the line.
167 "-- " ~[$docsym \#] .* { lineCommentToken }
168 "--" [^$symbol : \ ] .* { lineCommentToken }
170 -- Next, match Haddock comments if no -haddock flag
172 "-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
174 -- Now, when we've matched comments that begin with 2 dashes and continue
175 -- with a different character, we need to match comments that begin with three
176 -- or more dashes (which clearly can't be Haddock comments). We only need to
177 -- make sure that the first non-dash character isn't a symbol, and munch the
180 "---"\-* [^$symbol :] .* { lineCommentToken }
182 -- Since the previous rules all match dashes followed by at least one
183 -- character, we also need to match a whole line filled with just dashes.
185 "--"\-* / { atEOL } { lineCommentToken }
187 -- We need this rule since none of the other single line comment rules
188 -- actually match this case.
190 "-- " / { atEOL } { lineCommentToken }
192 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
193 -- blank lines) until we find a non-whitespace character, then do layout
196 -- One slight wibble here: what if the line begins with {-#? In
197 -- theory, we have to lex the pragma to see if it's one we recognise,
198 -- and if it is, then we backtrack and do_bol, otherwise we treat it
199 -- as a nested comment. We don't bother with this: if the line begins
200 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
203 ^\# (line)? { begin line_prag1 }
204 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
205 ^\# \! .* \n ; -- #!, for scripts
209 -- after a layout keyword (let, where, do, of), we begin a new layout
210 -- context if the curly brace is missing.
211 -- Careful! This stuff is quite delicate.
212 <layout, layout_do> {
213 \{ / { notFollowedBy '-' } { pop_and open_brace }
214 -- we might encounter {-# here, but {- has been handled already
216 ^\# (line)? { begin line_prag1 }
219 -- do is treated in a subtly different way, see new_layout_context
220 <layout> () { new_layout_context True }
221 <layout_do> () { new_layout_context False }
223 -- after a new layout context which was found to be to the left of the
224 -- previous context, we have generated a '{' token, and we now need to
225 -- generate a matching '}' token.
226 <layout_left> () { do_layout_left }
228 <0,option_prags> \n { begin bol }
230 "{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
231 { dispatch_pragmas linePrags }
233 -- single-line line pragmas, of the form
234 -- # <line> "<file>" <extra-stuff> \n
235 <line_prag1> $decdigit+ { setLine line_prag1a }
236 <line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
237 <line_prag1b> .* { pop }
239 -- Haskell-style line pragmas, of the form
240 -- {-# LINE <line> "<file>" #-}
241 <line_prag2> $decdigit+ { setLine line_prag2a }
242 <line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
243 <line_prag2b> "#-}"|"-}" { pop }
244 -- NOTE: accept -} at the end of a LINE pragma, for compatibility
245 -- with older versions of GHC which generated these.
248 "{-#" $whitechar* $pragmachar+
249 $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
250 { dispatch_pragmas twoWordPrags }
252 "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
253 { dispatch_pragmas oneWordPrags }
255 -- We ignore all these pragmas, but don't generate a warning for them
256 "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
257 { dispatch_pragmas ignoredPrags }
259 -- ToDo: should only be valid inside a pragma:
264 "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
265 { dispatch_pragmas fileHeaderPrags }
267 "-- #" { multiline_doc_comment }
271 -- In the "0" mode we ignore these pragmas
272 "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
273 { nested_comment lexToken }
277 "-- #" .* { lineCommentToken }
281 "{-#" { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
282 (nested_comment lexToken) }
285 -- '0' state: ordinary lexemes
290 "-- " $docsym / { ifExtension haddockEnabled } { multiline_doc_comment }
291 "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
297 "[:" / { ifExtension parrEnabled } { token ITopabrack }
298 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
302 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
303 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
304 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
305 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
306 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
307 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
308 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
309 "$(" / { ifExtension thEnabled } { token ITparenEscape }
311 "[" @varid "|" / { ifExtension qqEnabled }
312 { lex_quasiquote_tok }
316 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
317 { special IToparenbar }
318 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
322 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
326 "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
327 { token IToubxparen }
328 "#)" / { ifExtension unboxedTuplesEnabled }
329 { token ITcubxparen }
333 "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
334 "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
338 \( { special IToparen }
339 \) { special ITcparen }
340 \[ { special ITobrack }
341 \] { special ITcbrack }
342 \, { special ITcomma }
343 \; { special ITsemi }
344 \` { special ITbackquote }
351 @qual @varid { idtoken qvarid }
352 @qual @conid { idtoken qconid }
354 @conid { idtoken conid }
358 @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
359 @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
360 @varid "#"+ / { ifExtension magicHashEnabled } { varid }
361 @conid "#"+ / { ifExtension magicHashEnabled } { idtoken conid }
364 -- ToDo: - move `var` and (sym) into lexical syntax?
365 -- - remove backquote from $special?
367 @qual @varsym / { ifExtension oldQualOps } { idtoken qvarsym }
368 @qual @consym / { ifExtension oldQualOps } { idtoken qconsym }
369 @qual \( @varsym \) / { ifExtension newQualOps } { idtoken prefixqvarsym }
370 @qual \( @consym \) / { ifExtension newQualOps } { idtoken prefixqconsym }
375 -- For the normal boxed literals we need to be careful
376 -- when trying to be close to Haskell98
378 -- Normal integral literals (:: Num a => a, from Integer)
379 @decimal { tok_num positive 0 0 decimal }
380 0[oO] @octal { tok_num positive 2 2 octal }
381 0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal }
383 -- Normal rational literals (:: Fractional a => a, from Rational)
384 @floating_point { strtoken tok_float }
388 -- Unboxed ints (:: Int#) and words (:: Word#)
389 -- It's simpler (and faster?) to give separate cases to the negatives,
390 -- especially considering octal/hexadecimal prefixes.
391 @decimal \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
392 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
393 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
394 @negative @decimal \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
395 @negative 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
396 @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
398 @decimal \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
399 0[oO] @octal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
400 0[xX] @hexadecimal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }
402 -- Unboxed floats and doubles (:: Float#, :: Double#)
403 -- prim_{float,double} work with signed literals
404 @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
405 @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
408 -- Strings and chars are lexed by hand-written code. The reason is
409 -- that even if we recognise the string or char here in the regex
410 -- lexer, we would still have to parse the string afterward in order
411 -- to convert it to a String.
