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 RealSrcLoc
13 -----------------------------------------------------------------------------
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
23 -- qualified operator NOTES.
25 -- - If M.(+) is a single lexeme, then..
26 -- - Probably (+) should be a single lexeme too, for consistency.
27 -- Otherwise ( + ) would be a prefix operator, but M.( + ) would not be.
28 -- - But we have to rule out reserved operators, otherwise (..) becomes
29 -- a different lexeme.
30 -- - Should we therefore also rule out reserved operators in the qualified
31 -- form? This is quite difficult to achieve. We don't do it for
35 -- XXX The above flags turn off warnings in the generated code:
36 {-# LANGUAGE BangPatterns #-}
37 {-# OPTIONS_GHC -fno-warn-unused-matches #-}
38 {-# OPTIONS_GHC -fno-warn-unused-binds #-}
39 {-# OPTIONS_GHC -fno-warn-unused-imports #-}
40 {-# OPTIONS_GHC -fno-warn-missing-signatures #-}
41 -- But alex still generates some code that causes the "lazy unlifted bindings"
42 -- warning, and old compilers don't know about it so we can't easily turn
43 -- it off, so for now we use the sledge hammer:
44 {-# OPTIONS_GHC -w #-}
46 {-# OPTIONS_GHC -funbox-strict-fields #-}
49 Token(..), lexer, pragState, mkPState, PState(..),
50 P(..), ParseResult(..), getSrcLoc,
51 getPState, getDynFlags, withThisPackage,
52 failLocMsgP, failSpanMsgP, srcParseFail,
54 popContext, pushCurrentContext, setLastToken, setSrcLoc,
55 activeContext, nextIsEOF,
56 getLexState, popLexState, pushLexState,
57 extension, bangPatEnabled, datatypeContextsEnabled,
72 import BasicTypes ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) )
73 import Util ( readRational )
81 import qualified Data.Map as Map
85 $unispace = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
86 $whitechar = [\ \n\r\f\v $unispace]
87 $white_no_nl = $whitechar # \n
91 $unidigit = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
92 $decdigit = $ascdigit -- for now, should really be $digit (ToDo)
93 $digit = [$ascdigit $unidigit]
95 $special = [\(\)\,\;\[\]\`\{\}]
96 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
97 $unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
98 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
100 $unilarge = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
102 $large = [$asclarge $unilarge]
104 $unismall = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
106 $small = [$ascsmall $unismall \_]
108 $unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
109 $graphic = [$small $large $symbol $digit $special $unigraphic \:\"\']
112 $hexit = [$decdigit A-F a-f]
113 $symchar = [$symbol \:]
115 $idchar = [$small $large $digit \']
117 $pragmachar = [$small $large $digit]
119 $docsym = [\| \^ \* \$]
121 @varid = $small $idchar*
122 @conid = $large $idchar*
124 @varsym = $symbol $symchar*
125 @consym = \: $symchar*
127 @decimal = $decdigit+
129 @hexadecimal = $hexit+
130 @exponent = [eE] [\-\+]? @decimal
132 -- we support the hierarchical module name extension:
135 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
137 -- normal signed numerical literals can only be explicitly negative,
138 -- not explicitly positive (contrast @exponent)
140 @signed = @negative ?
144 -- everywhere: skip whitespace and comments
146 $tab+ { warn Opt_WarnTabs (text "Warning: Tab character") }
148 -- Everywhere: deal with nested comments. We explicitly rule out
149 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
150 -- (this can happen even though pragmas will normally take precedence due to
151 -- longest-match, because pragmas aren't valid in every state, but comments
152 -- are). We also rule out nested Haddock comments, if the -haddock flag is
155 "{-" / { isNormalComment } { nested_comment lexToken }
157 -- Single-line comments are a bit tricky. Haskell 98 says that two or
158 -- more dashes followed by a symbol should be parsed as a varsym, so we
159 -- have to exclude those.
161 -- Since Haddock comments aren't valid in every state, we need to rule them
164 -- The following two rules match comments that begin with two dashes, but
165 -- continue with a different character. The rules test that this character
166 -- is not a symbol (in which case we'd have a varsym), and that it's not a
167 -- space followed by a Haddock comment symbol (docsym) (in which case we'd
168 -- have a Haddock comment). The rules then munch the rest of the line.
170 "-- " ~[$docsym \#] .* { lineCommentToken }
171 "--" [^$symbol : \ ] .* { lineCommentToken }
173 -- Next, match Haddock comments if no -haddock flag
175 "-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
177 -- Now, when we've matched comments that begin with 2 dashes and continue
178 -- with a different character, we need to match comments that begin with three
179 -- or more dashes (which clearly can't be Haddock comments). We only need to
180 -- make sure that the first non-dash character isn't a symbol, and munch the
183 "---"\-* [^$symbol :] .* { lineCommentToken }
185 -- Since the previous rules all match dashes followed by at least one
186 -- character, we also need to match a whole line filled with just dashes.
188 "--"\-* / { atEOL } { lineCommentToken }
190 -- We need this rule since none of the other single line comment rules
191 -- actually match this case.
193 "-- " / { atEOL } { lineCommentToken }
195 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
196 -- blank lines) until we find a non-whitespace character, then do layout
199 -- One slight wibble here: what if the line begins with {-#? In
200 -- theory, we have to lex the pragma to see if it's one we recognise,
201 -- and if it is, then we backtrack and do_bol, otherwise we treat it
202 -- as a nested comment. We don't bother with this: if the line begins
203 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
206 ^\# (line)? { begin line_prag1 }
207 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
208 ^\# \! .* \n ; -- #!, for scripts
212 -- after a layout keyword (let, where, do, of), we begin a new layout
213 -- context if the curly brace is missing.
214 -- Careful! This stuff is quite delicate.
215 <layout, layout_do> {
216 \{ / { notFollowedBy '-' } { hopefully_open_brace }
217 -- we might encounter {-# here, but {- has been handled already
219 ^\# (line)? { begin line_prag1 }
222 -- do is treated in a subtly different way, see new_layout_context
223 <layout> () { new_layout_context True }
224 <layout_do> () { new_layout_context False }
226 -- after a new layout context which was found to be to the left of the
227 -- previous context, we have generated a '{' token, and we now need to
228 -- generate a matching '}' token.
229 <layout_left> () { do_layout_left }
231 <0,option_prags> \n { begin bol }
233 "{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
234 { dispatch_pragmas linePrags }
236 -- single-line line pragmas, of the form
237 -- # <line> "<file>" <extra-stuff> \n
238 <line_prag1> $decdigit+ { setLine line_prag1a }
239 <line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
240 <line_prag1b> .* { pop }
242 -- Haskell-style line pragmas, of the form
243 -- {-# LINE <line> "<file>" #-}
244 <line_prag2> $decdigit+ { setLine line_prag2a }
245 <line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
246 <line_prag2b> "#-}"|"-}" { pop }
247 -- NOTE: accept -} at the end of a LINE pragma, for compatibility
248 -- with older versions of GHC which generated these.
251 "{-#" $whitechar* $pragmachar+
252 $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
253 { dispatch_pragmas twoWordPrags }
255 "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
256 { dispatch_pragmas oneWordPrags }
258 -- We ignore all these pragmas, but don't generate a warning for them
259 "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
260 { dispatch_pragmas ignoredPrags }
262 -- ToDo: should only be valid inside a pragma:
267 "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
268 { dispatch_pragmas fileHeaderPrags }
270 "-- #" { multiline_doc_comment }
274 -- In the "0" mode we ignore these pragmas
275 "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
276 { nested_comment lexToken }
280 "-- #" .* { lineCommentToken }
284 "{-#" { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
285 (nested_comment lexToken) }
288 -- '0' state: ordinary lexemes
293 "-- " $docsym / { ifExtension haddockEnabled } { multiline_doc_comment }
294 "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
300 "[:" / { ifExtension parrEnabled } { token ITopabrack }
301 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
305 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
306 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
307 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
308 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
309 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
310 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
311 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
312 "$(" / { ifExtension thEnabled } { token ITparenEscape }
314 -- For backward compatibility, accept the old dollar syntax
315 "[$" @varid "|" / { ifExtension qqEnabled }
316 { lex_quasiquote_tok }
318 "[" @varid "|" / { ifExtension qqEnabled }
319 { lex_quasiquote_tok }
323 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
324 { special IToparenbar }
325 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
329 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
333 "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
334 { token IToubxparen }
335 "#)" / { ifExtension unboxedTuplesEnabled }
336 { token ITcubxparen }
340 \( { special IToparen }
341 \) { special ITcparen }
342 \[ { special ITobrack }
343 \] { special ITcbrack }
344 \, { special ITcomma }
345 \; { special ITsemi }
346 \` { special ITbackquote }
353 @qual @varid { idtoken qvarid }
354 @qual @conid { idtoken qconid }
356 @conid { idtoken conid }
360 @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
361 @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
362 @varid "#"+ / { ifExtension magicHashEnabled } { varid }
363 @conid "#"+ / { ifExtension magicHashEnabled } { idtoken conid }
366 -- ToDo: - move `var` and (sym) into lexical syntax?
367 -- - remove backquote from $special?
