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 {-# LANGUAGE BangPatterns #-}
36 {-# OPTIONS_GHC -fno-warn-unused-matches #-}
37 {-# OPTIONS_GHC -fno-warn-unused-binds #-}
38 {-# OPTIONS_GHC -fno-warn-unused-imports #-}
39 {-# OPTIONS_GHC -fno-warn-missing-signatures #-}
40 -- But alex still generates some code that causes the "lazy unlifted bindings"
41 -- warning, and old compilers don't know about it so we can't easily turn
42 -- it off, so for now we use the sledge hammer:
43 {-# OPTIONS_GHC -w #-}
45 {-# OPTIONS_GHC -funbox-strict-fields #-}
48 Token(..), lexer, pragState, mkPState, PState(..),
49 P(..), ParseResult(..), getSrcLoc,
50 getPState, getDynFlags, withThisPackage,
51 failLocMsgP, failSpanMsgP, srcParseFail,
53 popContext, pushCurrentContext, setLastToken, setSrcLoc,
54 getLexState, popLexState, pushLexState,
55 extension, bangPatEnabled, datatypeContextsEnabled,
70 import BasicTypes ( InlineSpec(..), RuleMatchInfo(..) )
71 import Util ( readRational )
79 import qualified Data.Map as Map
83 $unispace = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
84 $whitechar = [\ \n\r\f\v $unispace]
85 $white_no_nl = $whitechar # \n
89 $unidigit = \x03 -- Trick Alex into handling Unicode. See alexGetChar.
90 $decdigit = $ascdigit -- for now, should really be $digit (ToDo)
91 $digit = [$ascdigit $unidigit]
93 $special = [\(\)\,\;\[\]\`\{\}]
94 $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~]
95 $unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar.
96 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\']
98 $unilarge = \x01 -- Trick Alex into handling Unicode. See alexGetChar.
100 $large = [$asclarge $unilarge]
102 $unismall = \x02 -- Trick Alex into handling Unicode. See alexGetChar.
104 $small = [$ascsmall $unismall \_]
106 $unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar.
107 $graphic = [$small $large $symbol $digit $special $unigraphic \:\"\']
110 $hexit = [$decdigit A-F a-f]
111 $symchar = [$symbol \:]
113 $idchar = [$small $large $digit \']
115 $pragmachar = [$small $large $digit]
117 $docsym = [\| \^ \* \$]
119 @varid = $small $idchar*
120 @conid = $large $idchar*
122 @varsym = $symbol $symchar*
123 @consym = \: $symchar*
125 @decimal = $decdigit+
127 @hexadecimal = $hexit+
128 @exponent = [eE] [\-\+]? @decimal
130 -- we support the hierarchical module name extension:
133 @floating_point = @decimal \. @decimal @exponent? | @decimal @exponent
135 -- normal signed numerical literals can only be explicitly negative,
136 -- not explicitly positive (contrast @exponent)
138 @signed = @negative ?
142 -- everywhere: skip whitespace and comments
144 $tab+ { warn Opt_WarnTabs (text "Warning: Tab character") }
146 -- Everywhere: deal with nested comments. We explicitly rule out
147 -- pragmas, "{-#", so that we don't accidentally treat them as comments.
148 -- (this can happen even though pragmas will normally take precedence due to
149 -- longest-match, because pragmas aren't valid in every state, but comments
150 -- are). We also rule out nested Haddock comments, if the -haddock flag is
153 "{-" / { isNormalComment } { nested_comment lexToken }
155 -- Single-line comments are a bit tricky. Haskell 98 says that two or
156 -- more dashes followed by a symbol should be parsed as a varsym, so we
157 -- have to exclude those.
159 -- Since Haddock comments aren't valid in every state, we need to rule them
162 -- The following two rules match comments that begin with two dashes, but
163 -- continue with a different character. The rules test that this character
164 -- is not a symbol (in which case we'd have a varsym), and that it's not a
165 -- space followed by a Haddock comment symbol (docsym) (in which case we'd
166 -- have a Haddock comment). The rules then munch the rest of the line.
168 "-- " ~[$docsym \#] .* { lineCommentToken }
169 "--" [^$symbol : \ ] .* { lineCommentToken }
171 -- Next, match Haddock comments if no -haddock flag
173 "-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken }
175 -- Now, when we've matched comments that begin with 2 dashes and continue
176 -- with a different character, we need to match comments that begin with three
177 -- or more dashes (which clearly can't be Haddock comments). We only need to
178 -- make sure that the first non-dash character isn't a symbol, and munch the
181 "---"\-* [^$symbol :] .* { lineCommentToken }
183 -- Since the previous rules all match dashes followed by at least one
184 -- character, we also need to match a whole line filled with just dashes.
186 "--"\-* / { atEOL } { lineCommentToken }
188 -- We need this rule since none of the other single line comment rules
189 -- actually match this case.
191 "-- " / { atEOL } { lineCommentToken }
193 -- 'bol' state: beginning of a line. Slurp up all the whitespace (including
194 -- blank lines) until we find a non-whitespace character, then do layout
197 -- One slight wibble here: what if the line begins with {-#? In
198 -- theory, we have to lex the pragma to see if it's one we recognise,
199 -- and if it is, then we backtrack and do_bol, otherwise we treat it
200 -- as a nested comment. We don't bother with this: if the line begins
201 -- with {-#, then we'll assume it's a pragma we know about and go for do_bol.
204 ^\# (line)? { begin line_prag1 }
205 ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently
206 ^\# \! .* \n ; -- #!, for scripts
210 -- after a layout keyword (let, where, do, of), we begin a new layout
211 -- context if the curly brace is missing.
212 -- Careful! This stuff is quite delicate.
213 <layout, layout_do> {
214 \{ / { notFollowedBy '-' } { pop_and open_brace }
215 -- we might encounter {-# here, but {- has been handled already
217 ^\# (line)? { begin line_prag1 }
220 -- do is treated in a subtly different way, see new_layout_context
221 <layout> () { new_layout_context True }
222 <layout_do> () { new_layout_context False }
224 -- after a new layout context which was found to be to the left of the
225 -- previous context, we have generated a '{' token, and we now need to
226 -- generate a matching '}' token.