414 \" { lex_string_tok }
418 -- -----------------------------------------------------------------------------
422 = ITas -- Haskell keywords
446 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
448 | ITforall -- GHC extension keywords
466 | ITinline_prag InlineSpec RuleMatchInfo
467 | ITspec_prag -- SPECIALISE
468 | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE)
476 | ITcore_prag -- hdaume: core annotations
480 | IToptions_prag String
481 | ITinclude_prag String
484 | ITdotdot -- reserved symbols
500 | ITbiglam -- GHC-extension symbols
502 | ITocurly -- special symbols
504 | ITocurlybar -- {|, for type applications
505 | ITccurlybar -- |}, for type applications
509 | ITopabrack -- [:, for parallel arrays with -XParr
510 | ITcpabrack -- :], for parallel arrays with -XParr
521 | ITvarid FastString -- identifiers
523 | ITvarsym FastString
524 | ITconsym FastString
525 | ITqvarid (FastString,FastString)
526 | ITqconid (FastString,FastString)
527 | ITqvarsym (FastString,FastString)
528 | ITqconsym (FastString,FastString)
529 | ITprefixqvarsym (FastString,FastString)
530 | ITprefixqconsym (FastString,FastString)
532 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
535 | ITstring FastString
537 | ITrational Rational
540 | ITprimstring FastString
543 | ITprimfloat Rational
544 | ITprimdouble Rational
546 -- Template Haskell extension tokens
547 | ITopenExpQuote -- [| or [e|
548 | ITopenPatQuote -- [p|
549 | ITopenDecQuote -- [d|
550 | ITopenTypQuote -- [t|
552 | ITidEscape FastString -- $x
553 | ITparenEscape -- $(
556 | ITquasiQuote (FastString,FastString,SrcSpan) -- [:...|...|]
558 -- Arrow notation extension
565 | ITLarrowtail -- -<<
566 | ITRarrowtail -- >>-
568 | ITunknown String -- Used when the lexer can't make sense of it
569 | ITeof -- end of file token
571 -- Documentation annotations
572 | ITdocCommentNext String -- something beginning '-- |'
573 | ITdocCommentPrev String -- something beginning '-- ^'
574 | ITdocCommentNamed String -- something beginning '-- $'
575 | ITdocSection Int String -- a section heading
576 | ITdocOptions String -- doc options (prune, ignore-exports, etc)
577 | ITdocOptionsOld String -- doc options declared "-- # ..."-style
578 | ITlineComment String -- comment starting by "--"
579 | ITblockComment String -- comment in {- -}
582 deriving Show -- debugging
586 isSpecial :: Token -> Bool
587 -- If we see M.x, where x is a keyword, but
588 -- is special, we treat is as just plain M.x,
590 isSpecial ITas = True
591 isSpecial IThiding = True
592 isSpecial ITqualified = True
593 isSpecial ITforall = True
594 isSpecial ITexport = True
595 isSpecial ITlabel = True
596 isSpecial ITdynamic = True
597 isSpecial ITsafe = True
598 isSpecial ITthreadsafe = True
599 isSpecial ITunsafe = True
600 isSpecial ITccallconv = True
601 isSpecial ITstdcallconv = True
602 isSpecial ITprimcallconv = True
603 isSpecial ITmdo = True
604 isSpecial ITfamily = True
605 isSpecial ITgroup = True
606 isSpecial ITby = True
607 isSpecial ITusing = True
611 -- the bitmap provided as the third component indicates whether the
612 -- corresponding extension keyword is valid under the extension options
613 -- provided to the compiler; if the extension corresponding to *any* of the
614 -- bits set in the bitmap is enabled, the keyword is valid (this setup
615 -- facilitates using a keyword in two different extensions that can be
616 -- activated independently)
618 reservedWordsFM :: UniqFM (Token, Int)
619 reservedWordsFM = listToUFM $
620 map (\(x, y, z) -> (mkFastString x, (y, z)))
621 [( "_", ITunderscore, 0 ),
623 ( "case", ITcase, 0 ),
624 ( "class", ITclass, 0 ),
625 ( "data", ITdata, 0 ),
626 ( "default", ITdefault, 0 ),
627 ( "deriving", ITderiving, 0 ),
629 ( "else", ITelse, 0 ),
630 ( "hiding", IThiding, 0 ),
632 ( "import", ITimport, 0 ),
634 ( "infix", ITinfix, 0 ),
635 ( "infixl", ITinfixl, 0 ),
636 ( "infixr", ITinfixr, 0 ),
637 ( "instance", ITinstance, 0 ),
639 ( "module", ITmodule, 0 ),
640 ( "newtype", ITnewtype, 0 ),
642 ( "qualified", ITqualified, 0 ),
643 ( "then", ITthen, 0 ),
644 ( "type", ITtype, 0 ),
645 ( "where", ITwhere, 0 ),
646 ( "_scc_", ITscc, 0 ), -- ToDo: remove
648 ( "forall", ITforall, bit explicitForallBit .|. bit inRulePragBit),
649 ( "mdo", ITmdo, bit recursiveDoBit),
650 ( "family", ITfamily, bit tyFamBit),
651 ( "group", ITgroup, bit transformComprehensionsBit),
652 ( "by", ITby, bit transformComprehensionsBit),
653 ( "using", ITusing, bit transformComprehensionsBit),
655 ( "foreign", ITforeign, bit ffiBit),
656 ( "export", ITexport, bit ffiBit),
657 ( "label", ITlabel, bit ffiBit),
658 ( "dynamic", ITdynamic, bit ffiBit),
659 ( "safe", ITsafe, bit ffiBit),
660 ( "threadsafe", ITthreadsafe, bit ffiBit), -- ToDo: remove
661 ( "unsafe", ITunsafe, bit ffiBit),
662 ( "stdcall", ITstdcallconv, bit ffiBit),
663 ( "ccall", ITccallconv, bit ffiBit),
664 ( "prim", ITprimcallconv, bit ffiBit),
666 ( "rec", ITrec, bit recBit),
667 ( "proc", ITproc, bit arrowsBit)
670 reservedSymsFM :: UniqFM (Token, Int -> Bool)
671 reservedSymsFM = listToUFM $
672 map (\ (x,y,z) -> (mkFastString x,(y,z)))
673 [ ("..", ITdotdot, always)
674 -- (:) is a reserved op, meaning only list cons
675 ,(":", ITcolon, always)
676 ,("::", ITdcolon, always)
677 ,("=", ITequal, always)
678 ,("\\", ITlam, always)
679 ,("|", ITvbar, always)
680 ,("<-", ITlarrow, always)
681 ,("->", ITrarrow, always)
683 ,("~", ITtilde, always)
684 ,("=>", ITdarrow, always)
685 ,("-", ITminus, always)
686 ,("!", ITbang, always)
688 -- For data T (a::*) = MkT
689 ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
690 -- For 'forall a . t'
691 ,(".", ITdot, always) -- \i -> explicitForallEnabled i || inRulePrag i)
693 ,("-<", ITlarrowtail, arrowsEnabled)
694 ,(">-", ITrarrowtail, arrowsEnabled)
695 ,("-<<", ITLarrowtail, arrowsEnabled)
696 ,(">>-", ITRarrowtail, arrowsEnabled)
698 ,("∷", ITdcolon, unicodeSyntaxEnabled)
699 ,("⇒", ITdarrow, unicodeSyntaxEnabled)
700 ,("∀", ITforall, \i -> unicodeSyntaxEnabled i &&
701 explicitForallEnabled i)
702 ,("→", ITrarrow, unicodeSyntaxEnabled)
703 ,("←", ITlarrow, unicodeSyntaxEnabled)
705 ,("⤙", ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
706 ,("⤚", ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
707 ,("⤛", ITLarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
708 ,("⤜", ITRarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
710 ,("★", ITstar, unicodeSyntaxEnabled)
712 -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
713 -- form part of a large operator. This would let us have a better
714 -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
717 -- -----------------------------------------------------------------------------
720 type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
722 special :: Token -> Action
723 special tok span _buf _len = return (L span tok)
725 token, layout_token :: Token -> Action
726 token t span _buf _len = return (L span t)
727 layout_token t span _buf _len = pushLexState layout >> return (L span t)
729 idtoken :: (StringBuffer -> Int -> Token) -> Action
730 idtoken f span buf len = return (L span $! (f buf len))
732 skip_one_varid :: (FastString -> Token) -> Action
733 skip_one_varid f span buf len
734 = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
736 strtoken :: (String -> Token) -> Action
737 strtoken f span buf len =
738 return (L span $! (f $! lexemeToString buf len))
740 init_strtoken :: Int -> (String -> Token) -> Action
741 -- like strtoken, but drops the last N character(s)
742 init_strtoken drop f span buf len =
743 return (L span $! (f $! lexemeToString buf (len-drop)))
745 begin :: Int -> Action
746 begin code _span _str _len = do pushLexState code; lexToken
749 pop _span _buf _len = do _ <- popLexState
752 pop_and :: Action -> Action
753 pop_and act span buf len = do _ <- popLexState
756 {-# INLINE nextCharIs #-}
757 nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
758 nextCharIs buf p = not (atEnd buf) && p (currentChar buf)
760 notFollowedBy :: Char -> AlexAccPred Int
761 notFollowedBy char _ _ _ (AI _ buf)
762 = nextCharIs buf (/=char)
764 notFollowedBySymbol :: AlexAccPred Int
765 notFollowedBySymbol _ _ _ (AI _ buf)
766 = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
768 -- We must reject doc comments as being ordinary comments everywhere.
769 -- In some cases the doc comment will be selected as the lexeme due to
770 -- maximal munch, but not always, because the nested comment rule is
771 -- valid in all states, but the doc-comment rules are only valid in
772 -- the non-layout states.