369 @qual @varsym { idtoken qvarsym }
370 @qual @consym { idtoken qconsym }
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
467 | ITinline_prag InlineSpec RuleMatchInfo
468 | ITspec_prag -- SPECIALISE
469 | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE)
477 | ITcore_prag -- hdaume: core annotations
481 | IToptions_prag String
482 | ITinclude_prag String
488 | ITdotdot -- reserved symbols
504 | ITbiglam -- GHC-extension symbols
506 | ITocurly -- special symbols
508 | ITocurlybar -- {|, for type applications
509 | ITccurlybar -- |}, for type applications
513 | ITopabrack -- [:, for parallel arrays with -XParallelArrays
514 | ITcpabrack -- :], for parallel arrays with -XParallelArrays
525 | ITvarid FastString -- identifiers
527 | ITvarsym FastString
528 | ITconsym FastString
529 | ITqvarid (FastString,FastString)
530 | ITqconid (FastString,FastString)
531 | ITqvarsym (FastString,FastString)
532 | ITqconsym (FastString,FastString)
533 | ITprefixqvarsym (FastString,FastString)
534 | ITprefixqconsym (FastString,FastString)
536 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
539 | ITstring FastString
541 | ITrational FractionalLit
544 | ITprimstring FastString
547 | ITprimfloat FractionalLit
548 | ITprimdouble FractionalLit
550 -- Template Haskell extension tokens
551 | ITopenExpQuote -- [| or [e|
552 | ITopenPatQuote -- [p|
553 | ITopenDecQuote -- [d|
554 | ITopenTypQuote -- [t|
556 | ITidEscape FastString -- $x
557 | ITparenEscape -- $(
560 | ITquasiQuote (FastString,FastString,RealSrcSpan) -- [:...|...|]
562 -- Arrow notation extension
569 | ITLarrowtail -- -<<
570 | ITRarrowtail -- >>-
572 | ITunknown String -- Used when the lexer can't make sense of it
573 | ITeof -- end of file token
575 -- Documentation annotations
576 | ITdocCommentNext String -- something beginning '-- |'
577 | ITdocCommentPrev String -- something beginning '-- ^'
578 | ITdocCommentNamed String -- something beginning '-- $'
579 | ITdocSection Int String -- a section heading
580 | ITdocOptions String -- doc options (prune, ignore-exports, etc)
581 | ITdocOptionsOld String -- doc options declared "-- # ..."-style
582 | ITlineComment String -- comment starting by "--"
583 | ITblockComment String -- comment in {- -}
586 deriving Show -- debugging
590 isSpecial :: Token -> Bool
591 -- If we see M.x, where x is a keyword, but
592 -- is special, we treat is as just plain M.x,
594 isSpecial ITas = True
595 isSpecial IThiding = True
596 isSpecial ITqualified = True
597 isSpecial ITforall = True
598 isSpecial ITexport = True
599 isSpecial ITlabel = True
600 isSpecial ITdynamic = True
601 isSpecial ITsafe = True
602 isSpecial ITthreadsafe = True
603 isSpecial ITinterruptible = True
604 isSpecial ITunsafe = True
605 isSpecial ITccallconv = True
606 isSpecial ITstdcallconv = True
607 isSpecial ITprimcallconv = True
608 isSpecial ITmdo = True
609 isSpecial ITfamily = True
610 isSpecial ITgroup = True
611 isSpecial ITby = True
612 isSpecial ITusing = True
616 -- the bitmap provided as the third component indicates whether the
617 -- corresponding extension keyword is valid under the extension options
618 -- provided to the compiler; if the extension corresponding to *any* of the
619 -- bits set in the bitmap is enabled, the keyword is valid (this setup
620 -- facilitates using a keyword in two different extensions that can be
621 -- activated independently)
623 reservedWordsFM :: UniqFM (Token, Int)
624 reservedWordsFM = listToUFM $
625 map (\(x, y, z) -> (mkFastString x, (y, z)))
626 [( "_", ITunderscore, 0 ),
628 ( "case", ITcase, 0 ),
629 ( "class", ITclass, 0 ),
630 ( "data", ITdata, 0 ),
631 ( "default", ITdefault, 0 ),
632 ( "deriving", ITderiving, 0 ),
634 ( "else", ITelse, 0 ),
635 ( "hiding", IThiding, 0 ),
637 ( "import", ITimport, 0 ),
639 ( "infix", ITinfix, 0 ),
640 ( "infixl", ITinfixl, 0 ),
641 ( "infixr", ITinfixr, 0 ),
642 ( "instance", ITinstance, 0 ),
644 ( "module", ITmodule, 0 ),
645 ( "newtype", ITnewtype, 0 ),
647 ( "qualified", ITqualified, 0 ),
648 ( "then", ITthen, 0 ),
649 ( "type", ITtype, 0 ),
650 ( "where", ITwhere, 0 ),
651 ( "_scc_", ITscc, 0 ), -- ToDo: remove
653 ( "forall", ITforall, bit explicitForallBit .|. bit inRulePragBit),
654 ( "mdo", ITmdo, bit recursiveDoBit),
655 ( "family", ITfamily, bit tyFamBit),
656 ( "group", ITgroup, bit transformComprehensionsBit),
657 ( "by", ITby, bit transformComprehensionsBit),
658 ( "using", ITusing, bit transformComprehensionsBit),
660 ( "foreign", ITforeign, bit ffiBit),
661 ( "export", ITexport, bit ffiBit),
662 ( "label", ITlabel, bit ffiBit),
663 ( "dynamic", ITdynamic, bit ffiBit),
664 ( "safe", ITsafe, bit ffiBit),
665 ( "threadsafe", ITthreadsafe, bit ffiBit), -- ToDo: remove
666 ( "interruptible", ITinterruptible, bit ffiBit),
667 ( "unsafe", ITunsafe, bit ffiBit),
668 ( "stdcall", ITstdcallconv, bit ffiBit),
669 ( "ccall", ITccallconv, bit ffiBit),
670 ( "prim", ITprimcallconv, bit ffiBit),
672 ( "rec", ITrec, bit recBit),
673 ( "proc", ITproc, bit arrowsBit)
676 reservedSymsFM :: UniqFM (Token, Int -> Bool)
677 reservedSymsFM = listToUFM $
678 map (\ (x,y,z) -> (mkFastString x,(y,z)))
679 [ ("..", ITdotdot, always)
680 -- (:) is a reserved op, meaning only list cons
681 ,(":", ITcolon, always)
682 ,("::", ITdcolon, always)
683 ,("=", ITequal, always)
684 ,("\\", ITlam, always)
685 ,("|", ITvbar, always)
686 ,("<-", ITlarrow, always)
687 ,("->", ITrarrow, always)
689 ,("~", ITtilde, always)
690 ,("=>", ITdarrow, always)
691 ,("-", ITminus, always)
692 ,("!", ITbang, always)
694 -- For data T (a::*) = MkT
695 ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
696 -- For 'forall a . t'
697 ,(".", ITdot, always) -- \i -> explicitForallEnabled i || inRulePrag i)
699 ,("-<", ITlarrowtail, arrowsEnabled)
700 ,(">-", ITrarrowtail, arrowsEnabled)
701 ,("-<<", ITLarrowtail, arrowsEnabled)
702 ,(">>-", ITRarrowtail, arrowsEnabled)
704 ,("∷", ITdcolon, unicodeSyntaxEnabled)
705 ,("⇒", ITdarrow, unicodeSyntaxEnabled)
706 ,("∀", ITforall, \i -> unicodeSyntaxEnabled i &&
707 explicitForallEnabled i)
708 ,("→", ITrarrow, unicodeSyntaxEnabled)
709 ,("←", ITlarrow, unicodeSyntaxEnabled)
711 ,("⤙", ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
712 ,("⤚", ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
713 ,("⤛", ITLarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
714 ,("⤜", ITRarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
716 ,("★", ITstar, unicodeSyntaxEnabled)
718 -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
719 -- form part of a large operator. This would let us have a better
720 -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
723 -- -----------------------------------------------------------------------------
726 type Action = RealSrcSpan -> StringBuffer -> Int -> P (RealLocated Token)
728 special :: Token -> Action
729 special tok span _buf _len = return (L span tok)
731 token, layout_token :: Token -> Action
732 token t span _buf _len = return (L span t)
733 layout_token t span _buf _len = pushLexState layout >> return (L span t)
735 idtoken :: (StringBuffer -> Int -> Token) -> Action
736 idtoken f span buf len = return (L span $! (f buf len))
738 skip_one_varid :: (FastString -> Token) -> Action
739 skip_one_varid f span buf len
740 = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
742 strtoken :: (String -> Token) -> Action
743 strtoken f span buf len =
744 return (L span $! (f $! lexemeToString buf len))
746 init_strtoken :: Int -> (String -> Token) -> Action
747 -- like strtoken, but drops the last N character(s)
748 init_strtoken drop f span buf len =
749 return (L span $! (f $! lexemeToString buf (len-drop)))
751 begin :: Int -> Action
752 begin code _span _str _len = do pushLexState code; lexToken
755 pop _span _buf _len = do _ <- popLexState
758 hopefully_open_brace :: Action
759 hopefully_open_brace span buf len
760 = do relaxed <- extension relaxedLayout
763 let offset = srcLocCol l
766 Layout prev_off : _ -> prev_off < offset
768 if isOK then pop_and open_brace span buf len
769 else failSpanMsgP (RealSrcSpan span) (text "Missing block")
771 pop_and :: Action -> Action
772 pop_and act span buf len = do _ <- popLexState
775 {-# INLINE nextCharIs #-}
776 nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
777 nextCharIs buf p = not (atEnd buf) && p (currentChar buf)
779 notFollowedBy :: Char -> AlexAccPred Int
780 notFollowedBy char _ _ _ (AI _ buf)
781 = nextCharIs buf (/=char)
783 notFollowedBySymbol :: AlexAccPred Int
784 notFollowedBySymbol _ _ _ (AI _ buf)
785 = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
787 -- We must reject doc comments as being ordinary comments everywhere.