227 <layout_left> () { do_layout_left }
229 <0,option_prags> \n { begin bol }
231 "{-#" $whitechar* $pragmachar+ / { known_pragma linePrags }
232 { dispatch_pragmas linePrags }
234 -- single-line line pragmas, of the form
235 -- # <line> "<file>" <extra-stuff> \n
236 <line_prag1> $decdigit+ { setLine line_prag1a }
237 <line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
238 <line_prag1b> .* { pop }
240 -- Haskell-style line pragmas, of the form
241 -- {-# LINE <line> "<file>" #-}
242 <line_prag2> $decdigit+ { setLine line_prag2a }
243 <line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
244 <line_prag2b> "#-}"|"-}" { pop }
245 -- NOTE: accept -} at the end of a LINE pragma, for compatibility
246 -- with older versions of GHC which generated these.
249 "{-#" $whitechar* $pragmachar+
250 $whitechar+ $pragmachar+ / { known_pragma twoWordPrags }
251 { dispatch_pragmas twoWordPrags }
253 "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags }
254 { dispatch_pragmas oneWordPrags }
256 -- We ignore all these pragmas, but don't generate a warning for them
257 "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags }
258 { dispatch_pragmas ignoredPrags }
260 -- ToDo: should only be valid inside a pragma:
265 "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
266 { dispatch_pragmas fileHeaderPrags }
268 "-- #" { multiline_doc_comment }
272 -- In the "0" mode we ignore these pragmas
273 "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags }
274 { nested_comment lexToken }
278 "-- #" .* { lineCommentToken }
282 "{-#" { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma")
283 (nested_comment lexToken) }
286 -- '0' state: ordinary lexemes
291 "-- " $docsym / { ifExtension haddockEnabled } { multiline_doc_comment }
292 "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment }
298 "[:" / { ifExtension parrEnabled } { token ITopabrack }
299 ":]" / { ifExtension parrEnabled } { token ITcpabrack }
303 "[|" / { ifExtension thEnabled } { token ITopenExpQuote }
304 "[e|" / { ifExtension thEnabled } { token ITopenExpQuote }
305 "[p|" / { ifExtension thEnabled } { token ITopenPatQuote }
306 "[d|" / { ifExtension thEnabled } { layout_token ITopenDecQuote }
307 "[t|" / { ifExtension thEnabled } { token ITopenTypQuote }
308 "|]" / { ifExtension thEnabled } { token ITcloseQuote }
309 \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape }
310 "$(" / { ifExtension thEnabled } { token ITparenEscape }
312 -- For backward compatibility, accept the old dollar syntax
313 "[$" @varid "|" / { ifExtension qqEnabled }
314 { lex_quasiquote_tok }
316 "[" @varid "|" / { ifExtension qqEnabled }
317 { lex_quasiquote_tok }
321 "(|" / { ifExtension arrowsEnabled `alexAndPred` notFollowedBySymbol }
322 { special IToparenbar }
323 "|)" / { ifExtension arrowsEnabled } { special ITcparenbar }
327 \? @varid / { ifExtension ipEnabled } { skip_one_varid ITdupipvarid }
331 "(#" / { ifExtension unboxedTuplesEnabled `alexAndPred` notFollowedBySymbol }
332 { token IToubxparen }
333 "#)" / { ifExtension unboxedTuplesEnabled }
334 { token ITcubxparen }
338 "{|" / { ifExtension genericsEnabled } { token ITocurlybar }
339 "|}" / { ifExtension genericsEnabled } { token ITccurlybar }
343 \( { special IToparen }
344 \) { special ITcparen }
345 \[ { special ITobrack }
346 \] { special ITcbrack }
347 \, { special ITcomma }
348 \; { special ITsemi }
349 \` { special ITbackquote }
356 @qual @varid { idtoken qvarid }
357 @qual @conid { idtoken qconid }
359 @conid { idtoken conid }
363 @qual @varid "#"+ / { ifExtension magicHashEnabled } { idtoken qvarid }
364 @qual @conid "#"+ / { ifExtension magicHashEnabled } { idtoken qconid }
365 @varid "#"+ / { ifExtension magicHashEnabled } { varid }
366 @conid "#"+ / { ifExtension magicHashEnabled } { idtoken conid }
369 -- ToDo: - move `var` and (sym) into lexical syntax?
370 -- - remove backquote from $special?
372 @qual @varsym / { ifExtension oldQualOps } { idtoken qvarsym }
373 @qual @consym / { ifExtension oldQualOps } { idtoken qconsym }
374 @qual \( @varsym \) / { ifExtension newQualOps } { idtoken prefixqvarsym }
375 @qual \( @consym \) / { ifExtension newQualOps } { idtoken prefixqconsym }
380 -- For the normal boxed literals we need to be careful
381 -- when trying to be close to Haskell98
383 -- Normal integral literals (:: Num a => a, from Integer)
384 @decimal { tok_num positive 0 0 decimal }
385 0[oO] @octal { tok_num positive 2 2 octal }
386 0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal }
388 -- Normal rational literals (:: Fractional a => a, from Rational)
389 @floating_point { strtoken tok_float }
393 -- Unboxed ints (:: Int#) and words (:: Word#)
394 -- It's simpler (and faster?) to give separate cases to the negatives,
395 -- especially considering octal/hexadecimal prefixes.
396 @decimal \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal }
397 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal }
398 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal }
399 @negative @decimal \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal }
400 @negative 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal }
401 @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal }
403 @decimal \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal }
404 0[oO] @octal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal }
405 0[xX] @hexadecimal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal }
407 -- Unboxed floats and doubles (:: Float#, :: Double#)
408 -- prim_{float,double} work with signed literals
409 @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat }
410 @signed @floating_point \# \# / { ifExtension magicHashEnabled } { init_strtoken 2 tok_primdouble }
413 -- Strings and chars are lexed by hand-written code. The reason is
414 -- that even if we recognise the string or char here in the regex
415 -- lexer, we would still have to parse the string afterward in order
416 -- to convert it to a String.