773 isNormalComment :: AlexAccPred Int
774 isNormalComment bits _ _ (AI _ buf)
775 | haddockEnabled bits = notFollowedByDocOrPragma
776 | otherwise = nextCharIs buf (/='#')
778 notFollowedByDocOrPragma
779 = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))
781 spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
782 spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
785 haddockDisabledAnd p bits _ _ (AI _ buf)
786 = if haddockEnabled bits then False else (p buf)
789 atEOL :: AlexAccPred Int
790 atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
792 ifExtension :: (Int -> Bool) -> AlexAccPred Int
793 ifExtension pred bits _ _ _ = pred bits
795 multiline_doc_comment :: Action
796 multiline_doc_comment span buf _len = withLexedDocType (worker "")
798 worker commentAcc input docType oneLine = case alexGetChar input of
800 | oneLine -> docCommentEnd input commentAcc docType buf span
801 | otherwise -> case checkIfCommentLine input' of
802 Just input -> worker ('\n':commentAcc) input docType False
803 Nothing -> docCommentEnd input commentAcc docType buf span
804 Just (c, input) -> worker (c:commentAcc) input docType oneLine
805 Nothing -> docCommentEnd input commentAcc docType buf span
807 checkIfCommentLine input = check (dropNonNewlineSpace input)
809 check input = case alexGetChar input of
810 Just ('-', input) -> case alexGetChar input of
811 Just ('-', input) -> case alexGetChar input of
812 Just (c, _) | c /= '-' -> Just input
817 dropNonNewlineSpace input = case alexGetChar input of
819 | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
823 lineCommentToken :: Action
824 lineCommentToken span buf len = do
825 b <- extension rawTokenStreamEnabled
826 if b then strtoken ITlineComment span buf len else lexToken
829 nested comments require traversing by hand, they can't be parsed
830 using regular expressions.
832 nested_comment :: P (Located Token) -> Action
833 nested_comment cont span _str _len = do
837 go commentAcc 0 input = do setInput input
838 b <- extension rawTokenStreamEnabled
840 then docCommentEnd input commentAcc ITblockComment _str span
842 go commentAcc n input = case alexGetChar input of
843 Nothing -> errBrace input span
844 Just ('-',input) -> case alexGetChar input of
845 Nothing -> errBrace input span
846 Just ('\125',input) -> go commentAcc (n-1) input
847 Just (_,_) -> go ('-':commentAcc) n input
848 Just ('\123',input) -> case alexGetChar input of
849 Nothing -> errBrace input span
850 Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
851 Just (_,_) -> go ('\123':commentAcc) n input
852 Just (c,input) -> go (c:commentAcc) n input
854 nested_doc_comment :: Action
855 nested_doc_comment span buf _len = withLexedDocType (go "")
857 go commentAcc input docType _ = case alexGetChar input of
858 Nothing -> errBrace input span
859 Just ('-',input) -> case alexGetChar input of
860 Nothing -> errBrace input span
861 Just ('\125',input) ->
862 docCommentEnd input commentAcc docType buf span
863 Just (_,_) -> go ('-':commentAcc) input docType False
864 Just ('\123', input) -> case alexGetChar input of
865 Nothing -> errBrace input span
866 Just ('-',input) -> do
868 let cont = do input <- getInput; go commentAcc input docType False
869 nested_comment cont span buf _len
870 Just (_,_) -> go ('\123':commentAcc) input docType False
871 Just (c,input) -> go (c:commentAcc) input docType False
873 withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
875 withLexedDocType lexDocComment = do
876 input@(AI _ buf) <- getInput
877 case prevChar buf ' ' of
878 '|' -> lexDocComment input ITdocCommentNext False
879 '^' -> lexDocComment input ITdocCommentPrev False
880 '$' -> lexDocComment input ITdocCommentNamed False
881 '*' -> lexDocSection 1 input
882 '#' -> lexDocComment input ITdocOptionsOld False
883 _ -> panic "withLexedDocType: Bad doc type"
885 lexDocSection n input = case alexGetChar input of
886 Just ('*', input) -> lexDocSection (n+1) input
887 Just (_, _) -> lexDocComment input (ITdocSection n) True
888 Nothing -> do setInput input; lexToken -- eof reached, lex it normally
890 -- RULES pragmas turn on the forall and '.' keywords, and we turn them
891 -- off again at the end of the pragma.
893 rulePrag span _buf _len = do
894 setExts (.|. bit inRulePragBit)
895 return (L span ITrules_prag)
898 endPrag span _buf _len = do
899 setExts (.&. complement (bit inRulePragBit))
900 return (L span ITclose_prag)
903 -------------------------------------------------------------------------------
904 -- This function is quite tricky. We can't just return a new token, we also
905 -- need to update the state of the parser. Why? Because the token is longer
906 -- than what was lexed by Alex, and the lexToken function doesn't know this, so
907 -- it writes the wrong token length to the parser state. This function is
908 -- called afterwards, so it can just update the state.
910 docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
911 SrcSpan -> P (Located Token)
912 docCommentEnd input commentAcc docType buf span = do
914 let (AI loc nextBuf) = input
915 comment = reverse commentAcc
916 span' = mkSrcSpan (srcSpanStart span) loc
917 last_len = byteDiff buf nextBuf
919 span `seq` setLastToken span' last_len
920 return (L span' (docType comment))
922 errBrace :: AlexInput -> SrcSpan -> P a
923 errBrace (AI end _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
925 open_brace, close_brace :: Action
926 open_brace span _str _len = do
928 setContext (NoLayout:ctx)
929 return (L span ITocurly)
930 close_brace span _str _len = do
932 return (L span ITccurly)
934 qvarid, qconid :: StringBuffer -> Int -> Token
935 qvarid buf len = ITqvarid $! splitQualName buf len False
936 qconid buf len = ITqconid $! splitQualName buf len False
938 splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
939 -- takes a StringBuffer and a length, and returns the module name
940 -- and identifier parts of a qualified name. Splits at the *last* dot,
941 -- because of hierarchical module names.
942 splitQualName orig_buf len parens = split orig_buf orig_buf
945 | orig_buf `byteDiff` buf >= len = done dot_buf
946 | c == '.' = found_dot buf'
947 | otherwise = split buf' dot_buf
949 (c,buf') = nextChar buf
951 -- careful, we might get names like M....
952 -- so, if the character after the dot is not upper-case, this is
953 -- the end of the qualifier part.
954 found_dot buf -- buf points after the '.'
955 | isUpper c = split buf' buf
956 | otherwise = done buf
958 (c,buf') = nextChar buf
961 (lexemeToFastString orig_buf (qual_size - 1),
962 if parens -- Prelude.(+)
963 then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
964 else lexemeToFastString dot_buf (len - qual_size))
966 qual_size = orig_buf `byteDiff` dot_buf
971 case lookupUFM reservedWordsFM fs of
972 Just (keyword,0) -> do
974 return (L span keyword)
975 Just (keyword,exts) -> do
976 b <- extension (\i -> exts .&. i /= 0)
977 if b then do maybe_layout keyword
978 return (L span keyword)
979 else return (L span (ITvarid fs))
980 _other -> return (L span (ITvarid fs))
982 fs = lexemeToFastString buf len
984 conid :: StringBuffer -> Int -> Token
985 conid buf len = ITconid fs
986 where fs = lexemeToFastString buf len
988 qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
989 qvarsym buf len = ITqvarsym $! splitQualName buf len False
990 qconsym buf len = ITqconsym $! splitQualName buf len False
991 prefixqvarsym buf len = ITprefixqvarsym $! splitQualName buf len True
992 prefixqconsym buf len = ITprefixqconsym $! splitQualName buf len True
994 varsym, consym :: Action
995 varsym = sym ITvarsym
996 consym = sym ITconsym
998 sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
1000 sym con span buf len =
1001 case lookupUFM reservedSymsFM fs of
1002 Just (keyword,exts) -> do
1004 if b then return (L span keyword)
1005 else return (L span $! con fs)
1006 _other -> return (L span $! con fs)
1008 fs = lexemeToFastString buf len
1010 -- Variations on the integral numeric literal.
1011 tok_integral :: (Integer -> Token)
1012 -> (Integer -> Integer)
1013 -- -> (StringBuffer -> StringBuffer) -> (Int -> Int)
1015 -> (Integer, (Char->Int)) -> Action
1016 tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
1017 return $ L span $ itint $! transint $ parseUnsignedInteger
1018 (offsetBytes transbuf buf) (subtract translen len) radix char_to_int
1020 -- some conveniences for use with tok_integral
1021 tok_num :: (Integer -> Integer)
1023 -> (Integer, (Char->Int)) -> Action
1024 tok_num = tok_integral ITinteger
1025 tok_primint :: (Integer -> Integer)
1027 -> (Integer, (Char->Int)) -> Action
1028 tok_primint = tok_integral ITprimint
1029 tok_primword :: Int -> Int
1030 -> (Integer, (Char->Int)) -> Action
1031 tok_primword = tok_integral ITprimword positive
1032 positive, negative :: (Integer -> Integer)
1035 decimal, octal, hexadecimal :: (Integer, Char -> Int)
1036 decimal = (10,octDecDigit)
1037 octal = (8,octDecDigit)
1038 hexadecimal = (16,hexDigit)
1040 -- readRational can understand negative rationals, exponents, everything.