788 -- In some cases the doc comment will be selected as the lexeme due to
789 -- maximal munch, but not always, because the nested comment rule is
790 -- valid in all states, but the doc-comment rules are only valid in
791 -- the non-layout states.
792 isNormalComment :: AlexAccPred Int
793 isNormalComment bits _ _ (AI _ buf)
794 | haddockEnabled bits = notFollowedByDocOrPragma
795 | otherwise = nextCharIs buf (/='#')
797 notFollowedByDocOrPragma
798 = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))
800 spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
801 spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
804 haddockDisabledAnd p bits _ _ (AI _ buf)
805 = if haddockEnabled bits then False else (p buf)
808 atEOL :: AlexAccPred Int
809 atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
811 ifExtension :: (Int -> Bool) -> AlexAccPred Int
812 ifExtension pred bits _ _ _ = pred bits
814 multiline_doc_comment :: Action
815 multiline_doc_comment span buf _len = withLexedDocType (worker "")
817 worker commentAcc input docType oneLine = case alexGetChar input of
819 | oneLine -> docCommentEnd input commentAcc docType buf span
820 | otherwise -> case checkIfCommentLine input' of
821 Just input -> worker ('\n':commentAcc) input docType False
822 Nothing -> docCommentEnd input commentAcc docType buf span
823 Just (c, input) -> worker (c:commentAcc) input docType oneLine
824 Nothing -> docCommentEnd input commentAcc docType buf span
826 checkIfCommentLine input = check (dropNonNewlineSpace input)
828 check input = case alexGetChar input of
829 Just ('-', input) -> case alexGetChar input of
830 Just ('-', input) -> case alexGetChar input of
831 Just (c, _) | c /= '-' -> Just input
836 dropNonNewlineSpace input = case alexGetChar input of
838 | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
842 lineCommentToken :: Action
843 lineCommentToken span buf len = do
844 b <- extension rawTokenStreamEnabled
845 if b then strtoken ITlineComment span buf len else lexToken
848 nested comments require traversing by hand, they can't be parsed
849 using regular expressions.
851 nested_comment :: P (RealLocated Token) -> Action
852 nested_comment cont span _str _len = do
856 go commentAcc 0 input = do setInput input
857 b <- extension rawTokenStreamEnabled
859 then docCommentEnd input commentAcc ITblockComment _str span
861 go commentAcc n input = case alexGetChar input of
862 Nothing -> errBrace input span
863 Just ('-',input) -> case alexGetChar input of
864 Nothing -> errBrace input span
865 Just ('\125',input) -> go commentAcc (n-1) input
866 Just (_,_) -> go ('-':commentAcc) n input
867 Just ('\123',input) -> case alexGetChar input of
868 Nothing -> errBrace input span
869 Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
870 Just (_,_) -> go ('\123':commentAcc) n input
871 Just (c,input) -> go (c:commentAcc) n input
873 nested_doc_comment :: Action
874 nested_doc_comment span buf _len = withLexedDocType (go "")
876 go commentAcc input docType _ = case alexGetChar input of
877 Nothing -> errBrace input span
878 Just ('-',input) -> case alexGetChar input of
879 Nothing -> errBrace input span
880 Just ('\125',input) ->
881 docCommentEnd input commentAcc docType buf span
882 Just (_,_) -> go ('-':commentAcc) input docType False
883 Just ('\123', input) -> case alexGetChar input of
884 Nothing -> errBrace input span
885 Just ('-',input) -> do
887 let cont = do input <- getInput; go commentAcc input docType False
888 nested_comment cont span buf _len
889 Just (_,_) -> go ('\123':commentAcc) input docType False
890 Just (c,input) -> go (c:commentAcc) input docType False
892 withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (RealLocated Token))
893 -> P (RealLocated Token)
894 withLexedDocType lexDocComment = do
895 input@(AI _ buf) <- getInput
896 case prevChar buf ' ' of
897 '|' -> lexDocComment input ITdocCommentNext False
898 '^' -> lexDocComment input ITdocCommentPrev False
899 '$' -> lexDocComment input ITdocCommentNamed False
900 '*' -> lexDocSection 1 input
901 '#' -> lexDocComment input ITdocOptionsOld False
902 _ -> panic "withLexedDocType: Bad doc type"
904 lexDocSection n input = case alexGetChar input of
905 Just ('*', input) -> lexDocSection (n+1) input
906 Just (_, _) -> lexDocComment input (ITdocSection n) True
907 Nothing -> do setInput input; lexToken -- eof reached, lex it normally
909 -- RULES pragmas turn on the forall and '.' keywords, and we turn them
910 -- off again at the end of the pragma.
912 rulePrag span _buf _len = do
913 setExts (.|. bit inRulePragBit)
914 return (L span ITrules_prag)
917 endPrag span _buf _len = do
918 setExts (.&. complement (bit inRulePragBit))
919 return (L span ITclose_prag)
922 -------------------------------------------------------------------------------
923 -- This function is quite tricky. We can't just return a new token, we also
924 -- need to update the state of the parser. Why? Because the token is longer
925 -- than what was lexed by Alex, and the lexToken function doesn't know this, so
926 -- it writes the wrong token length to the parser state. This function is
927 -- called afterwards, so it can just update the state.
929 docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
930 RealSrcSpan -> P (RealLocated Token)
931 docCommentEnd input commentAcc docType buf span = do
933 let (AI loc nextBuf) = input
934 comment = reverse commentAcc
935 span' = mkRealSrcSpan (realSrcSpanStart span) loc
936 last_len = byteDiff buf nextBuf
938 span `seq` setLastToken span' last_len
939 return (L span' (docType comment))
941 errBrace :: AlexInput -> RealSrcSpan -> P a
942 errBrace (AI end _) span = failLocMsgP (realSrcSpanStart span) end "unterminated `{-'"
944 open_brace, close_brace :: Action
945 open_brace span _str _len = do
947 setContext (NoLayout:ctx)
948 return (L span ITocurly)
949 close_brace span _str _len = do
951 return (L span ITccurly)
953 qvarid, qconid :: StringBuffer -> Int -> Token
954 qvarid buf len = ITqvarid $! splitQualName buf len False
955 qconid buf len = ITqconid $! splitQualName buf len False
957 splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
958 -- takes a StringBuffer and a length, and returns the module name
959 -- and identifier parts of a qualified name. Splits at the *last* dot,
960 -- because of hierarchical module names.
961 splitQualName orig_buf len parens = split orig_buf orig_buf
964 | orig_buf `byteDiff` buf >= len = done dot_buf
965 | c == '.' = found_dot buf'
966 | otherwise = split buf' dot_buf
968 (c,buf') = nextChar buf
970 -- careful, we might get names like M....
971 -- so, if the character after the dot is not upper-case, this is
972 -- the end of the qualifier part.
973 found_dot buf -- buf points after the '.'
974 | isUpper c = split buf' buf
975 | otherwise = done buf
977 (c,buf') = nextChar buf
980 (lexemeToFastString orig_buf (qual_size - 1),
981 if parens -- Prelude.(+)
982 then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
983 else lexemeToFastString dot_buf (len - qual_size))
985 qual_size = orig_buf `byteDiff` dot_buf
990 case lookupUFM reservedWordsFM fs of
991 Just (keyword,0) -> do
993 return (L span keyword)
994 Just (keyword,exts) -> do
995 b <- extension (\i -> exts .&. i /= 0)
996 if b then do maybe_layout keyword
997 return (L span keyword)
998 else return (L span (ITvarid fs))
999 _other -> return (L span (ITvarid fs))
1001 fs = lexemeToFastString buf len
1003 conid :: StringBuffer -> Int -> Token
1004 conid buf len = ITconid fs
1005 where fs = lexemeToFastString buf len
1007 qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
1008 qvarsym buf len = ITqvarsym $! splitQualName buf len False
1009 qconsym buf len = ITqconsym $! splitQualName buf len False
1010 prefixqvarsym buf len = ITprefixqvarsym $! splitQualName buf len True
1011 prefixqconsym buf len = ITprefixqconsym $! splitQualName buf len True
1013 varsym, consym :: Action
1014 varsym = sym ITvarsym
1015 consym = sym ITconsym
1017 sym :: (FastString -> Token) -> RealSrcSpan -> StringBuffer -> Int
1018 -> P (RealLocated Token)
1019 sym con span buf len =
1020 case lookupUFM reservedSymsFM fs of
1021 Just (keyword,exts) -> do
1023 if b then return (L span keyword)
1024 else return (L span $! con fs)
1025 _other -> return (L span $! con fs)
1027 fs = lexemeToFastString buf len
1029 -- Variations on the integral numeric literal.