419 \" { lex_string_tok }
423 -- -----------------------------------------------------------------------------
427 = ITas -- Haskell keywords
451 | ITscc -- ToDo: remove (we use {-# SCC "..." #-} now)
453 | ITforall -- GHC extension keywords
472 | ITinline_prag InlineSpec RuleMatchInfo
473 | ITspec_prag -- SPECIALISE
474 | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE)
482 | ITcore_prag -- hdaume: core annotations
486 | IToptions_prag String
487 | ITinclude_prag String
490 | ITdotdot -- reserved symbols
506 | ITbiglam -- GHC-extension symbols
508 | ITocurly -- special symbols
510 | ITocurlybar -- {|, for type applications
511 | ITccurlybar -- |}, for type applications
515 | ITopabrack -- [:, for parallel arrays with -XParr
516 | ITcpabrack -- :], for parallel arrays with -XParr
527 | ITvarid FastString -- identifiers
529 | ITvarsym FastString
530 | ITconsym FastString
531 | ITqvarid (FastString,FastString)
532 | ITqconid (FastString,FastString)
533 | ITqvarsym (FastString,FastString)
534 | ITqconsym (FastString,FastString)
535 | ITprefixqvarsym (FastString,FastString)
536 | ITprefixqconsym (FastString,FastString)
538 | ITdupipvarid FastString -- GHC extension: implicit param: ?x
541 | ITstring FastString
543 | ITrational Rational
546 | ITprimstring FastString
549 | ITprimfloat Rational
550 | ITprimdouble Rational
552 -- Template Haskell extension tokens
553 | ITopenExpQuote -- [| or [e|
554 | ITopenPatQuote -- [p|
555 | ITopenDecQuote -- [d|
556 | ITopenTypQuote -- [t|
558 | ITidEscape FastString -- $x
559 | ITparenEscape -- $(
562 | ITquasiQuote (FastString,FastString,SrcSpan) -- [:...|...|]
564 -- Arrow notation extension
571 | ITLarrowtail -- -<<
572 | ITRarrowtail -- >>-
574 | ITunknown String -- Used when the lexer can't make sense of it
575 | ITeof -- end of file token
577 -- Documentation annotations
578 | ITdocCommentNext String -- something beginning '-- |'
579 | ITdocCommentPrev String -- something beginning '-- ^'
580 | ITdocCommentNamed String -- something beginning '-- $'
581 | ITdocSection Int String -- a section heading
582 | ITdocOptions String -- doc options (prune, ignore-exports, etc)
583 | ITdocOptionsOld String -- doc options declared "-- # ..."-style
584 | ITlineComment String -- comment starting by "--"
585 | ITblockComment String -- comment in {- -}
588 deriving Show -- debugging
592 isSpecial :: Token -> Bool
593 -- If we see M.x, where x is a keyword, but
594 -- is special, we treat is as just plain M.x,
596 isSpecial ITas = True
597 isSpecial IThiding = True
598 isSpecial ITqualified = True
599 isSpecial ITforall = True
600 isSpecial ITexport = True
601 isSpecial ITlabel = True
602 isSpecial ITdynamic = True
603 isSpecial ITsafe = True
604 isSpecial ITthreadsafe = True
605 isSpecial ITinterruptible = True
606 isSpecial ITunsafe = True
607 isSpecial ITccallconv = True
608 isSpecial ITstdcallconv = True
609 isSpecial ITprimcallconv = True
610 isSpecial ITmdo = True
611 isSpecial ITfamily = True
612 isSpecial ITgroup = True
613 isSpecial ITby = True
614 isSpecial ITusing = True
618 -- the bitmap provided as the third component indicates whether the
619 -- corresponding extension keyword is valid under the extension options
620 -- provided to the compiler; if the extension corresponding to *any* of the
621 -- bits set in the bitmap is enabled, the keyword is valid (this setup
622 -- facilitates using a keyword in two different extensions that can be
623 -- activated independently)
625 reservedWordsFM :: UniqFM (Token, Int)
626 reservedWordsFM = listToUFM $
627 map (\(x, y, z) -> (mkFastString x, (y, z)))
628 [( "_", ITunderscore, 0 ),
630 ( "case", ITcase, 0 ),
631 ( "class", ITclass, 0 ),
632 ( "data", ITdata, 0 ),
633 ( "default", ITdefault, 0 ),
634 ( "deriving", ITderiving, 0 ),
636 ( "else", ITelse, 0 ),
637 ( "hiding", IThiding, 0 ),
639 ( "import", ITimport, 0 ),
641 ( "infix", ITinfix, 0 ),
642 ( "infixl", ITinfixl, 0 ),
643 ( "infixr", ITinfixr, 0 ),
644 ( "instance", ITinstance, 0 ),
646 ( "module", ITmodule, 0 ),
647 ( "newtype", ITnewtype, 0 ),
649 ( "qualified", ITqualified, 0 ),
650 ( "then", ITthen, 0 ),
651 ( "type", ITtype, 0 ),
652 ( "where", ITwhere, 0 ),
653 ( "_scc_", ITscc, 0 ), -- ToDo: remove
655 ( "forall", ITforall, bit explicitForallBit .|. bit inRulePragBit),
656 ( "mdo", ITmdo, bit recursiveDoBit),
657 ( "family", ITfamily, bit tyFamBit),
658 ( "group", ITgroup, bit transformComprehensionsBit),
659 ( "by", ITby, bit transformComprehensionsBit),
660 ( "using", ITusing, bit transformComprehensionsBit),
662 ( "foreign", ITforeign, bit ffiBit),
663 ( "export", ITexport, bit ffiBit),
664 ( "label", ITlabel, bit ffiBit),
665 ( "dynamic", ITdynamic, bit ffiBit),
666 ( "safe", ITsafe, bit ffiBit),
667 ( "threadsafe", ITthreadsafe, bit ffiBit), -- ToDo: remove
668 ( "interruptible", ITinterruptible, bit ffiBit),
669 ( "unsafe", ITunsafe, bit ffiBit),
670 ( "stdcall", ITstdcallconv, bit ffiBit),
671 ( "ccall", ITccallconv, bit ffiBit),
672 ( "prim", ITprimcallconv, bit ffiBit),
674 ( "rec", ITrec, bit recBit),
675 ( "proc", ITproc, bit arrowsBit)
678 reservedSymsFM :: UniqFM (Token, Int -> Bool)
679 reservedSymsFM = listToUFM $
680 map (\ (x,y,z) -> (mkFastString x,(y,z)))
681 [ ("..", ITdotdot, always)
682 -- (:) is a reserved op, meaning only list cons