1041 tok_float, tok_primfloat, tok_primdouble :: String -> Token
1042 tok_float str = ITrational $! readRational str
1043 tok_primfloat str = ITprimfloat $! readRational str
1044 tok_primdouble str = ITprimdouble $! readRational str
1046 -- -----------------------------------------------------------------------------
1047 -- Layout processing
1049 -- we're at the first token on a line, insert layout tokens if necessary
1051 do_bol span _str _len = do
1055 --trace "layout: inserting '}'" $ do
1057 -- do NOT pop the lex state, we might have a ';' to insert
1058 return (L span ITvccurly)
1060 --trace "layout: inserting ';'" $ do
1062 return (L span ITsemi)
1067 -- certain keywords put us in the "layout" state, where we might
1068 -- add an opening curly brace.
1069 maybe_layout :: Token -> P ()
1070 maybe_layout t = do -- If the alternative layout rule is enabled then
1071 -- we never create an implicit layout context here.
1072 -- Layout is handled XXX instead.
1073 -- The code for closing implicit contexts, or
1074 -- inserting implicit semi-colons, is therefore
1075 -- irrelevant as it only applies in an implicit
1077 alr <- extension alternativeLayoutRule
1079 where f ITdo = pushLexState layout_do
1080 f ITmdo = pushLexState layout_do
1081 f ITof = pushLexState layout
1082 f ITlet = pushLexState layout
1083 f ITwhere = pushLexState layout
1084 f ITrec = pushLexState layout
1087 -- Pushing a new implicit layout context. If the indentation of the
1088 -- next token is not greater than the previous layout context, then
1089 -- Haskell 98 says that the new layout context should be empty; that is
1090 -- the lexer must generate {}.
1092 -- We are slightly more lenient than this: when the new context is started
1093 -- by a 'do', then we allow the new context to be at the same indentation as
1094 -- the previous context. This is what the 'strict' argument is for.
1096 new_layout_context :: Bool -> Action
1097 new_layout_context strict span _buf _len = do
1099 (AI l _) <- getInput
1100 let offset = srcLocCol l
1103 Layout prev_off : _ |
1104 (strict && prev_off >= offset ||
1105 not strict && prev_off > offset) -> do
1106 -- token is indented to the left of the previous context.
1107 -- we must generate a {} sequence now.
1108 pushLexState layout_left
1109 return (L span ITvocurly)
1111 setContext (Layout offset : ctx)
1112 return (L span ITvocurly)
1114 do_layout_left :: Action
1115 do_layout_left span _buf _len = do
1117 pushLexState bol -- we must be at the start of a line
1118 return (L span ITvccurly)
1120 -- -----------------------------------------------------------------------------
1123 setLine :: Int -> Action
1124 setLine code span buf len = do
1125 let line = parseUnsignedInteger buf len 10 octDecDigit
1126 setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 1)
1127 -- subtract one: the line number refers to the *following* line
1132 setFile :: Int -> Action
1133 setFile code span buf len = do
1134 let file = lexemeToFastString (stepOn buf) (len-2)
1135 setAlrLastLoc noSrcSpan
1136 setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1142 -- -----------------------------------------------------------------------------
1143 -- Options, includes and language pragmas.
1145 lex_string_prag :: (String -> Token) -> Action
1146 lex_string_prag mkTok span _buf _len
1147 = do input <- getInput
1151 return (L (mkSrcSpan start end) tok)
1153 = if isString input "#-}"
1154 then do setInput input
1155 return (mkTok (reverse acc))
1156 else case alexGetChar input of
1157 Just (c,i) -> go (c:acc) i
1158 Nothing -> err input
1159 isString _ [] = True
1161 = case alexGetChar i of
1162 Just (c,i') | c == x -> isString i' xs
1164 err (AI end _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"
1167 -- -----------------------------------------------------------------------------
1170 -- This stuff is horrible. I hates it.
1172 lex_string_tok :: Action
1173 lex_string_tok span _buf _len = do
1174 tok <- lex_string ""
1176 return (L (mkSrcSpan (srcSpanStart span) end) tok)
1178 lex_string :: String -> P Token
1181 case alexGetChar' i of
1182 Nothing -> lit_error i
1186 magicHash <- extension magicHashEnabled
1190 case alexGetChar' i of
1194 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1195 else let s' = mkZFastString (reverse s) in
1196 return (ITprimstring s')
1197 -- mkZFastString is a hack to avoid encoding the
1198 -- string in UTF-8. We just want the exact bytes.
1200 return (ITstring (mkFastString (reverse s)))
1202 return (ITstring (mkFastString (reverse s)))
1205 | Just ('&',i) <- next -> do
1206 setInput i; lex_string s
1207 | Just (c,i) <- next, c <= '\x7f' && is_space c -> do
1208 -- is_space only works for <= '\x7f' (#3751)
1209 setInput i; lex_stringgap s
1210 where next = alexGetChar' i
1214 '\\' -> do setInput i1; c' <- lex_escape; lex_string (c':s)
1215 c | isAny c -> do setInput i1; lex_string (c:s)
1216 _other -> lit_error i
1218 lex_stringgap :: String -> P Token
1219 lex_stringgap s = do
1221 c <- getCharOrFail i
1223 '\\' -> lex_string s
1224 c | is_space c -> lex_stringgap s
1225 _other -> lit_error i
1228 lex_char_tok :: Action
1229 -- Here we are basically parsing character literals, such as 'x' or '\n'
1230 -- but, when Template Haskell is on, we additionally spot
1231 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
1232 -- but WITHOUT CONSUMING the x or T part (the parser does that).
1233 -- So we have to do two characters of lookahead: when we see 'x we need to
1234 -- see if there's a trailing quote
1235 lex_char_tok span _buf _len = do -- We've seen '
1236 i1 <- getInput -- Look ahead to first character
1237 let loc = srcSpanStart span
1238 case alexGetChar' i1 of
1239 Nothing -> lit_error i1
1241 Just ('\'', i2@(AI end2 _)) -> do -- We've seen ''
1242 th_exts <- extension thEnabled
1245 return (L (mkSrcSpan loc end2) ITtyQuote)
1248 Just ('\\', i2@(AI _end2 _)) -> do -- We've seen 'backslash
1250 lit_ch <- lex_escape
1252 mc <- getCharOrFail i3 -- Trailing quote
1253 if mc == '\'' then finish_char_tok loc lit_ch
1256 Just (c, i2@(AI _end2 _))
1257 | not (isAny c) -> lit_error i1
1260 -- We've seen 'x, where x is a valid character
1261 -- (i.e. not newline etc) but not a quote or backslash
1262 case alexGetChar' i2 of -- Look ahead one more character
1263 Just ('\'', i3) -> do -- We've seen 'x'
1265 finish_char_tok loc c
1266 _other -> do -- We've seen 'x not followed by quote
1267 -- (including the possibility of EOF)
1268 -- If TH is on, just parse the quote only
1269 th_exts <- extension thEnabled
1271 if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1274 finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1275 finish_char_tok loc ch -- We've already seen the closing quote
1276 -- Just need to check for trailing #
1277 = do magicHash <- extension magicHashEnabled
1278 i@(AI end _) <- getInput
1279 if magicHash then do
1280 case alexGetChar' i of
1281 Just ('#',i@(AI end _)) -> do
1283 return (L (mkSrcSpan loc end) (ITprimchar ch))
1285 return (L (mkSrcSpan loc end) (ITchar ch))
1287 return (L (mkSrcSpan loc end) (ITchar ch))
1289 isAny :: Char -> Bool
1290 isAny c | c > '\x7f' = isPrint c
1291 | otherwise = is_any c
1293 lex_escape :: P Char
1296 c <- getCharOrFail i0
1308 '^' -> do i1 <- getInput
1309 c <- getCharOrFail i1
1310 if c >= '@' && c <= '_'
1311 then return (chr (ord c - ord '@'))
1314 'x' -> readNum is_hexdigit 16 hexDigit
1315 'o' -> readNum is_octdigit 8 octDecDigit
1316 x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x)
1320 case alexGetChar' i of
1321 Nothing -> lit_error i0
1323 case alexGetChar' i2 of
1324 Nothing -> do lit_error i0
1326 let str = [c1,c2,c3] in
1327 case [ (c,rest) | (p,c) <- silly_escape_chars,
1328 Just rest <- [stripPrefix p str] ] of
1329 (escape_char,[]):_ -> do
1332 (escape_char,_:_):_ -> do
1337 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
1338 readNum is_digit base conv = do
1340 c <- getCharOrFail i
1342 then readNum2 is_digit base conv (conv c)
1345 readNum2 :: (Char -> Bool) -> Int -> (Char -> Int) -> Int -> P Char
1346 readNum2 is_digit base conv i = do
1349 where read i input = do
1350 case alexGetChar' input of
1351 Just (c,input') | is_digit c -> do
1352 let i' = i*base + conv c
1354 then setInput input >> lexError "numeric escape sequence out of range"
1357 setInput input; return (chr i)
1360 silly_escape_chars :: [(String, Char)]
1361 silly_escape_chars = [
1398 -- before calling lit_error, ensure that the current input is pointing to
1399 -- the position of the error in the buffer. This is so that we can report
1400 -- a correct location to the user, but also so we can detect UTF-8 decoding
1401 -- errors if they occur.