1030 tok_integral :: (Integer -> Token)
1031 -> (Integer -> Integer)
1032 -- -> (StringBuffer -> StringBuffer) -> (Int -> Int)
1034 -> (Integer, (Char->Int)) -> Action
1035 tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
1036 return $ L span $ itint $! transint $ parseUnsignedInteger
1037 (offsetBytes transbuf buf) (subtract translen len) radix char_to_int
1039 -- some conveniences for use with tok_integral
1040 tok_num :: (Integer -> Integer)
1042 -> (Integer, (Char->Int)) -> Action
1043 tok_num = tok_integral ITinteger
1044 tok_primint :: (Integer -> Integer)
1046 -> (Integer, (Char->Int)) -> Action
1047 tok_primint = tok_integral ITprimint
1048 tok_primword :: Int -> Int
1049 -> (Integer, (Char->Int)) -> Action
1050 tok_primword = tok_integral ITprimword positive
1051 positive, negative :: (Integer -> Integer)
1054 decimal, octal, hexadecimal :: (Integer, Char -> Int)
1055 decimal = (10,octDecDigit)
1056 octal = (8,octDecDigit)
1057 hexadecimal = (16,hexDigit)
1059 -- readRational can understand negative rationals, exponents, everything.
1060 tok_float, tok_primfloat, tok_primdouble :: String -> Token
1061 tok_float str = ITrational $! readFractionalLit str
1062 tok_primfloat str = ITprimfloat $! readFractionalLit str
1063 tok_primdouble str = ITprimdouble $! readFractionalLit str
1065 readFractionalLit :: String -> FractionalLit
1066 readFractionalLit str = (FL $! str) $! readRational str
1068 -- -----------------------------------------------------------------------------
1069 -- Layout processing
1071 -- we're at the first token on a line, insert layout tokens if necessary
1073 do_bol span _str _len = do
1077 --trace "layout: inserting '}'" $ do
1079 -- do NOT pop the lex state, we might have a ';' to insert
1080 return (L span ITvccurly)
1082 --trace "layout: inserting ';'" $ do
1084 return (L span ITsemi)
1089 -- certain keywords put us in the "layout" state, where we might
1090 -- add an opening curly brace.
1091 maybe_layout :: Token -> P ()
1092 maybe_layout t = do -- If the alternative layout rule is enabled then
1093 -- we never create an implicit layout context here.
1094 -- Layout is handled XXX instead.
1095 -- The code for closing implicit contexts, or
1096 -- inserting implicit semi-colons, is therefore
1097 -- irrelevant as it only applies in an implicit
1099 alr <- extension alternativeLayoutRule
1101 where f ITdo = pushLexState layout_do
1102 f ITmdo = pushLexState layout_do
1103 f ITof = pushLexState layout
1104 f ITlet = pushLexState layout
1105 f ITwhere = pushLexState layout
1106 f ITrec = pushLexState layout
1109 -- Pushing a new implicit layout context. If the indentation of the
1110 -- next token is not greater than the previous layout context, then
1111 -- Haskell 98 says that the new layout context should be empty; that is
1112 -- the lexer must generate {}.
1114 -- We are slightly more lenient than this: when the new context is started
1115 -- by a 'do', then we allow the new context to be at the same indentation as
1116 -- the previous context. This is what the 'strict' argument is for.
1118 new_layout_context :: Bool -> Action
1119 new_layout_context strict span _buf _len = do
1121 (AI l _) <- getInput
1122 let offset = srcLocCol l
1124 nondecreasing <- extension nondecreasingIndentation
1125 let strict' = strict || not nondecreasing
1127 Layout prev_off : _ |
1128 (strict' && prev_off >= offset ||
1129 not strict' && prev_off > offset) -> do
1130 -- token is indented to the left of the previous context.
1131 -- we must generate a {} sequence now.
1132 pushLexState layout_left
1133 return (L span ITvocurly)
1135 setContext (Layout offset : ctx)
1136 return (L span ITvocurly)
1138 do_layout_left :: Action
1139 do_layout_left span _buf _len = do
1141 pushLexState bol -- we must be at the start of a line
1142 return (L span ITvccurly)
1144 -- -----------------------------------------------------------------------------
1147 setLine :: Int -> Action
1148 setLine code span buf len = do
1149 let line = parseUnsignedInteger buf len 10 octDecDigit
1150 setSrcLoc (mkRealSrcLoc (srcSpanFile span) (fromIntegral line - 1) 1)
1151 -- subtract one: the line number refers to the *following* line
1156 setFile :: Int -> Action
1157 setFile code span buf len = do
1158 let file = lexemeToFastString (stepOn buf) (len-2)
1159 setAlrLastLoc $ alrInitialLoc file
1160 setSrcLoc (mkRealSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1165 alrInitialLoc :: FastString -> RealSrcSpan
1166 alrInitialLoc file = mkRealSrcSpan loc loc
1167 where -- This is a hack to ensure that the first line in a file
1168 -- looks like it is after the initial location:
1169 loc = mkRealSrcLoc file (-1) (-1)
1171 -- -----------------------------------------------------------------------------
1172 -- Options, includes and language pragmas.
1174 lex_string_prag :: (String -> Token) -> Action
1175 lex_string_prag mkTok span _buf _len
1176 = do input <- getInput
1180 return (L (mkRealSrcSpan start end) tok)
1182 = if isString input "#-}"
1183 then do setInput input
1184 return (mkTok (reverse acc))
1185 else case alexGetChar input of
1186 Just (c,i) -> go (c:acc) i
1187 Nothing -> err input
1188 isString _ [] = True
1190 = case alexGetChar i of
1191 Just (c,i') | c == x -> isString i' xs
1193 err (AI end _) = failLocMsgP (realSrcSpanStart span) end "unterminated options pragma"
1196 -- -----------------------------------------------------------------------------
1199 -- This stuff is horrible. I hates it.
1201 lex_string_tok :: Action
1202 lex_string_tok span _buf _len = do
1203 tok <- lex_string ""
1205 return (L (mkRealSrcSpan (realSrcSpanStart span) end) tok)
1207 lex_string :: String -> P Token
1210 case alexGetChar' i of
1211 Nothing -> lit_error i
1215 magicHash <- extension magicHashEnabled
1219 case alexGetChar' i of
1223 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1224 else let s' = mkZFastString (reverse s) in
1225 return (ITprimstring s')
1226 -- mkZFastString is a hack to avoid encoding the
1227 -- string in UTF-8. We just want the exact bytes.
1229 return (ITstring (mkFastString (reverse s)))
1231 return (ITstring (mkFastString (reverse s)))
1234 | Just ('&',i) <- next -> do
1235 setInput i; lex_string s
1236 | Just (c,i) <- next, c <= '\x7f' && is_space c -> do
1237 -- is_space only works for <= '\x7f' (#3751)
1238 setInput i; lex_stringgap s
1239 where next = alexGetChar' i
1243 '\\' -> do setInput i1; c' <- lex_escape; lex_string (c':s)
1244 c | isAny c -> do setInput i1; lex_string (c:s)
1245 _other -> lit_error i
1247 lex_stringgap :: String -> P Token
1248 lex_stringgap s = do
1250 c <- getCharOrFail i
1252 '\\' -> lex_string s
1253 c | is_space c -> lex_stringgap s
1254 _other -> lit_error i
1257 lex_char_tok :: Action
1258 -- Here we are basically parsing character literals, such as 'x' or '\n'
1259 -- but, when Template Haskell is on, we additionally spot
1260 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
1261 -- but WITHOUT CONSUMING the x or T part (the parser does that).