683 ,(":", ITcolon, always)
684 ,("::", ITdcolon, always)
685 ,("=", ITequal, always)
686 ,("\\", ITlam, always)
687 ,("|", ITvbar, always)
688 ,("<-", ITlarrow, always)
689 ,("->", ITrarrow, always)
691 ,("~", ITtilde, always)
692 ,("=>", ITdarrow, always)
693 ,("-", ITminus, always)
694 ,("!", ITbang, always)
696 -- For data T (a::*) = MkT
697 ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i)
698 -- For 'forall a . t'
699 ,(".", ITdot, always) -- \i -> explicitForallEnabled i || inRulePrag i)
701 ,("-<", ITlarrowtail, arrowsEnabled)
702 ,(">-", ITrarrowtail, arrowsEnabled)
703 ,("-<<", ITLarrowtail, arrowsEnabled)
704 ,(">>-", ITRarrowtail, arrowsEnabled)
706 ,("∷", ITdcolon, unicodeSyntaxEnabled)
707 ,("⇒", ITdarrow, unicodeSyntaxEnabled)
708 ,("∀", ITforall, \i -> unicodeSyntaxEnabled i &&
709 explicitForallEnabled i)
710 ,("→", ITrarrow, unicodeSyntaxEnabled)
711 ,("←", ITlarrow, unicodeSyntaxEnabled)
713 ,("⤙", ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
714 ,("⤚", ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
715 ,("⤛", ITLarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
716 ,("⤜", ITRarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i)
718 ,("★", ITstar, unicodeSyntaxEnabled)
720 -- ToDo: ideally, → and ∷ should be "specials", so that they cannot
721 -- form part of a large operator. This would let us have a better
722 -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe).
725 -- -----------------------------------------------------------------------------
728 type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token)
730 special :: Token -> Action
731 special tok span _buf _len = return (L span tok)
733 token, layout_token :: Token -> Action
734 token t span _buf _len = return (L span t)
735 layout_token t span _buf _len = pushLexState layout >> return (L span t)
737 idtoken :: (StringBuffer -> Int -> Token) -> Action
738 idtoken f span buf len = return (L span $! (f buf len))
740 skip_one_varid :: (FastString -> Token) -> Action
741 skip_one_varid f span buf len
742 = return (L span $! f (lexemeToFastString (stepOn buf) (len-1)))
744 strtoken :: (String -> Token) -> Action
745 strtoken f span buf len =
746 return (L span $! (f $! lexemeToString buf len))
748 init_strtoken :: Int -> (String -> Token) -> Action
749 -- like strtoken, but drops the last N character(s)
750 init_strtoken drop f span buf len =
751 return (L span $! (f $! lexemeToString buf (len-drop)))
753 begin :: Int -> Action
754 begin code _span _str _len = do pushLexState code; lexToken
757 pop _span _buf _len = do _ <- popLexState
760 pop_and :: Action -> Action
761 pop_and act span buf len = do _ <- popLexState
764 {-# INLINE nextCharIs #-}
765 nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
766 nextCharIs buf p = not (atEnd buf) && p (currentChar buf)
768 notFollowedBy :: Char -> AlexAccPred Int
769 notFollowedBy char _ _ _ (AI _ buf)
770 = nextCharIs buf (/=char)
772 notFollowedBySymbol :: AlexAccPred Int
773 notFollowedBySymbol _ _ _ (AI _ buf)
774 = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
776 -- We must reject doc comments as being ordinary comments everywhere.
777 -- In some cases the doc comment will be selected as the lexeme due to
778 -- maximal munch, but not always, because the nested comment rule is
779 -- valid in all states, but the doc-comment rules are only valid in
780 -- the non-layout states.
781 isNormalComment :: AlexAccPred Int
782 isNormalComment bits _ _ (AI _ buf)
783 | haddockEnabled bits = notFollowedByDocOrPragma
784 | otherwise = nextCharIs buf (/='#')
786 notFollowedByDocOrPragma
787 = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))
789 spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
790 spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
793 haddockDisabledAnd p bits _ _ (AI _ buf)
794 = if haddockEnabled bits then False else (p buf)
797 atEOL :: AlexAccPred Int
798 atEOL _ _ _ (AI _ buf) = atEnd buf || currentChar buf == '\n'
800 ifExtension :: (Int -> Bool) -> AlexAccPred Int
801 ifExtension pred bits _ _ _ = pred bits
803 multiline_doc_comment :: Action
804 multiline_doc_comment span buf _len = withLexedDocType (worker "")
806 worker commentAcc input docType oneLine = case alexGetChar input of
808 | oneLine -> docCommentEnd input commentAcc docType buf span
809 | otherwise -> case checkIfCommentLine input' of
810 Just input -> worker ('\n':commentAcc) input docType False
811 Nothing -> docCommentEnd input commentAcc docType buf span
812 Just (c, input) -> worker (c:commentAcc) input docType oneLine
813 Nothing -> docCommentEnd input commentAcc docType buf span
815 checkIfCommentLine input = check (dropNonNewlineSpace input)
817 check input = case alexGetChar input of
818 Just ('-', input) -> case alexGetChar input of
819 Just ('-', input) -> case alexGetChar input of
820 Just (c, _) | c /= '-' -> Just input
825 dropNonNewlineSpace input = case alexGetChar input of
827 | isSpace c && c /= '\n' -> dropNonNewlineSpace input'
831 lineCommentToken :: Action
832 lineCommentToken span buf len = do
833 b <- extension rawTokenStreamEnabled
834 if b then strtoken ITlineComment span buf len else lexToken
837 nested comments require traversing by hand, they can't be parsed
838 using regular expressions.