1402 lit_error :: AlexInput -> P a
1403 lit_error i = do setInput i; lexError "lexical error in string/character literal"
1405 getCharOrFail :: AlexInput -> P Char
1406 getCharOrFail i = do
1407 case alexGetChar' i of
1408 Nothing -> lexError "unexpected end-of-file in string/character literal"
1409 Just (c,i) -> do setInput i; return c
1411 -- -----------------------------------------------------------------------------
1414 lex_quasiquote_tok :: Action
1415 lex_quasiquote_tok span buf len = do
1416 let quoter = tail (lexemeToString buf (len - 1))
1417 -- 'tail' drops the initial '[',
1418 -- while the -1 drops the trailing '|'
1419 quoteStart <- getSrcLoc
1420 quote <- lex_quasiquote ""
1422 return (L (mkSrcSpan (srcSpanStart span) end)
1423 (ITquasiQuote (mkFastString quoter,
1424 mkFastString (reverse quote),
1425 mkSrcSpan quoteStart end)))
1427 lex_quasiquote :: String -> P String
1428 lex_quasiquote s = do
1430 case alexGetChar' i of
1431 Nothing -> lit_error i
1434 | Just ('|',i) <- next -> do
1435 setInput i; lex_quasiquote ('|' : s)
1436 | Just (']',i) <- next -> do
1437 setInput i; lex_quasiquote (']' : s)
1438 where next = alexGetChar' i
1441 | Just (']',i) <- next -> do
1442 setInput i; return s
1443 where next = alexGetChar' i
1446 setInput i; lex_quasiquote (c : s)
1448 -- -----------------------------------------------------------------------------
1451 warn :: DynFlag -> SDoc -> Action
1452 warn option warning srcspan _buf _len = do
1453 addWarning option srcspan warning
1456 warnThen :: DynFlag -> SDoc -> Action -> Action
1457 warnThen option warning action srcspan buf len = do
1458 addWarning option srcspan warning
1459 action srcspan buf len
1461 -- -----------------------------------------------------------------------------
1472 SrcSpan -- The start and end of the text span related to
1473 -- the error. Might be used in environments which can
1474 -- show this span, e.g. by highlighting it.
1475 Message -- The error message
1477 data PState = PState {
1478 buffer :: StringBuffer,
1480 messages :: Messages,
1481 last_loc :: SrcSpan, -- pos of previous token
1482 last_len :: !Int, -- len of previous token
1483 loc :: SrcLoc, -- current loc (end of prev token + 1)
1484 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1485 context :: [LayoutContext],
1487 -- Used in the alternative layout rule:
1488 -- These tokens are the next ones to be sent out. They are
1489 -- just blindly emitted, without the rule looking at them again:
1490 alr_pending_implicit_tokens :: [Located Token],
1491 -- This is the next token to be considered or, if it is Nothing,
1492 -- we need to get the next token from the input stream:
1493 alr_next_token :: Maybe (Located Token),
1494 -- This is what we consider to be the locatino of the last token
1496 alr_last_loc :: SrcSpan,
1497 -- The stack of layout contexts:
1498 alr_context :: [ALRContext],
1499 -- Are we expecting a '{'? If it's Just, then the ALRLayout tells
1500 -- us what sort of layout the '{' will open:
1501 alr_expecting_ocurly :: Maybe ALRLayout,
1502 -- Have we just had the '}' for a let block? If so, than an 'in'
1503 -- token doesn't need to close anything:
1504 alr_justClosedExplicitLetBlock :: Bool
1506 -- last_loc and last_len are used when generating error messages,
1507 -- and in pushCurrentContext only. Sigh, if only Happy passed the
1508 -- current token to happyError, we could at least get rid of last_len.
1509 -- Getting rid of last_loc would require finding another way to
1510 -- implement pushCurrentContext (which is only called from one place).
1512 data ALRContext = ALRNoLayout Bool{- does it contain commas? -}
1513 Bool{- is it a 'let' block? -}
1514 | ALRLayout ALRLayout Int
1515 data ALRLayout = ALRLayoutLet
1520 newtype P a = P { unP :: PState -> ParseResult a }
1522 instance Monad P where
1528 returnP a = a `seq` (P $ \s -> POk s a)
1530 thenP :: P a -> (a -> P b) -> P b
1531 (P m) `thenP` k = P $ \ s ->
1533 POk s1 a -> (unP (k a)) s1
1534 PFailed span err -> PFailed span err
1536 failP :: String -> P a
1537 failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1539 failMsgP :: String -> P a
1540 failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1542 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1543 failLocMsgP loc1 loc2 str = P $ \_ -> PFailed (mkSrcSpan loc1 loc2) (text str)
1545 failSpanMsgP :: SrcSpan -> SDoc -> P a
1546 failSpanMsgP span msg = P $ \_ -> PFailed span msg
1548 getPState :: P PState
1549 getPState = P $ \s -> POk s s
1551 getDynFlags :: P DynFlags
1552 getDynFlags = P $ \s -> POk s (dflags s)
1554 withThisPackage :: (PackageId -> a) -> P a
1556 = do pkg <- liftM thisPackage getDynFlags
1559 extension :: (Int -> Bool) -> P Bool
1560 extension p = P $ \s -> POk s (p $! extsBitmap s)
1563 getExts = P $ \s -> POk s (extsBitmap s)
1565 setExts :: (Int -> Int) -> P ()
1566 setExts f = P $ \s -> POk s{ extsBitmap = f (extsBitmap s) } ()
1568 setSrcLoc :: SrcLoc -> P ()
1569 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1571 getSrcLoc :: P SrcLoc
1572 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1574 setLastToken :: SrcSpan -> Int -> P ()
1575 setLastToken loc len = P $ \s -> POk s {
1580 data AlexInput = AI SrcLoc StringBuffer
1582 alexInputPrevChar :: AlexInput -> Char
1583 alexInputPrevChar (AI _ buf) = prevChar buf '\n'
1585 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1586 alexGetChar (AI loc s)
1588 | otherwise = adj_c `seq` loc' `seq` s' `seq`
1589 --trace (show (ord c)) $
1590 Just (adj_c, (AI loc' s'))
1591 where (c,s') = nextChar s
1592 loc' = advanceSrcLoc loc c
1600 other_graphic = '\x6'
1603 | c <= '\x06' = non_graphic
1605 -- Alex doesn't handle Unicode, so when Unicode
1606 -- character is encountered we output these values
1607 -- with the actual character value hidden in the state.