1262 -- So we have to do two characters of lookahead: when we see 'x we need to
1263 -- see if there's a trailing quote
1264 lex_char_tok span _buf _len = do -- We've seen '
1265 i1 <- getInput -- Look ahead to first character
1266 let loc = realSrcSpanStart span
1267 case alexGetChar' i1 of
1268 Nothing -> lit_error i1
1270 Just ('\'', i2@(AI end2 _)) -> do -- We've seen ''
1271 th_exts <- extension thEnabled
1274 return (L (mkRealSrcSpan loc end2) ITtyQuote)
1277 Just ('\\', i2@(AI _end2 _)) -> do -- We've seen 'backslash
1279 lit_ch <- lex_escape
1281 mc <- getCharOrFail i3 -- Trailing quote
1282 if mc == '\'' then finish_char_tok loc lit_ch
1285 Just (c, i2@(AI _end2 _))
1286 | not (isAny c) -> lit_error i1
1289 -- We've seen 'x, where x is a valid character
1290 -- (i.e. not newline etc) but not a quote or backslash
1291 case alexGetChar' i2 of -- Look ahead one more character
1292 Just ('\'', i3) -> do -- We've seen 'x'
1294 finish_char_tok loc c
1295 _other -> do -- We've seen 'x not followed by quote
1296 -- (including the possibility of EOF)
1297 -- If TH is on, just parse the quote only
1298 th_exts <- extension thEnabled
1300 if th_exts then return (L (mkRealSrcSpan loc end) ITvarQuote)
1303 finish_char_tok :: RealSrcLoc -> Char -> P (RealLocated Token)
1304 finish_char_tok loc ch -- We've already seen the closing quote
1305 -- Just need to check for trailing #
1306 = do magicHash <- extension magicHashEnabled
1307 i@(AI end _) <- getInput
1308 if magicHash then do
1309 case alexGetChar' i of
1310 Just ('#',i@(AI end _)) -> do
1312 return (L (mkRealSrcSpan loc end) (ITprimchar ch))
1314 return (L (mkRealSrcSpan loc end) (ITchar ch))
1316 return (L (mkRealSrcSpan loc end) (ITchar ch))
1318 isAny :: Char -> Bool
1319 isAny c | c > '\x7f' = isPrint c
1320 | otherwise = is_any c
1322 lex_escape :: P Char
1325 c <- getCharOrFail i0
1337 '^' -> do i1 <- getInput
1338 c <- getCharOrFail i1
1339 if c >= '@' && c <= '_'
1340 then return (chr (ord c - ord '@'))
1343 'x' -> readNum is_hexdigit 16 hexDigit
1344 'o' -> readNum is_octdigit 8 octDecDigit
1345 x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x)
1349 case alexGetChar' i of
1350 Nothing -> lit_error i0
1352 case alexGetChar' i2 of
1353 Nothing -> do lit_error i0
1355 let str = [c1,c2,c3] in
1356 case [ (c,rest) | (p,c) <- silly_escape_chars,
1357 Just rest <- [stripPrefix p str] ] of
1358 (escape_char,[]):_ -> do
1361 (escape_char,_:_):_ -> do
1366 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
1367 readNum is_digit base conv = do
1369 c <- getCharOrFail i
1371 then readNum2 is_digit base conv (conv c)
1374 readNum2 :: (Char -> Bool) -> Int -> (Char -> Int) -> Int -> P Char
1375 readNum2 is_digit base conv i = do
1378 where read i input = do
1379 case alexGetChar' input of
1380 Just (c,input') | is_digit c -> do
1381 let i' = i*base + conv c
1383 then setInput input >> lexError "numeric escape sequence out of range"
1386 setInput input; return (chr i)
1389 silly_escape_chars :: [(String, Char)]
1390 silly_escape_chars = [
1427 -- before calling lit_error, ensure that the current input is pointing to
1428 -- the position of the error in the buffer. This is so that we can report
1429 -- a correct location to the user, but also so we can detect UTF-8 decoding
1430 -- errors if they occur.
1431 lit_error :: AlexInput -> P a
1432 lit_error i = do setInput i; lexError "lexical error in string/character literal"
1434 getCharOrFail :: AlexInput -> P Char
1435 getCharOrFail i = do
1436 case alexGetChar' i of
1437 Nothing -> lexError "unexpected end-of-file in string/character literal"
1438 Just (c,i) -> do setInput i; return c
1440 -- -----------------------------------------------------------------------------
1443 lex_quasiquote_tok :: Action
1444 lex_quasiquote_tok span buf len = do
1445 let quoter = tail (lexemeToString buf (len - 1))
1446 -- 'tail' drops the initial '[',
1447 -- while the -1 drops the trailing '|'
1448 quoteStart <- getSrcLoc
1449 quote <- lex_quasiquote ""
1451 return (L (mkRealSrcSpan (realSrcSpanStart span) end)
1452 (ITquasiQuote (mkFastString quoter,
1453 mkFastString (reverse quote),
1454 mkRealSrcSpan quoteStart end)))
1456 lex_quasiquote :: String -> P String
1457 lex_quasiquote s = do
1459 case alexGetChar' i of
1460 Nothing -> lit_error i
1463 | Just ('|',i) <- next -> do
1464 setInput i; lex_quasiquote ('|' : s)
1465 | Just (']',i) <- next -> do
1466 setInput i; lex_quasiquote (']' : s)
1467 where next = alexGetChar' i
1470 | Just (']',i) <- next -> do
1471 setInput i; return s
1472 where next = alexGetChar' i
1475 setInput i; lex_quasiquote (c : s)
1477 -- -----------------------------------------------------------------------------
1480 warn :: DynFlag -> SDoc -> Action
1481 warn option warning srcspan _buf _len = do
1482 addWarning option (RealSrcSpan srcspan) warning
1485 warnThen :: DynFlag -> SDoc -> Action -> Action
1486 warnThen option warning action srcspan buf len = do
1487 addWarning option (RealSrcSpan srcspan) warning
1488 action srcspan buf len
1490 -- -----------------------------------------------------------------------------
1501 SrcSpan -- The start and end of the text span related to
1502 -- the error. Might be used in environments which can
1503 -- show this span, e.g. by highlighting it.
1504 Message -- The error message
1506 data PState = PState {
1507 buffer :: StringBuffer,
1509 messages :: Messages,
1510 last_loc :: RealSrcSpan, -- pos of previous token
1511 last_len :: !Int, -- len of previous token
1512 loc :: RealSrcLoc, -- current loc (end of prev token + 1)
1513 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1514 context :: [LayoutContext],
1516 -- Used in the alternative layout rule:
1517 -- These tokens are the next ones to be sent out. They are
1518 -- just blindly emitted, without the rule looking at them again:
1519 alr_pending_implicit_tokens :: [RealLocated Token],
1520 -- This is the next token to be considered or, if it is Nothing,
1521 -- we need to get the next token from the input stream:
1522 alr_next_token :: Maybe (RealLocated Token),
1523 -- This is what we consider to be the locatino of the last token
1525 alr_last_loc :: RealSrcSpan,
1526 -- The stack of layout contexts:
1527 alr_context :: [ALRContext],
1528 -- Are we expecting a '{'? If it's Just, then the ALRLayout tells
1529 -- us what sort of layout the '{' will open:
1530 alr_expecting_ocurly :: Maybe ALRLayout,
1531 -- Have we just had the '}' for a let block? If so, than an 'in'
1532 -- token doesn't need to close anything:
1533 alr_justClosedExplicitLetBlock :: Bool
1535 -- last_loc and last_len are used when generating error messages,
1536 -- and in pushCurrentContext only. Sigh, if only Happy passed the
1537 -- current token to happyError, we could at least get rid of last_len.
1538 -- Getting rid of last_loc would require finding another way to
1539 -- implement pushCurrentContext (which is only called from one place).
1541 data ALRContext = ALRNoLayout Bool{- does it contain commas? -}
1542 Bool{- is it a 'let' block? -}
1543 | ALRLayout ALRLayout Int
1544 data ALRLayout = ALRLayoutLet
1549 newtype P a = P { unP :: PState -> ParseResult a }
1551 instance Monad P where
1557 returnP a = a `seq` (P $ \s -> POk s a)
1559 thenP :: P a -> (a -> P b) -> P b
1560 (P m) `thenP` k = P $ \ s ->
1562 POk s1 a -> (unP (k a)) s1
1563 PFailed span err -> PFailed span err
1565 failP :: String -> P a
1566 failP msg = P $ \s -> PFailed (RealSrcSpan (last_loc s)) (text msg)
1568 failMsgP :: String -> P a
1569 failMsgP msg = P $ \s -> PFailed (RealSrcSpan (last_loc s)) (text msg)
1571 failLocMsgP :: RealSrcLoc -> RealSrcLoc -> String -> P a
1572 failLocMsgP loc1 loc2 str = P $ \_ -> PFailed (RealSrcSpan (mkRealSrcSpan loc1 loc2)) (text str)
1574 failSpanMsgP :: SrcSpan -> SDoc -> P a
1575 failSpanMsgP span msg = P $ \_ -> PFailed span msg
1577 getPState :: P PState
1578 getPState = P $ \s -> POk s s
1580 getDynFlags :: P DynFlags
1581 getDynFlags = P $ \s -> POk s (dflags s)
1583 withThisPackage :: (PackageId -> a) -> P a
1585 = do pkg <- liftM thisPackage getDynFlags
1588 extension :: (Int -> Bool) -> P Bool
1589 extension p = P $ \s -> POk s (p $! extsBitmap s)
1592 getExts = P $ \s -> POk s (extsBitmap s)
1594 setExts :: (Int -> Int) -> P ()
1595 setExts f = P $ \s -> POk s{ extsBitmap = f (extsBitmap s) } ()
1597 setSrcLoc :: RealSrcLoc -> P ()
1598 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1600 getSrcLoc :: P RealSrcLoc
1601 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1603 setLastToken :: RealSrcSpan -> Int -> P ()
1604 setLastToken loc len = P $ \s -> POk s {
1609 data AlexInput = AI RealSrcLoc StringBuffer
1611 alexInputPrevChar :: AlexInput -> Char
1612 alexInputPrevChar (AI _ buf) = prevChar buf '\n'
1614 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1615 alexGetChar (AI loc s)
1617 | otherwise = adj_c `seq` loc' `seq` s' `seq`
1618 --trace (show (ord c)) $
1619 Just (adj_c, (AI loc' s'))
1620 where (c,s') = nextChar s
1621 loc' = advanceSrcLoc loc c
1629 other_graphic = '\x6'
1632 | c <= '\x06' = non_graphic
1634 -- Alex doesn't handle Unicode, so when Unicode
1635 -- character is encountered we output these values
1636 -- with the actual character value hidden in the state.