840 nested_comment :: P (Located Token) -> Action
841 nested_comment cont span _str _len = do
845 go commentAcc 0 input = do setInput input
846 b <- extension rawTokenStreamEnabled
848 then docCommentEnd input commentAcc ITblockComment _str span
850 go commentAcc n input = case alexGetChar input of
851 Nothing -> errBrace input span
852 Just ('-',input) -> case alexGetChar input of
853 Nothing -> errBrace input span
854 Just ('\125',input) -> go commentAcc (n-1) input
855 Just (_,_) -> go ('-':commentAcc) n input
856 Just ('\123',input) -> case alexGetChar input of
857 Nothing -> errBrace input span
858 Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input
859 Just (_,_) -> go ('\123':commentAcc) n input
860 Just (c,input) -> go (c:commentAcc) n input
862 nested_doc_comment :: Action
863 nested_doc_comment span buf _len = withLexedDocType (go "")
865 go commentAcc input docType _ = case alexGetChar input of
866 Nothing -> errBrace input span
867 Just ('-',input) -> case alexGetChar input of
868 Nothing -> errBrace input span
869 Just ('\125',input) ->
870 docCommentEnd input commentAcc docType buf span
871 Just (_,_) -> go ('-':commentAcc) input docType False
872 Just ('\123', input) -> case alexGetChar input of
873 Nothing -> errBrace input span
874 Just ('-',input) -> do
876 let cont = do input <- getInput; go commentAcc input docType False
877 nested_comment cont span buf _len
878 Just (_,_) -> go ('\123':commentAcc) input docType False
879 Just (c,input) -> go (c:commentAcc) input docType False
881 withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
883 withLexedDocType lexDocComment = do
884 input@(AI _ buf) <- getInput
885 case prevChar buf ' ' of
886 '|' -> lexDocComment input ITdocCommentNext False
887 '^' -> lexDocComment input ITdocCommentPrev False
888 '$' -> lexDocComment input ITdocCommentNamed False
889 '*' -> lexDocSection 1 input
890 '#' -> lexDocComment input ITdocOptionsOld False
891 _ -> panic "withLexedDocType: Bad doc type"
893 lexDocSection n input = case alexGetChar input of
894 Just ('*', input) -> lexDocSection (n+1) input
895 Just (_, _) -> lexDocComment input (ITdocSection n) True
896 Nothing -> do setInput input; lexToken -- eof reached, lex it normally
898 -- RULES pragmas turn on the forall and '.' keywords, and we turn them
899 -- off again at the end of the pragma.
901 rulePrag span _buf _len = do
902 setExts (.|. bit inRulePragBit)
903 return (L span ITrules_prag)
906 endPrag span _buf _len = do
907 setExts (.&. complement (bit inRulePragBit))
908 return (L span ITclose_prag)
911 -------------------------------------------------------------------------------
912 -- This function is quite tricky. We can't just return a new token, we also
913 -- need to update the state of the parser. Why? Because the token is longer
914 -- than what was lexed by Alex, and the lexToken function doesn't know this, so
915 -- it writes the wrong token length to the parser state. This function is
916 -- called afterwards, so it can just update the state.
918 docCommentEnd :: AlexInput -> String -> (String -> Token) -> StringBuffer ->
919 SrcSpan -> P (Located Token)
920 docCommentEnd input commentAcc docType buf span = do
922 let (AI loc nextBuf) = input
923 comment = reverse commentAcc
924 span' = mkSrcSpan (srcSpanStart span) loc
925 last_len = byteDiff buf nextBuf
927 span `seq` setLastToken span' last_len
928 return (L span' (docType comment))
930 errBrace :: AlexInput -> SrcSpan -> P a
931 errBrace (AI end _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'"
933 open_brace, close_brace :: Action
934 open_brace span _str _len = do
936 setContext (NoLayout:ctx)
937 return (L span ITocurly)
938 close_brace span _str _len = do
940 return (L span ITccurly)
942 qvarid, qconid :: StringBuffer -> Int -> Token
943 qvarid buf len = ITqvarid $! splitQualName buf len False
944 qconid buf len = ITqconid $! splitQualName buf len False
946 splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString)
947 -- takes a StringBuffer and a length, and returns the module name
948 -- and identifier parts of a qualified name. Splits at the *last* dot,
949 -- because of hierarchical module names.
950 splitQualName orig_buf len parens = split orig_buf orig_buf
953 | orig_buf `byteDiff` buf >= len = done dot_buf
954 | c == '.' = found_dot buf'
955 | otherwise = split buf' dot_buf
957 (c,buf') = nextChar buf
959 -- careful, we might get names like M....
960 -- so, if the character after the dot is not upper-case, this is
961 -- the end of the qualifier part.
962 found_dot buf -- buf points after the '.'