1609 case generalCategory c of
1610 UppercaseLetter -> upper
1611 LowercaseLetter -> lower
1612 TitlecaseLetter -> upper
1613 ModifierLetter -> other_graphic
1614 OtherLetter -> lower -- see #1103
1615 NonSpacingMark -> other_graphic
1616 SpacingCombiningMark -> other_graphic
1617 EnclosingMark -> other_graphic
1618 DecimalNumber -> digit
1619 LetterNumber -> other_graphic
1620 OtherNumber -> other_graphic
1621 ConnectorPunctuation -> symbol
1622 DashPunctuation -> symbol
1623 OpenPunctuation -> other_graphic
1624 ClosePunctuation -> other_graphic
1625 InitialQuote -> other_graphic
1626 FinalQuote -> other_graphic
1627 OtherPunctuation -> symbol
1628 MathSymbol -> symbol
1629 CurrencySymbol -> symbol
1630 ModifierSymbol -> symbol
1631 OtherSymbol -> symbol
1633 _other -> non_graphic
1635 -- This version does not squash unicode characters, it is used when
1637 alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
1638 alexGetChar' (AI loc s)
1640 | otherwise = c `seq` loc' `seq` s' `seq`
1641 --trace (show (ord c)) $
1642 Just (c, (AI loc' s'))
1643 where (c,s') = nextChar s
1644 loc' = advanceSrcLoc loc c
1646 getInput :: P AlexInput
1647 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (AI l b)
1649 setInput :: AlexInput -> P ()
1650 setInput (AI l b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1652 pushLexState :: Int -> P ()
1653 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1655 popLexState :: P Int
1656 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1658 getLexState :: P Int
1659 getLexState = P $ \s@PState{ lex_state=ls:_ } -> POk s ls
1661 popNextToken :: P (Maybe (Located Token))
1663 = P $ \s@PState{ alr_next_token = m } ->
1664 POk (s {alr_next_token = Nothing}) m
1666 setAlrLastLoc :: SrcSpan -> P ()
1667 setAlrLastLoc l = P $ \s -> POk (s {alr_last_loc = l}) ()
1669 getAlrLastLoc :: P SrcSpan
1670 getAlrLastLoc = P $ \s@(PState {alr_last_loc = l}) -> POk s l
1672 getALRContext :: P [ALRContext]
1673 getALRContext = P $ \s@(PState {alr_context = cs}) -> POk s cs
1675 setALRContext :: [ALRContext] -> P ()
1676 setALRContext cs = P $ \s -> POk (s {alr_context = cs}) ()
1678 getJustClosedExplicitLetBlock :: P Bool
1679 getJustClosedExplicitLetBlock
1680 = P $ \s@(PState {alr_justClosedExplicitLetBlock = b}) -> POk s b
1682 setJustClosedExplicitLetBlock :: Bool -> P ()
1683 setJustClosedExplicitLetBlock b
1684 = P $ \s -> POk (s {alr_justClosedExplicitLetBlock = b}) ()
1686 setNextToken :: Located Token -> P ()
1687 setNextToken t = P $ \s -> POk (s {alr_next_token = Just t}) ()
1689 popPendingImplicitToken :: P (Maybe (Located Token))
1690 popPendingImplicitToken
1691 = P $ \s@PState{ alr_pending_implicit_tokens = ts } ->
1694 (t : ts') -> POk (s {alr_pending_implicit_tokens = ts'}) (Just t)
1696 setPendingImplicitTokens :: [Located Token] -> P ()
1697 setPendingImplicitTokens ts = P $ \s -> POk (s {alr_pending_implicit_tokens = ts}) ()
1699 getAlrExpectingOCurly :: P (Maybe ALRLayout)
1700 getAlrExpectingOCurly = P $ \s@(PState {alr_expecting_ocurly = b}) -> POk s b
1702 setAlrExpectingOCurly :: Maybe ALRLayout -> P ()
1703 setAlrExpectingOCurly b = P $ \s -> POk (s {alr_expecting_ocurly = b}) ()
1705 -- for reasons of efficiency, flags indicating language extensions (eg,
1706 -- -fglasgow-exts or -XParr) are represented by a bitmap stored in an unboxed
1710 genericsBit = 0 -- {| and |}
1721 explicitForallBit :: Int
1722 explicitForallBit = 7 -- the 'forall' keyword and '.' symbol
1724 bangPatBit = 8 -- Tells the parser to understand bang-patterns
1725 -- (doesn't affect the lexer)
1727 tyFamBit = 9 -- indexed type families: 'family' keyword and kind sigs
1729 haddockBit = 10 -- Lex and parse Haddock comments
1731 magicHashBit = 11 -- "#" in both functions and operators
1733 kindSigsBit = 12 -- Kind signatures on type variables
1734 recursiveDoBit :: Int
1735 recursiveDoBit = 13 -- mdo
1736 unicodeSyntaxBit :: Int
1737 unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc
1738 unboxedTuplesBit :: Int
1739 unboxedTuplesBit = 15 -- (# and #)
1740 datatypeContextsBit :: Int
1741 datatypeContextsBit = 16
1742 transformComprehensionsBit :: Int
1743 transformComprehensionsBit = 17
1745 qqBit = 18 -- enable quasiquoting
1746 inRulePragBit :: Int
1748 rawTokenStreamBit :: Int
1749 rawTokenStreamBit = 20 -- producing a token stream with all comments included
1750 newQualOpsBit :: Int
1751 newQualOpsBit = 21 -- Haskell' qualified operator syntax, e.g. Prelude.(+)
1754 alternativeLayoutRuleBit :: Int
1755 alternativeLayoutRuleBit = 23
1757 always :: Int -> Bool
1759 genericsEnabled :: Int -> Bool
1760 genericsEnabled flags = testBit flags genericsBit
1761 parrEnabled :: Int -> Bool
1762 parrEnabled flags = testBit flags parrBit
1763 arrowsEnabled :: Int -> Bool
1764 arrowsEnabled flags = testBit flags arrowsBit
1765 thEnabled :: Int -> Bool
1766 thEnabled flags = testBit flags thBit
1767 ipEnabled :: Int -> Bool
1768 ipEnabled flags = testBit flags ipBit
1769 explicitForallEnabled :: Int -> Bool
1770 explicitForallEnabled flags = testBit flags explicitForallBit
1771 bangPatEnabled :: Int -> Bool
1772 bangPatEnabled flags = testBit flags bangPatBit
1773 -- tyFamEnabled :: Int -> Bool
1774 -- tyFamEnabled flags = testBit flags tyFamBit
1775 haddockEnabled :: Int -> Bool
1776 haddockEnabled flags = testBit flags haddockBit
1777 magicHashEnabled :: Int -> Bool
1778 magicHashEnabled flags = testBit flags magicHashBit
1779 -- kindSigsEnabled :: Int -> Bool
1780 -- kindSigsEnabled flags = testBit flags kindSigsBit
1781 unicodeSyntaxEnabled :: Int -> Bool
1782 unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit
1783 unboxedTuplesEnabled :: Int -> Bool
1784 unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit
1785 datatypeContextsEnabled :: Int -> Bool
1786 datatypeContextsEnabled flags = testBit flags datatypeContextsBit
1787 qqEnabled :: Int -> Bool
1788 qqEnabled flags = testBit flags qqBit
1789 -- inRulePrag :: Int -> Bool
1790 -- inRulePrag flags = testBit flags inRulePragBit
1791 rawTokenStreamEnabled :: Int -> Bool
1792 rawTokenStreamEnabled flags = testBit flags rawTokenStreamBit
1793 newQualOps :: Int -> Bool
1794 newQualOps flags = testBit flags newQualOpsBit
1795 oldQualOps :: Int -> Bool
1796 oldQualOps flags = not (newQualOps flags)
1797 alternativeLayoutRule :: Int -> Bool
1798 alternativeLayoutRule flags = testBit flags alternativeLayoutRuleBit
1800 -- PState for parsing options pragmas
1802 pragState :: DynFlags -> StringBuffer -> SrcLoc -> PState
1803 pragState dynflags buf loc = (mkPState dynflags buf loc) {
1804 lex_state = [bol, option_prags, 0]
1807 -- create a parse state
1809 mkPState :: DynFlags -> StringBuffer -> SrcLoc -> PState
1810 mkPState flags buf loc =
1814 messages = emptyMessages,
1815 last_loc = mkSrcSpan loc loc,
1818 extsBitmap = fromIntegral bitmap,
1820 lex_state = [bol, 0],
1821 alr_pending_implicit_tokens = [],
1822 alr_next_token = Nothing,
1823 alr_last_loc = noSrcSpan,
1825 alr_expecting_ocurly = Nothing,
1826 alr_justClosedExplicitLetBlock = False
1829 bitmap = genericsBit `setBitIf` xopt Opt_Generics flags
1830 .|. ffiBit `setBitIf` xopt Opt_ForeignFunctionInterface flags
1831 .|. parrBit `setBitIf` xopt Opt_PArr flags
1832 .|. arrowsBit `setBitIf` xopt Opt_Arrows flags
1833 .|. thBit `setBitIf` xopt Opt_TemplateHaskell flags
1834 .|. qqBit `setBitIf` xopt Opt_QuasiQuotes flags
1835 .|. ipBit `setBitIf` xopt Opt_ImplicitParams flags
1836 .|. explicitForallBit `setBitIf` xopt Opt_ExplicitForAll flags
1837 .|. bangPatBit `setBitIf` xopt Opt_BangPatterns flags
1838 .|. tyFamBit `setBitIf` xopt Opt_TypeFamilies flags
1839 .|. haddockBit `setBitIf` dopt Opt_Haddock flags
1840 .|. magicHashBit `setBitIf` xopt Opt_MagicHash flags
1841 .|. kindSigsBit `setBitIf` xopt Opt_KindSignatures flags
1842 .|. recursiveDoBit `setBitIf` xopt Opt_RecursiveDo flags
1843 .|. recBit `setBitIf` xopt Opt_DoRec flags
1844 .|. recBit `setBitIf` xopt Opt_Arrows flags
1845 .|. unicodeSyntaxBit `setBitIf` xopt Opt_UnicodeSyntax flags
1846 .|. unboxedTuplesBit `setBitIf` xopt Opt_UnboxedTuples flags
1847 .|. datatypeContextsBit `setBitIf` xopt Opt_DatatypeContexts flags
1848 .|. transformComprehensionsBit `setBitIf` xopt Opt_TransformListComp flags
1849 .|. rawTokenStreamBit `setBitIf` dopt Opt_KeepRawTokenStream flags
1850 .|. newQualOpsBit `setBitIf` xopt Opt_NewQualifiedOperators flags
1851 .|. alternativeLayoutRuleBit `setBitIf` xopt Opt_AlternativeLayoutRule flags
1853 setBitIf :: Int -> Bool -> Int
1854 b `setBitIf` cond | cond = bit b
1857 addWarning :: DynFlag -> SrcSpan -> SDoc -> P ()
1858 addWarning option srcspan warning
1859 = P $ \s@PState{messages=(ws,es), dflags=d} ->
1860 let warning' = mkWarnMsg srcspan alwaysQualify warning
1861 ws' = if dopt option d then ws `snocBag` warning' else ws
1862 in POk s{messages=(ws', es)} ()
1864 getMessages :: PState -> Messages
1865 getMessages PState{messages=ms} = ms
1867 getContext :: P [LayoutContext]
1868 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1870 setContext :: [LayoutContext] -> P ()
1871 setContext ctx = P $ \s -> POk s{context=ctx} ()
1874 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1875 last_len = len, last_loc = last_loc }) ->
1877 (_:tl) -> POk s{ context = tl } ()
1878 [] -> PFailed last_loc (srcParseErr buf len)
1880 -- Push a new layout context at the indentation of the last token read.