1638 case generalCategory c of
1639 UppercaseLetter -> upper
1640 LowercaseLetter -> lower
1641 TitlecaseLetter -> upper
1642 ModifierLetter -> other_graphic
1643 OtherLetter -> lower -- see #1103
1644 NonSpacingMark -> other_graphic
1645 SpacingCombiningMark -> other_graphic
1646 EnclosingMark -> other_graphic
1647 DecimalNumber -> digit
1648 LetterNumber -> other_graphic
1649 OtherNumber -> digit -- see #4373
1650 ConnectorPunctuation -> symbol
1651 DashPunctuation -> symbol
1652 OpenPunctuation -> other_graphic
1653 ClosePunctuation -> other_graphic
1654 InitialQuote -> other_graphic
1655 FinalQuote -> other_graphic
1656 OtherPunctuation -> symbol
1657 MathSymbol -> symbol
1658 CurrencySymbol -> symbol
1659 ModifierSymbol -> symbol
1660 OtherSymbol -> symbol
1662 _other -> non_graphic
1664 -- This version does not squash unicode characters, it is used when
1666 alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
1667 alexGetChar' (AI loc s)
1669 | otherwise = c `seq` loc' `seq` s' `seq`
1670 --trace (show (ord c)) $
1671 Just (c, (AI loc' s'))
1672 where (c,s') = nextChar s
1673 loc' = advanceSrcLoc loc c
1675 getInput :: P AlexInput
1676 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (AI l b)
1678 setInput :: AlexInput -> P ()
1679 setInput (AI l b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1686 pushLexState :: Int -> P ()
1687 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1689 popLexState :: P Int
1690 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1692 getLexState :: P Int
1693 getLexState = P $ \s@PState{ lex_state=ls:_ } -> POk s ls
1695 popNextToken :: P (Maybe (RealLocated Token))
1697 = P $ \s@PState{ alr_next_token = m } ->
1698 POk (s {alr_next_token = Nothing}) m
1700 activeContext :: P Bool
1702 ctxt <- getALRContext
1703 expc <- getAlrExpectingOCurly
1704 impt <- implicitTokenPending
1706 ([],Nothing) -> return impt
1707 _other -> return True
1709 setAlrLastLoc :: RealSrcSpan -> P ()
1710 setAlrLastLoc l = P $ \s -> POk (s {alr_last_loc = l}) ()
1712 getAlrLastLoc :: P RealSrcSpan
1713 getAlrLastLoc = P $ \s@(PState {alr_last_loc = l}) -> POk s l
1715 getALRContext :: P [ALRContext]
1716 getALRContext = P $ \s@(PState {alr_context = cs}) -> POk s cs
1718 setALRContext :: [ALRContext] -> P ()
1719 setALRContext cs = P $ \s -> POk (s {alr_context = cs}) ()
1721 getJustClosedExplicitLetBlock :: P Bool
1722 getJustClosedExplicitLetBlock
1723 = P $ \s@(PState {alr_justClosedExplicitLetBlock = b}) -> POk s b
1725 setJustClosedExplicitLetBlock :: Bool -> P ()
1726 setJustClosedExplicitLetBlock b
1727 = P $ \s -> POk (s {alr_justClosedExplicitLetBlock = b}) ()
1729 setNextToken :: RealLocated Token -> P ()
1730 setNextToken t = P $ \s -> POk (s {alr_next_token = Just t}) ()
1732 implicitTokenPending :: P Bool
1733 implicitTokenPending
1734 = P $ \s@PState{ alr_pending_implicit_tokens = ts } ->
1739 popPendingImplicitToken :: P (Maybe (RealLocated Token))
1740 popPendingImplicitToken
1741 = P $ \s@PState{ alr_pending_implicit_tokens = ts } ->
1744 (t : ts') -> POk (s {alr_pending_implicit_tokens = ts'}) (Just t)
1746 setPendingImplicitTokens :: [RealLocated Token] -> P ()
1747 setPendingImplicitTokens ts = P $ \s -> POk (s {alr_pending_implicit_tokens = ts}) ()
1749 getAlrExpectingOCurly :: P (Maybe ALRLayout)
1750 getAlrExpectingOCurly = P $ \s@(PState {alr_expecting_ocurly = b}) -> POk s b
1752 setAlrExpectingOCurly :: Maybe ALRLayout -> P ()
1753 setAlrExpectingOCurly b = P $ \s -> POk (s {alr_expecting_ocurly = b}) ()
1755 -- for reasons of efficiency, flags indicating language extensions (eg,
1756 -- -fglasgow-exts or -XParallelArrays) are represented by a bitmap stored in an unboxed
1759 -- The "genericsBit" is now unused, available for others
1760 -- genericsBit :: Int
1761 -- genericsBit = 0 -- {|, |} and "generic"
1773 explicitForallBit :: Int
1774 explicitForallBit = 7 -- the 'forall' keyword and '.' symbol
1776 bangPatBit = 8 -- Tells the parser to understand bang-patterns
1777 -- (doesn't affect the lexer)
1779 tyFamBit = 9 -- indexed type families: 'family' keyword and kind sigs
1781 haddockBit = 10 -- Lex and parse Haddock comments
1783 magicHashBit = 11 -- "#" in both functions and operators
1785 kindSigsBit = 12 -- Kind signatures on type variables
1786 recursiveDoBit :: Int
1787 recursiveDoBit = 13 -- mdo
1788 unicodeSyntaxBit :: Int
1789 unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc
1790 unboxedTuplesBit :: Int
1791 unboxedTuplesBit = 15 -- (# and #)
1792 datatypeContextsBit :: Int
1793 datatypeContextsBit = 16
1794 transformComprehensionsBit :: Int
1795 transformComprehensionsBit = 17
1797 qqBit = 18 -- enable quasiquoting
1798 inRulePragBit :: Int
1800 rawTokenStreamBit :: Int
1801 rawTokenStreamBit = 20 -- producing a token stream with all comments included
1804 alternativeLayoutRuleBit :: Int
1805 alternativeLayoutRuleBit = 23
1806 relaxedLayoutBit :: Int
1807 relaxedLayoutBit = 24
1808 nondecreasingIndentationBit :: Int
1809 nondecreasingIndentationBit = 25
1811 always :: Int -> Bool
1813 parrEnabled :: Int -> Bool
1814 parrEnabled flags = testBit flags parrBit
1815 arrowsEnabled :: Int -> Bool
1816 arrowsEnabled flags = testBit flags arrowsBit
1817 thEnabled :: Int -> Bool
1818 thEnabled flags = testBit flags thBit
1819 ipEnabled :: Int -> Bool
1820 ipEnabled flags = testBit flags ipBit
1821 explicitForallEnabled :: Int -> Bool
1822 explicitForallEnabled flags = testBit flags explicitForallBit
1823 bangPatEnabled :: Int -> Bool
1824 bangPatEnabled flags = testBit flags bangPatBit
1825 -- tyFamEnabled :: Int -> Bool
1826 -- tyFamEnabled flags = testBit flags tyFamBit
1827 haddockEnabled :: Int -> Bool
1828 haddockEnabled flags = testBit flags haddockBit
1829 magicHashEnabled :: Int -> Bool
1830 magicHashEnabled flags = testBit flags magicHashBit
1831 -- kindSigsEnabled :: Int -> Bool
1832 -- kindSigsEnabled flags = testBit flags kindSigsBit
1833 unicodeSyntaxEnabled :: Int -> Bool
1834 unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit
1835 unboxedTuplesEnabled :: Int -> Bool
1836 unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit
1837 datatypeContextsEnabled :: Int -> Bool
1838 datatypeContextsEnabled flags = testBit flags datatypeContextsBit
1839 qqEnabled :: Int -> Bool
1840 qqEnabled flags = testBit flags qqBit
1841 -- inRulePrag :: Int -> Bool
1842 -- inRulePrag flags = testBit flags inRulePragBit
1843 rawTokenStreamEnabled :: Int -> Bool
1844 rawTokenStreamEnabled flags = testBit flags rawTokenStreamBit
1845 alternativeLayoutRule :: Int -> Bool
1846 alternativeLayoutRule flags = testBit flags alternativeLayoutRuleBit
1847 relaxedLayout :: Int -> Bool
1848 relaxedLayout flags = testBit flags relaxedLayoutBit
1849 nondecreasingIndentation :: Int -> Bool
1850 nondecreasingIndentation flags = testBit flags nondecreasingIndentationBit
1852 -- PState for parsing options pragmas
1854 pragState :: DynFlags -> StringBuffer -> RealSrcLoc -> PState
1855 pragState dynflags buf loc = (mkPState dynflags buf loc) {
1856 lex_state = [bol, option_prags, 0]
1859 -- create a parse state
1861 mkPState :: DynFlags -> StringBuffer -> RealSrcLoc -> PState
1862 mkPState flags buf loc =
1866 messages = emptyMessages,
1867 last_loc = mkRealSrcSpan loc loc,
1870 extsBitmap = fromIntegral bitmap,
1872 lex_state = [bol, 0],
1873 alr_pending_implicit_tokens = [],
1874 alr_next_token = Nothing,
1875 alr_last_loc = alrInitialLoc (fsLit "<no file>"),