963 | isUpper c = split buf' buf
964 | otherwise = done buf
966 (c,buf') = nextChar buf
969 (lexemeToFastString orig_buf (qual_size - 1),
970 if parens -- Prelude.(+)
971 then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2)
972 else lexemeToFastString dot_buf (len - qual_size))
974 qual_size = orig_buf `byteDiff` dot_buf
979 case lookupUFM reservedWordsFM fs of
980 Just (keyword,0) -> do
982 return (L span keyword)
983 Just (keyword,exts) -> do
984 b <- extension (\i -> exts .&. i /= 0)
985 if b then do maybe_layout keyword
986 return (L span keyword)
987 else return (L span (ITvarid fs))
988 _other -> return (L span (ITvarid fs))
990 fs = lexemeToFastString buf len
992 conid :: StringBuffer -> Int -> Token
993 conid buf len = ITconid fs
994 where fs = lexemeToFastString buf len
996 qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token
997 qvarsym buf len = ITqvarsym $! splitQualName buf len False
998 qconsym buf len = ITqconsym $! splitQualName buf len False
999 prefixqvarsym buf len = ITprefixqvarsym $! splitQualName buf len True
1000 prefixqconsym buf len = ITprefixqconsym $! splitQualName buf len True
1002 varsym, consym :: Action
1003 varsym = sym ITvarsym
1004 consym = sym ITconsym
1006 sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int
1007 -> P (Located Token)
1008 sym con span buf len =
1009 case lookupUFM reservedSymsFM fs of
1010 Just (keyword,exts) -> do
1012 if b then return (L span keyword)
1013 else return (L span $! con fs)
1014 _other -> return (L span $! con fs)
1016 fs = lexemeToFastString buf len
1018 -- Variations on the integral numeric literal.
1019 tok_integral :: (Integer -> Token)
1020 -> (Integer -> Integer)
1021 -- -> (StringBuffer -> StringBuffer) -> (Int -> Int)
1023 -> (Integer, (Char->Int)) -> Action
1024 tok_integral itint transint transbuf translen (radix,char_to_int) span buf len =
1025 return $ L span $ itint $! transint $ parseUnsignedInteger
1026 (offsetBytes transbuf buf) (subtract translen len) radix char_to_int
1028 -- some conveniences for use with tok_integral
1029 tok_num :: (Integer -> Integer)
1031 -> (Integer, (Char->Int)) -> Action
1032 tok_num = tok_integral ITinteger
1033 tok_primint :: (Integer -> Integer)
1035 -> (Integer, (Char->Int)) -> Action
1036 tok_primint = tok_integral ITprimint
1037 tok_primword :: Int -> Int
1038 -> (Integer, (Char->Int)) -> Action
1039 tok_primword = tok_integral ITprimword positive
1040 positive, negative :: (Integer -> Integer)
1043 decimal, octal, hexadecimal :: (Integer, Char -> Int)
1044 decimal = (10,octDecDigit)
1045 octal = (8,octDecDigit)
1046 hexadecimal = (16,hexDigit)
1048 -- readRational can understand negative rationals, exponents, everything.
1049 tok_float, tok_primfloat, tok_primdouble :: String -> Token
1050 tok_float str = ITrational $! readRational str
1051 tok_primfloat str = ITprimfloat $! readRational str
1052 tok_primdouble str = ITprimdouble $! readRational str
1054 -- -----------------------------------------------------------------------------
1055 -- Layout processing
1057 -- we're at the first token on a line, insert layout tokens if necessary
1059 do_bol span _str _len = do
1063 --trace "layout: inserting '}'" $ do
1065 -- do NOT pop the lex state, we might have a ';' to insert
1066 return (L span ITvccurly)
1068 --trace "layout: inserting ';'" $ do
1070 return (L span ITsemi)
1075 -- certain keywords put us in the "layout" state, where we might
1076 -- add an opening curly brace.
1077 maybe_layout :: Token -> P ()
1078 maybe_layout t = do -- If the alternative layout rule is enabled then
1079 -- we never create an implicit layout context here.
1080 -- Layout is handled XXX instead.
1081 -- The code for closing implicit contexts, or
1082 -- inserting implicit semi-colons, is therefore
1083 -- irrelevant as it only applies in an implicit
1085 alr <- extension alternativeLayoutRule
1087 where f ITdo = pushLexState layout_do
1088 f ITmdo = pushLexState layout_do
1089 f ITof = pushLexState layout
1090 f ITlet = pushLexState layout
1091 f ITwhere = pushLexState layout
1092 f ITrec = pushLexState layout
1095 -- Pushing a new implicit layout context. If the indentation of the
1096 -- next token is not greater than the previous layout context, then
1097 -- Haskell 98 says that the new layout context should be empty; that is
1098 -- the lexer must generate {}.
1100 -- We are slightly more lenient than this: when the new context is started
1101 -- by a 'do', then we allow the new context to be at the same indentation as
1102 -- the previous context. This is what the 'strict' argument is for.
1104 new_layout_context :: Bool -> Action
1105 new_layout_context strict span _buf _len = do
1107 (AI l _) <- getInput
1108 let offset = srcLocCol l
1111 Layout prev_off : _ |
1112 (strict && prev_off >= offset ||
1113 not strict && prev_off > offset) -> do
1114 -- token is indented to the left of the previous context.
1115 -- we must generate a {} sequence now.
1116 pushLexState layout_left
1117 return (L span ITvocurly)
1119 setContext (Layout offset : ctx)
1120 return (L span ITvocurly)
1122 do_layout_left :: Action
1123 do_layout_left span _buf _len = do
1125 pushLexState bol -- we must be at the start of a line
1126 return (L span ITvccurly)
1128 -- -----------------------------------------------------------------------------
1131 setLine :: Int -> Action
1132 setLine code span buf len = do
1133 let line = parseUnsignedInteger buf len 10 octDecDigit
1134 setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 1)
1135 -- subtract one: the line number refers to the *following* line
1140 setFile :: Int -> Action
1141 setFile code span buf len = do
1142 let file = lexemeToFastString (stepOn buf) (len-2)
1143 setAlrLastLoc noSrcSpan
1144 setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span))
1150 -- -----------------------------------------------------------------------------
1151 -- Options, includes and language pragmas.