1881 -- This is only used at the outer level of a module when the 'module'
1882 -- keyword is missing.
1883 pushCurrentContext :: P ()
1884 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1885 POk s{context = Layout (srcSpanStartCol loc) : ctx} ()
1887 getOffside :: P Ordering
1888 getOffside = P $ \s@PState{last_loc=loc, context=stk} ->
1889 let offs = srcSpanStartCol loc in
1890 let ord = case stk of
1891 (Layout n:_) -> --trace ("layout: " ++ show n ++ ", offs: " ++ show offs) $
1896 -- ---------------------------------------------------------------------------
1897 -- Construct a parse error
1900 :: StringBuffer -- current buffer (placed just after the last token)
1901 -> Int -- length of the previous token
1904 = hcat [ if null token
1905 then ptext (sLit "parse error (possibly incorrect indentation)")
1906 else hcat [ptext (sLit "parse error on input "),
1907 char '`', text token, char '\'']
1909 where token = lexemeToString (offsetBytes (-len) buf) len
1911 -- Report a parse failure, giving the span of the previous token as
1912 -- the location of the error. This is the entry point for errors
1913 -- detected during parsing.
1915 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1916 last_loc = last_loc } ->
1917 PFailed last_loc (srcParseErr buf len)
1919 -- A lexical error is reported at a particular position in the source file,
1920 -- not over a token range.
1921 lexError :: String -> P a
1924 (AI end buf) <- getInput
1925 reportLexError loc end buf str
1927 -- -----------------------------------------------------------------------------
1928 -- This is the top-level function: called from the parser each time a
1929 -- new token is to be read from the input.
1931 lexer :: (Located Token -> P a) -> P a
1933 alr <- extension alternativeLayoutRule
1934 let lexTokenFun = if alr then lexTokenAlr else lexToken
1935 tok@(L _span _tok__) <- lexTokenFun
1936 --trace ("token: " ++ show _tok__) $ do
1939 lexTokenAlr :: P (Located Token)
1940 lexTokenAlr = do mPending <- popPendingImplicitToken
1941 t <- case mPending of
1943 do mNext <- popNextToken
1946 Just next -> return next
1947 alternativeLayoutRuleToken t
1950 setAlrLastLoc (getLoc t)
1952 ITwhere -> setAlrExpectingOCurly (Just ALRLayoutWhere)
1953 ITlet -> setAlrExpectingOCurly (Just ALRLayoutLet)
1954 ITof -> setAlrExpectingOCurly (Just ALRLayoutOf)
1955 ITdo -> setAlrExpectingOCurly (Just ALRLayoutDo)
1956 ITmdo -> setAlrExpectingOCurly (Just ALRLayoutDo)
1957 ITrec -> setAlrExpectingOCurly (Just ALRLayoutDo)
1961 alternativeLayoutRuleToken :: Located Token -> P (Located Token)
1962 alternativeLayoutRuleToken t
1963 = do context <- getALRContext
1964 lastLoc <- getAlrLastLoc
1965 mExpectingOCurly <- getAlrExpectingOCurly
1966 justClosedExplicitLetBlock <- getJustClosedExplicitLetBlock
1967 setJustClosedExplicitLetBlock False
1968 dflags <- getDynFlags
1969 let transitional = xopt Opt_AlternativeLayoutRuleTransitional dflags
1971 thisCol = srcSpanStartCol thisLoc
1972 newLine = (lastLoc == noSrcSpan)
1973 || (srcSpanStartLine thisLoc > srcSpanEndLine lastLoc)
1974 case (unLoc t, context, mExpectingOCurly) of
1975 -- This case handles a GHC extension to the original H98
1977 (ITocurly, _, Just alrLayout) ->
1978 do setAlrExpectingOCurly Nothing
1979 let isLet = case alrLayout of
1980 ALRLayoutLet -> True
1982 setALRContext (ALRNoLayout (containsCommas ITocurly) isLet : context)
1984 -- ...and makes this case unnecessary
1986 -- I think our implicit open-curly handling is slightly
1987 -- different to John's, in how it interacts with newlines
1989 (ITocurly, _, Just _) ->
1990 do setAlrExpectingOCurly Nothing
1994 (_, ALRLayout _ col : ls, Just expectingOCurly)
1995 | (thisCol > col) ||
1997 isNonDecreasingIntentation expectingOCurly) ->
1998 do setAlrExpectingOCurly Nothing
1999 setALRContext (ALRLayout expectingOCurly thisCol : context)
2001 return (L thisLoc ITocurly)
2003 do setAlrExpectingOCurly Nothing
2004 setPendingImplicitTokens [L lastLoc ITccurly]
2006 return (L lastLoc ITocurly)
2007 (_, _, Just expectingOCurly) ->
2008 do setAlrExpectingOCurly Nothing
2009 setALRContext (ALRLayout expectingOCurly thisCol : context)
2011 return (L thisLoc ITocurly)
2012 -- We do the [] cases earlier than in the spec, as we
2013 -- have an actual EOF token
2014 (ITeof, ALRLayout _ _ : ls, _) ->
2017 return (L thisLoc ITccurly)
2020 -- the other ITeof case omitted; general case below covers it
2022 | justClosedExplicitLetBlock ->
2024 (ITin, ALRLayout ALRLayoutLet _ : ls, _)
2026 do setPendingImplicitTokens [t]
2028 return (L thisLoc ITccurly)
2029 -- This next case is to handle a transitional issue:
2030 (ITwhere, ALRLayout _ col : ls, _)
2031 | newLine && thisCol == col && transitional ->
2032 do addWarning Opt_WarnAlternativeLayoutRuleTransitional
2034 (transitionalAlternativeLayoutWarning
2035 "`where' clause at the same depth as implicit layout block")
2038 -- Note that we use lastLoc, as we may need to close
2039 -- more layouts, or give a semicolon
2040 return (L lastLoc ITccurly)
2041 -- This next case is to handle a transitional issue:
2042 (ITvbar, ALRLayout _ col : ls, _)
2043 | newLine && thisCol == col && transitional ->
2044 do addWarning Opt_WarnAlternativeLayoutRuleTransitional
2046 (transitionalAlternativeLayoutWarning
2047 "`|' at the same depth as implicit layout block")
2050 -- Note that we use lastLoc, as we may need to close
2051 -- more layouts, or give a semicolon
2052 return (L lastLoc ITccurly)
2053 (_, ALRLayout _ col : ls, _)
2054 | newLine && thisCol == col ->
2056 return (L thisLoc ITsemi)
2057 | newLine && thisCol < col ->
2060 -- Note that we use lastLoc, as we may need to close
2061 -- more layouts, or give a semicolon
2062 return (L lastLoc ITccurly)
2063 -- We need to handle close before open, as 'then' is both
2064 -- an open and a close
2068 ALRLayout _ _ : ls ->
2071 return (L thisLoc ITccurly)
2072 ALRNoLayout _ isLet : ls ->
2073 do let ls' = if isALRopen u
2074 then ALRNoLayout (containsCommas u) False : ls