1877 alr_expecting_ocurly = Nothing,
1878 alr_justClosedExplicitLetBlock = False
1881 bitmap = ffiBit `setBitIf` xopt Opt_ForeignFunctionInterface flags
1882 .|. parrBit `setBitIf` xopt Opt_ParallelArrays flags
1883 .|. arrowsBit `setBitIf` xopt Opt_Arrows flags
1884 .|. thBit `setBitIf` xopt Opt_TemplateHaskell flags
1885 .|. qqBit `setBitIf` xopt Opt_QuasiQuotes flags
1886 .|. ipBit `setBitIf` xopt Opt_ImplicitParams flags
1887 .|. explicitForallBit `setBitIf` xopt Opt_ExplicitForAll flags
1888 .|. bangPatBit `setBitIf` xopt Opt_BangPatterns flags
1889 .|. tyFamBit `setBitIf` xopt Opt_TypeFamilies flags
1890 .|. haddockBit `setBitIf` dopt Opt_Haddock flags
1891 .|. magicHashBit `setBitIf` xopt Opt_MagicHash flags
1892 .|. kindSigsBit `setBitIf` xopt Opt_KindSignatures flags
1893 .|. recursiveDoBit `setBitIf` xopt Opt_RecursiveDo flags
1894 .|. recBit `setBitIf` xopt Opt_DoRec flags
1895 .|. recBit `setBitIf` xopt Opt_Arrows flags
1896 .|. unicodeSyntaxBit `setBitIf` xopt Opt_UnicodeSyntax flags
1897 .|. unboxedTuplesBit `setBitIf` xopt Opt_UnboxedTuples flags
1898 .|. datatypeContextsBit `setBitIf` xopt Opt_DatatypeContexts flags
1899 .|. transformComprehensionsBit `setBitIf` xopt Opt_TransformListComp flags
1900 .|. transformComprehensionsBit `setBitIf` xopt Opt_MonadComprehensions flags
1901 .|. rawTokenStreamBit `setBitIf` dopt Opt_KeepRawTokenStream flags
1902 .|. alternativeLayoutRuleBit `setBitIf` xopt Opt_AlternativeLayoutRule flags
1903 .|. relaxedLayoutBit `setBitIf` xopt Opt_RelaxedLayout flags
1904 .|. nondecreasingIndentationBit `setBitIf` xopt Opt_NondecreasingIndentation flags
1906 setBitIf :: Int -> Bool -> Int
1907 b `setBitIf` cond | cond = bit b
1910 addWarning :: DynFlag -> SrcSpan -> SDoc -> P ()
1911 addWarning option srcspan warning
1912 = P $ \s@PState{messages=(ws,es), dflags=d} ->
1913 let warning' = mkWarnMsg srcspan alwaysQualify warning
1914 ws' = if dopt option d then ws `snocBag` warning' else ws
1915 in POk s{messages=(ws', es)} ()
1917 getMessages :: PState -> Messages
1918 getMessages PState{messages=ms} = ms
1920 getContext :: P [LayoutContext]
1921 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1923 setContext :: [LayoutContext] -> P ()
1924 setContext ctx = P $ \s -> POk s{context=ctx} ()
1927 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1928 last_len = len, last_loc = last_loc }) ->
1930 (_:tl) -> POk s{ context = tl } ()
1931 [] -> PFailed (RealSrcSpan last_loc) (srcParseErr buf len)
1933 -- Push a new layout context at the indentation of the last token read.
1934 -- This is only used at the outer level of a module when the 'module'
1935 -- keyword is missing.
1936 pushCurrentContext :: P ()
1937 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1938 POk s{context = Layout (srcSpanStartCol loc) : ctx} ()
1940 getOffside :: P Ordering
1941 getOffside = P $ \s@PState{last_loc=loc, context=stk} ->
1942 let offs = srcSpanStartCol loc in
1943 let ord = case stk of
1944 (Layout n:_) -> --trace ("layout: " ++ show n ++ ", offs: " ++ show offs) $
1949 -- ---------------------------------------------------------------------------
1950 -- Construct a parse error
1953 :: StringBuffer -- current buffer (placed just after the last token)
1954 -> Int -- length of the previous token
1957 = hcat [ if null token
1958 then ptext (sLit "parse error (possibly incorrect indentation)")
1959 else hcat [ptext (sLit "parse error on input "),
1960 char '`', text token, char '\'']
1962 where token = lexemeToString (offsetBytes (-len) buf) len
1964 -- Report a parse failure, giving the span of the previous token as
1965 -- the location of the error. This is the entry point for errors
1966 -- detected during parsing.
1968 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1969 last_loc = last_loc } ->
1970 PFailed (RealSrcSpan last_loc) (srcParseErr buf len)
1972 -- A lexical error is reported at a particular position in the source file,
1973 -- not over a token range.
1974 lexError :: String -> P a
1977 (AI end buf) <- getInput
1978 reportLexError loc end buf str
1980 -- -----------------------------------------------------------------------------
1981 -- This is the top-level function: called from the parser each time a
1982 -- new token is to be read from the input.
1984 lexer :: (Located Token -> P a) -> P a
1986 alr <- extension alternativeLayoutRule
1987 let lexTokenFun = if alr then lexTokenAlr else lexToken
1988 (L span tok) <- lexTokenFun
1989 --trace ("token: " ++ show tok) $ do
1990 cont (L (RealSrcSpan span) tok)
1992 lexTokenAlr :: P (RealLocated Token)
1993 lexTokenAlr = do mPending <- popPendingImplicitToken
1994 t <- case mPending of
1996 do mNext <- popNextToken
1999 Just next -> return next
2000 alternativeLayoutRuleToken t
2003 setAlrLastLoc (getLoc t)
2005 ITwhere -> setAlrExpectingOCurly (Just ALRLayoutWhere)
2006 ITlet -> setAlrExpectingOCurly (Just ALRLayoutLet)
2007 ITof -> setAlrExpectingOCurly (Just ALRLayoutOf)
2008 ITdo -> setAlrExpectingOCurly (Just ALRLayoutDo)
2009 ITmdo -> setAlrExpectingOCurly (Just ALRLayoutDo)
2010 ITrec -> setAlrExpectingOCurly (Just ALRLayoutDo)
2014 alternativeLayoutRuleToken :: RealLocated Token -> P (RealLocated Token)
2015 alternativeLayoutRuleToken t
2016 = do context <- getALRContext
2017 lastLoc <- getAlrLastLoc
2018 mExpectingOCurly <- getAlrExpectingOCurly
2019 justClosedExplicitLetBlock <- getJustClosedExplicitLetBlock
2020 setJustClosedExplicitLetBlock False
2021 dflags <- getDynFlags
2022 let transitional = xopt Opt_AlternativeLayoutRuleTransitional dflags
2024 thisCol = srcSpanStartCol thisLoc
2025 newLine = srcSpanStartLine thisLoc > srcSpanEndLine lastLoc
2026 case (unLoc t, context, mExpectingOCurly) of
2027 -- This case handles a GHC extension to the original H98
2029 (ITocurly, _, Just alrLayout) ->
2030 do setAlrExpectingOCurly Nothing
2031 let isLet = case alrLayout of
2032 ALRLayoutLet -> True
2034 setALRContext (ALRNoLayout (containsCommas ITocurly) isLet : context)
2036 -- ...and makes this case unnecessary
2038 -- I think our implicit open-curly handling is slightly
2039 -- different to John's, in how it interacts with newlines
2041 (ITocurly, _, Just _) ->
2042 do setAlrExpectingOCurly Nothing
2046 (_, ALRLayout _ col : ls, Just expectingOCurly)
2047 | (thisCol > col) ||
2049 isNonDecreasingIntentation expectingOCurly) ->
2050 do setAlrExpectingOCurly Nothing
2051 setALRContext (ALRLayout expectingOCurly thisCol : context)
2053 return (L thisLoc ITocurly)
2055 do setAlrExpectingOCurly Nothing
2056 setPendingImplicitTokens [L lastLoc ITccurly]
2058 return (L lastLoc ITocurly)
2059 (_, _, Just expectingOCurly) ->
2060 do setAlrExpectingOCurly Nothing
2061 setALRContext (ALRLayout expectingOCurly thisCol : context)
2063 return (L thisLoc ITocurly)
2064 -- We do the [] cases earlier than in the spec, as we
2065 -- have an actual EOF token
2066 (ITeof, ALRLayout _ _ : ls, _) ->
2069 return (L thisLoc ITccurly)
2072 -- the other ITeof case omitted; general case below covers it
2074 | justClosedExplicitLetBlock ->
2076 (ITin, ALRLayout ALRLayoutLet _ : ls, _)
2078 do setPendingImplicitTokens [t]
2080 return (L thisLoc ITccurly)
2081 -- This next case is to handle a transitional issue:
2082 (ITwhere, ALRLayout _ col : ls, _)
2083 | newLine && thisCol == col && transitional ->
2084 do addWarning Opt_WarnAlternativeLayoutRuleTransitional
2085 (RealSrcSpan thisLoc)
2086 (transitionalAlternativeLayoutWarning