1153 lex_string_prag :: (String -> Token) -> Action
1154 lex_string_prag mkTok span _buf _len
1155 = do input <- getInput
1159 return (L (mkSrcSpan start end) tok)
1161 = if isString input "#-}"
1162 then do setInput input
1163 return (mkTok (reverse acc))
1164 else case alexGetChar input of
1165 Just (c,i) -> go (c:acc) i
1166 Nothing -> err input
1167 isString _ [] = True
1169 = case alexGetChar i of
1170 Just (c,i') | c == x -> isString i' xs
1172 err (AI end _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma"
1175 -- -----------------------------------------------------------------------------
1178 -- This stuff is horrible. I hates it.
1180 lex_string_tok :: Action
1181 lex_string_tok span _buf _len = do
1182 tok <- lex_string ""
1184 return (L (mkSrcSpan (srcSpanStart span) end) tok)
1186 lex_string :: String -> P Token
1189 case alexGetChar' i of
1190 Nothing -> lit_error i
1194 magicHash <- extension magicHashEnabled
1198 case alexGetChar' i of
1202 then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'"
1203 else let s' = mkZFastString (reverse s) in
1204 return (ITprimstring s')
1205 -- mkZFastString is a hack to avoid encoding the
1206 -- string in UTF-8. We just want the exact bytes.
1208 return (ITstring (mkFastString (reverse s)))
1210 return (ITstring (mkFastString (reverse s)))
1213 | Just ('&',i) <- next -> do
1214 setInput i; lex_string s
1215 | Just (c,i) <- next, c <= '\x7f' && is_space c -> do
1216 -- is_space only works for <= '\x7f' (#3751)
1217 setInput i; lex_stringgap s
1218 where next = alexGetChar' i
1222 '\\' -> do setInput i1; c' <- lex_escape; lex_string (c':s)
1223 c | isAny c -> do setInput i1; lex_string (c:s)
1224 _other -> lit_error i
1226 lex_stringgap :: String -> P Token
1227 lex_stringgap s = do
1229 c <- getCharOrFail i
1231 '\\' -> lex_string s
1232 c | is_space c -> lex_stringgap s
1233 _other -> lit_error i
1236 lex_char_tok :: Action
1237 -- Here we are basically parsing character literals, such as 'x' or '\n'
1238 -- but, when Template Haskell is on, we additionally spot
1239 -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively,
1240 -- but WITHOUT CONSUMING the x or T part (the parser does that).
1241 -- So we have to do two characters of lookahead: when we see 'x we need to
1242 -- see if there's a trailing quote
1243 lex_char_tok span _buf _len = do -- We've seen '
1244 i1 <- getInput -- Look ahead to first character
1245 let loc = srcSpanStart span
1246 case alexGetChar' i1 of
1247 Nothing -> lit_error i1
1249 Just ('\'', i2@(AI end2 _)) -> do -- We've seen ''
1250 th_exts <- extension thEnabled
1253 return (L (mkSrcSpan loc end2) ITtyQuote)
1256 Just ('\\', i2@(AI _end2 _)) -> do -- We've seen 'backslash
1258 lit_ch <- lex_escape
1260 mc <- getCharOrFail i3 -- Trailing quote
1261 if mc == '\'' then finish_char_tok loc lit_ch
1264 Just (c, i2@(AI _end2 _))
1265 | not (isAny c) -> lit_error i1
1268 -- We've seen 'x, where x is a valid character
1269 -- (i.e. not newline etc) but not a quote or backslash
1270 case alexGetChar' i2 of -- Look ahead one more character
1271 Just ('\'', i3) -> do -- We've seen 'x'
1273 finish_char_tok loc c
1274 _other -> do -- We've seen 'x not followed by quote
1275 -- (including the possibility of EOF)
1276 -- If TH is on, just parse the quote only
1277 th_exts <- extension thEnabled
1279 if th_exts then return (L (mkSrcSpan loc end) ITvarQuote)
1282 finish_char_tok :: SrcLoc -> Char -> P (Located Token)
1283 finish_char_tok loc ch -- We've already seen the closing quote
1284 -- Just need to check for trailing #
1285 = do magicHash <- extension magicHashEnabled
1286 i@(AI end _) <- getInput
1287 if magicHash then do
1288 case alexGetChar' i of
1289 Just ('#',i@(AI end _)) -> do
1291 return (L (mkSrcSpan loc end) (ITprimchar ch))
1293 return (L (mkSrcSpan loc end) (ITchar ch))
1295 return (L (mkSrcSpan loc end) (ITchar ch))
1297 isAny :: Char -> Bool
1298 isAny c | c > '\x7f' = isPrint c
1299 | otherwise = is_any c
1301 lex_escape :: P Char
1304 c <- getCharOrFail i0
1316 '^' -> do i1 <- getInput
1317 c <- getCharOrFail i1
1318 if c >= '@' && c <= '_'
1319 then return (chr (ord c - ord '@'))
1322 'x' -> readNum is_hexdigit 16 hexDigit
1323 'o' -> readNum is_octdigit 8 octDecDigit
1324 x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x)
1328 case alexGetChar' i of
1329 Nothing -> lit_error i0
1331 case alexGetChar' i2 of
1332 Nothing -> do lit_error i0
1334 let str = [c1,c2,c3] in
1335 case [ (c,rest) | (p,c) <- silly_escape_chars,
1336 Just rest <- [stripPrefix p str] ] of
1337 (escape_char,[]):_ -> do
1340 (escape_char,_:_):_ -> do
1345 readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char
1346 readNum is_digit base conv = do
1348 c <- getCharOrFail i
1350 then readNum2 is_digit base conv (conv c)
1353 readNum2 :: (Char -> Bool) -> Int -> (Char -> Int) -> Int -> P Char
1354 readNum2 is_digit base conv i = do
1357 where read i input = do
1358 case alexGetChar' input of
1359 Just (c,input') | is_digit c -> do
1360 let i' = i*base + conv c
1362 then setInput input >> lexError "numeric escape sequence out of range"
1365 setInput input; return (chr i)
1368 silly_escape_chars :: [(String, Char)]
1369 silly_escape_chars = [
1406 -- before calling lit_error, ensure that the current input is pointing to
1407 -- the position of the error in the buffer. This is so that we can report
1408 -- a correct location to the user, but also so we can detect UTF-8 decoding
1409 -- errors if they occur.