2077 when isLet $ setJustClosedExplicitLetBlock True
2080 do let ls = if isALRopen u
2081 then [ALRNoLayout (containsCommas u) False]
2084 -- XXX This is an error in John's code, but
2085 -- it looks reachable to me at first glance
2089 do setALRContext (ALRNoLayout (containsCommas u) False : context)
2091 (ITin, ALRLayout ALRLayoutLet _ : ls, _) ->
2093 setPendingImplicitTokens [t]
2094 return (L thisLoc ITccurly)
2095 (ITin, ALRLayout _ _ : ls, _) ->
2098 return (L thisLoc ITccurly)
2099 -- the other ITin case omitted; general case below covers it
2100 (ITcomma, ALRLayout _ _ : ls, _)
2101 | topNoLayoutContainsCommas ls ->
2104 return (L thisLoc ITccurly)
2105 (ITwhere, ALRLayout ALRLayoutDo _ : ls, _) ->
2107 setPendingImplicitTokens [t]
2108 return (L thisLoc ITccurly)
2109 -- the other ITwhere case omitted; general case below covers it
2110 (_, _, _) -> return t
2112 transitionalAlternativeLayoutWarning :: String -> SDoc
2113 transitionalAlternativeLayoutWarning msg
2114 = text "transitional layout will not be accepted in the future:"
2117 isALRopen :: Token -> Bool
2118 isALRopen ITcase = True
2119 isALRopen ITif = True
2120 isALRopen ITthen = True
2121 isALRopen IToparen = True
2122 isALRopen ITobrack = True
2123 isALRopen ITocurly = True
2125 isALRopen IToubxparen = True
2126 isALRopen ITparenEscape = True
2129 isALRclose :: Token -> Bool
2130 isALRclose ITof = True
2131 isALRclose ITthen = True
2132 isALRclose ITelse = True
2133 isALRclose ITcparen = True
2134 isALRclose ITcbrack = True
2135 isALRclose ITccurly = True
2137 isALRclose ITcubxparen = True
2138 isALRclose _ = False
2140 isNonDecreasingIntentation :: ALRLayout -> Bool
2141 isNonDecreasingIntentation ALRLayoutDo = True
2142 isNonDecreasingIntentation _ = False
2144 containsCommas :: Token -> Bool
2145 containsCommas IToparen = True
2146 containsCommas ITobrack = True
2147 -- John doesn't have {} as containing commas, but records contain them,
2148 -- which caused a problem parsing Cabal's Distribution.Simple.InstallDirs
2149 -- (defaultInstallDirs).
2150 containsCommas ITocurly = True
2152 containsCommas IToubxparen = True
2153 containsCommas _ = False
2155 topNoLayoutContainsCommas :: [ALRContext] -> Bool
2156 topNoLayoutContainsCommas [] = False
2157 topNoLayoutContainsCommas (ALRLayout _ _ : ls) = topNoLayoutContainsCommas ls
2158 topNoLayoutContainsCommas (ALRNoLayout b _ : _) = b
2160 lexToken :: P (Located Token)
2162 inp@(AI loc1 buf) <- getInput
2165 case alexScanUser exts inp sc of
2167 let span = mkSrcSpan loc1 loc1
2169 return (L span ITeof)
2170 AlexError (AI loc2 buf) ->
2171 reportLexError loc1 loc2 buf "lexical error"
2172 AlexSkip inp2 _ -> do
2175 AlexToken inp2@(AI end buf2) _ t -> do
2177 let span = mkSrcSpan loc1 end
2178 let bytes = byteDiff buf buf2
2179 span `seq` setLastToken span bytes
2182 reportLexError :: SrcLoc -> SrcLoc -> StringBuffer -> [Char] -> P a
2183 reportLexError loc1 loc2 buf str
2184 | atEnd buf = failLocMsgP loc1 loc2 (str ++ " at end of input")
2187 c = fst (nextChar buf)
2189 if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar#
2190 then failLocMsgP loc2 loc2 (str ++ " (UTF-8 decoding error)")
2191 else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c)
2193 lexTokenStream :: StringBuffer -> SrcLoc -> DynFlags -> ParseResult [Located Token]
2194 lexTokenStream buf loc dflags = unP go initState
2195 where dflags' = dopt_set (dopt_unset dflags Opt_Haddock) Opt_KeepRawTokenStream
2196 initState = mkPState dflags' buf loc
2198 ltok <- lexer return
2200 L _ ITeof -> return []
2201 _ -> liftM (ltok:) go
2203 linePrags = Map.singleton "line" (begin line_prag2)
2205 fileHeaderPrags = Map.fromList([("options", lex_string_prag IToptions_prag),
2206 ("options_ghc", lex_string_prag IToptions_prag),
2207 ("options_haddock", lex_string_prag ITdocOptions),
2208 ("language", token ITlanguage_prag),
2209 ("include", lex_string_prag ITinclude_prag)])
2211 ignoredPrags = Map.fromList (map ignored pragmas)
2212 where ignored opt = (opt, nested_comment lexToken)
2213 impls = ["hugs", "nhc98", "jhc", "yhc", "catch", "derive"]
2214 options_pragmas = map ("options_" ++) impls
2215 -- CFILES is a hugs-only thing.
2216 pragmas = options_pragmas ++ ["cfiles", "contract"]
2218 oneWordPrags = Map.fromList([("rules", rulePrag),
2219 ("inline", token (ITinline_prag Inline FunLike)),
2220 ("inlinable", token (ITinline_prag Inlinable FunLike)),
2221 ("inlineable", token (ITinline_prag Inlinable FunLike)),
2223 ("notinline", token (ITinline_prag NoInline FunLike)),
2224 ("specialize", token ITspec_prag),
2225 ("source", token ITsource_prag),
2226 ("warning", token ITwarning_prag),
2227 ("deprecated", token ITdeprecated_prag),
2228 ("scc", token ITscc_prag),
2229 ("generated", token ITgenerated_prag),
2230 ("core", token ITcore_prag),
2231 ("unpack", token ITunpack_prag),
2232 ("ann", token ITann_prag)])
2234 twoWordPrags = Map.fromList([("inline conlike", token (ITinline_prag Inline ConLike)),
2235 ("notinline conlike", token (ITinline_prag NoInline ConLike)),
2236 ("specialize inline", token (ITspec_inline_prag True)),
2237 ("specialize notinline", token (ITspec_inline_prag False))])
2240 dispatch_pragmas :: Map String Action -> Action
2241 dispatch_pragmas prags span buf len = case Map.lookup (clean_pragma (lexemeToString buf len)) prags of
2242 Just found -> found span buf len
2243 Nothing -> lexError "unknown pragma"
2245 known_pragma :: Map String Action -> AlexAccPred Int
2246 known_pragma prags _ _ len (AI _ buf) = (isJust $ Map.lookup (clean_pragma (lexemeToString (offsetBytes (- len) buf) len)) prags)
2247 && (nextCharIs buf (\c -> not (isAlphaNum c || c == '_')))
2249 clean_pragma :: String -> String
2250 clean_pragma prag = canon_ws (map toLower (unprefix prag))
2251 where unprefix prag' = case stripPrefix "{-#" prag' of
2254 canonical prag' = case prag' of
2255 "noinline" -> "notinline"
2256 "specialise" -> "specialize"
2257 "constructorlike" -> "conlike"
2259 canon_ws s = unwords (map canonical (words s))