2087 "`where' clause at the same depth as implicit layout block")
2090 -- Note that we use lastLoc, as we may need to close
2091 -- more layouts, or give a semicolon
2092 return (L lastLoc ITccurly)
2093 -- This next case is to handle a transitional issue:
2094 (ITvbar, ALRLayout _ col : ls, _)
2095 | newLine && thisCol == col && transitional ->
2096 do addWarning Opt_WarnAlternativeLayoutRuleTransitional
2097 (RealSrcSpan thisLoc)
2098 (transitionalAlternativeLayoutWarning
2099 "`|' at the same depth as implicit layout block")
2102 -- Note that we use lastLoc, as we may need to close
2103 -- more layouts, or give a semicolon
2104 return (L lastLoc ITccurly)
2105 (_, ALRLayout _ col : ls, _)
2106 | newLine && thisCol == col ->
2108 return (L thisLoc ITsemi)
2109 | newLine && thisCol < col ->
2112 -- Note that we use lastLoc, as we may need to close
2113 -- more layouts, or give a semicolon
2114 return (L lastLoc ITccurly)
2115 -- We need to handle close before open, as 'then' is both
2116 -- an open and a close
2120 ALRLayout _ _ : ls ->
2123 return (L thisLoc ITccurly)
2124 ALRNoLayout _ isLet : ls ->
2125 do let ls' = if isALRopen u
2126 then ALRNoLayout (containsCommas u) False : ls
2129 when isLet $ setJustClosedExplicitLetBlock True
2132 do let ls = if isALRopen u
2133 then [ALRNoLayout (containsCommas u) False]
2136 -- XXX This is an error in John's code, but
2137 -- it looks reachable to me at first glance
2141 do setALRContext (ALRNoLayout (containsCommas u) False : context)
2143 (ITin, ALRLayout ALRLayoutLet _ : ls, _) ->
2145 setPendingImplicitTokens [t]
2146 return (L thisLoc ITccurly)
2147 (ITin, ALRLayout _ _ : ls, _) ->
2150 return (L thisLoc ITccurly)
2151 -- the other ITin case omitted; general case below covers it
2152 (ITcomma, ALRLayout _ _ : ls, _)
2153 | topNoLayoutContainsCommas ls ->
2156 return (L thisLoc ITccurly)
2157 (ITwhere, ALRLayout ALRLayoutDo _ : ls, _) ->
2159 setPendingImplicitTokens [t]
2160 return (L thisLoc ITccurly)
2161 -- the other ITwhere case omitted; general case below covers it
2162 (_, _, _) -> return t
2164 transitionalAlternativeLayoutWarning :: String -> SDoc
2165 transitionalAlternativeLayoutWarning msg
2166 = text "transitional layout will not be accepted in the future:"
2169 isALRopen :: Token -> Bool
2170 isALRopen ITcase = True
2171 isALRopen ITif = True
2172 isALRopen ITthen = True
2173 isALRopen IToparen = True
2174 isALRopen ITobrack = True
2175 isALRopen ITocurly = True
2177 isALRopen IToubxparen = True
2178 isALRopen ITparenEscape = True
2181 isALRclose :: Token -> Bool
2182 isALRclose ITof = True
2183 isALRclose ITthen = True
2184 isALRclose ITelse = True
2185 isALRclose ITcparen = True
2186 isALRclose ITcbrack = True
2187 isALRclose ITccurly = True
2189 isALRclose ITcubxparen = True
2190 isALRclose _ = False
2192 isNonDecreasingIntentation :: ALRLayout -> Bool
2193 isNonDecreasingIntentation ALRLayoutDo = True
2194 isNonDecreasingIntentation _ = False
2196 containsCommas :: Token -> Bool
2197 containsCommas IToparen = True
2198 containsCommas ITobrack = True
2199 -- John doesn't have {} as containing commas, but records contain them,
2200 -- which caused a problem parsing Cabal's Distribution.Simple.InstallDirs
2201 -- (defaultInstallDirs).
2202 containsCommas ITocurly = True
2204 containsCommas IToubxparen = True
2205 containsCommas _ = False
2207 topNoLayoutContainsCommas :: [ALRContext] -> Bool
2208 topNoLayoutContainsCommas [] = False
2209 topNoLayoutContainsCommas (ALRLayout _ _ : ls) = topNoLayoutContainsCommas ls
2210 topNoLayoutContainsCommas (ALRNoLayout b _ : _) = b
2212 lexToken :: P (RealLocated Token)
2214 inp@(AI loc1 buf) <- getInput
2217 case alexScanUser exts inp sc of
2219 let span = mkRealSrcSpan loc1 loc1
2221 return (L span ITeof)
2222 AlexError (AI loc2 buf) ->
2223 reportLexError loc1 loc2 buf "lexical error"
2224 AlexSkip inp2 _ -> do
2227 AlexToken inp2@(AI end buf2) _ t -> do
2229 let span = mkRealSrcSpan loc1 end
2230 let bytes = byteDiff buf buf2
2231 span `seq` setLastToken span bytes
2234 reportLexError :: RealSrcLoc -> RealSrcLoc -> StringBuffer -> [Char] -> P a
2235 reportLexError loc1 loc2 buf str
2236 | atEnd buf = failLocMsgP loc1 loc2 (str ++ " at end of input")
2239 c = fst (nextChar buf)
2241 if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar#
2242 then failLocMsgP loc2 loc2 (str ++ " (UTF-8 decoding error)")
2243 else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c)
2245 lexTokenStream :: StringBuffer -> RealSrcLoc -> DynFlags -> ParseResult [Located Token]
2246 lexTokenStream buf loc dflags = unP go initState
2247 where dflags' = dopt_set (dopt_unset dflags Opt_Haddock) Opt_KeepRawTokenStream
2248 initState = mkPState dflags' buf loc
2250 ltok <- lexer return
2252 L _ ITeof -> return []
2253 _ -> liftM (ltok:) go
2255 linePrags = Map.singleton "line" (begin line_prag2)
2257 fileHeaderPrags = Map.fromList([("options", lex_string_prag IToptions_prag),
2258 ("options_ghc", lex_string_prag IToptions_prag),
2259 ("options_haddock", lex_string_prag ITdocOptions),
2260 ("language", token ITlanguage_prag),
2261 ("include", lex_string_prag ITinclude_prag)])
2263 ignoredPrags = Map.fromList (map ignored pragmas)
2264 where ignored opt = (opt, nested_comment lexToken)
2265 impls = ["hugs", "nhc98", "jhc", "yhc", "catch", "derive"]
2266 options_pragmas = map ("options_" ++) impls
2267 -- CFILES is a hugs-only thing.
2268 pragmas = options_pragmas ++ ["cfiles", "contract"]
2270 oneWordPrags = Map.fromList([("rules", rulePrag),
2271 ("inline", token (ITinline_prag Inline FunLike)),
2272 ("inlinable", token (ITinline_prag Inlinable FunLike)),
2273 ("inlineable", token (ITinline_prag Inlinable FunLike)),
2275 ("notinline", token (ITinline_prag NoInline FunLike)),
2276 ("specialize", token ITspec_prag),
2277 ("source", token ITsource_prag),
2278 ("warning", token ITwarning_prag),
2279 ("deprecated", token ITdeprecated_prag),
2280 ("scc", token ITscc_prag),
2281 ("generated", token ITgenerated_prag),
2282 ("core", token ITcore_prag),
2283 ("unpack", token ITunpack_prag),
2284 ("ann", token ITann_prag),
2285 ("vectorize", token ITvect_prag),
2286 ("novectorize", token ITnovect_prag)])
2288 twoWordPrags = Map.fromList([("inline conlike", token (ITinline_prag Inline ConLike)),
2289 ("notinline conlike", token (ITinline_prag NoInline ConLike)),
2290 ("specialize inline", token (ITspec_inline_prag True)),
2291 ("specialize notinline", token (ITspec_inline_prag False)),
2292 ("vectorize scalar", token ITvect_scalar_prag)])
2294 dispatch_pragmas :: Map String Action -> Action
2295 dispatch_pragmas prags span buf len = case Map.lookup (clean_pragma (lexemeToString buf len)) prags of
2296 Just found -> found span buf len
2297 Nothing -> lexError "unknown pragma"
2299 known_pragma :: Map String Action -> AlexAccPred Int
2300 known_pragma prags _ _ len (AI _ buf) = (isJust $ Map.lookup (clean_pragma (lexemeToString (offsetBytes (- len) buf) len)) prags)
2301 && (nextCharIs buf (\c -> not (isAlphaNum c || c == '_')))
2303 clean_pragma :: String -> String
2304 clean_pragma prag = canon_ws (map toLower (unprefix prag))
2305 where unprefix prag' = case stripPrefix "{-#" prag' of
2308 canonical prag' = case prag' of
2309 "noinline" -> "notinline"
2310 "specialise" -> "specialize"
2311 "vectorise" -> "vectorize"
2312 "novectorise" -> "novectorize"
2313 "constructorlike" -> "conlike"
2315 canon_ws s = unwords (map canonical (words s))