1410 lit_error :: AlexInput -> P a
1411 lit_error i = do setInput i; lexError "lexical error in string/character literal"
1413 getCharOrFail :: AlexInput -> P Char
1414 getCharOrFail i = do
1415 case alexGetChar' i of
1416 Nothing -> lexError "unexpected end-of-file in string/character literal"
1417 Just (c,i) -> do setInput i; return c
1419 -- -----------------------------------------------------------------------------
1422 lex_quasiquote_tok :: Action
1423 lex_quasiquote_tok span buf len = do
1424 let quoter = tail (lexemeToString buf (len - 1))
1425 -- 'tail' drops the initial '[',
1426 -- while the -1 drops the trailing '|'
1427 quoteStart <- getSrcLoc
1428 quote <- lex_quasiquote ""
1430 return (L (mkSrcSpan (srcSpanStart span) end)
1431 (ITquasiQuote (mkFastString quoter,
1432 mkFastString (reverse quote),
1433 mkSrcSpan quoteStart end)))
1435 lex_quasiquote :: String -> P String
1436 lex_quasiquote s = do
1438 case alexGetChar' i of
1439 Nothing -> lit_error i
1442 | Just ('|',i) <- next -> do
1443 setInput i; lex_quasiquote ('|' : s)
1444 | Just (']',i) <- next -> do
1445 setInput i; lex_quasiquote (']' : s)
1446 where next = alexGetChar' i
1449 | Just (']',i) <- next -> do
1450 setInput i; return s
1451 where next = alexGetChar' i
1454 setInput i; lex_quasiquote (c : s)
1456 -- -----------------------------------------------------------------------------
1459 warn :: DynFlag -> SDoc -> Action
1460 warn option warning srcspan _buf _len = do
1461 addWarning option srcspan warning
1464 warnThen :: DynFlag -> SDoc -> Action -> Action
1465 warnThen option warning action srcspan buf len = do
1466 addWarning option srcspan warning
1467 action srcspan buf len
1469 -- -----------------------------------------------------------------------------
1480 SrcSpan -- The start and end of the text span related to
1481 -- the error. Might be used in environments which can
1482 -- show this span, e.g. by highlighting it.
1483 Message -- The error message
1485 data PState = PState {
1486 buffer :: StringBuffer,
1488 messages :: Messages,
1489 last_loc :: SrcSpan, -- pos of previous token
1490 last_len :: !Int, -- len of previous token
1491 loc :: SrcLoc, -- current loc (end of prev token + 1)
1492 extsBitmap :: !Int, -- bitmap that determines permitted extensions
1493 context :: [LayoutContext],
1495 -- Used in the alternative layout rule:
1496 -- These tokens are the next ones to be sent out. They are
1497 -- just blindly emitted, without the rule looking at them again:
1498 alr_pending_implicit_tokens :: [Located Token],
1499 -- This is the next token to be considered or, if it is Nothing,
1500 -- we need to get the next token from the input stream:
1501 alr_next_token :: Maybe (Located Token),
1502 -- This is what we consider to be the locatino of the last token
1504 alr_last_loc :: SrcSpan,
1505 -- The stack of layout contexts:
1506 alr_context :: [ALRContext],
1507 -- Are we expecting a '{'? If it's Just, then the ALRLayout tells
1508 -- us what sort of layout the '{' will open:
1509 alr_expecting_ocurly :: Maybe ALRLayout,
1510 -- Have we just had the '}' for a let block? If so, than an 'in'
1511 -- token doesn't need to close anything:
1512 alr_justClosedExplicitLetBlock :: Bool
1514 -- last_loc and last_len are used when generating error messages,
1515 -- and in pushCurrentContext only. Sigh, if only Happy passed the
1516 -- current token to happyError, we could at least get rid of last_len.
1517 -- Getting rid of last_loc would require finding another way to
1518 -- implement pushCurrentContext (which is only called from one place).
1520 data ALRContext = ALRNoLayout Bool{- does it contain commas? -}
1521 Bool{- is it a 'let' block? -}
1522 | ALRLayout ALRLayout Int
1523 data ALRLayout = ALRLayoutLet
1528 newtype P a = P { unP :: PState -> ParseResult a }
1530 instance Monad P where
1536 returnP a = a `seq` (P $ \s -> POk s a)
1538 thenP :: P a -> (a -> P b) -> P b
1539 (P m) `thenP` k = P $ \ s ->
1541 POk s1 a -> (unP (k a)) s1
1542 PFailed span err -> PFailed span err
1544 failP :: String -> P a
1545 failP msg = P $ \s -> PFailed (last_loc s) (text msg)
1547 failMsgP :: String -> P a
1548 failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg)
1550 failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a
1551 failLocMsgP loc1 loc2 str = P $ \_ -> PFailed (mkSrcSpan loc1 loc2) (text str)
1553 failSpanMsgP :: SrcSpan -> SDoc -> P a
1554 failSpanMsgP span msg = P $ \_ -> PFailed span msg
1556 getPState :: P PState
1557 getPState = P $ \s -> POk s s
1559 getDynFlags :: P DynFlags
1560 getDynFlags = P $ \s -> POk s (dflags s)
1562 withThisPackage :: (PackageId -> a) -> P a
1564 = do pkg <- liftM thisPackage getDynFlags
1567 extension :: (Int -> Bool) -> P Bool
1568 extension p = P $ \s -> POk s (p $! extsBitmap s)
1571 getExts = P $ \s -> POk s (extsBitmap s)
1573 setExts :: (Int -> Int) -> P ()
1574 setExts f = P $ \s -> POk s{ extsBitmap = f (extsBitmap s) } ()
1576 setSrcLoc :: SrcLoc -> P ()
1577 setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} ()
1579 getSrcLoc :: P SrcLoc
1580 getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc
1582 setLastToken :: SrcSpan -> Int -> P ()
1583 setLastToken loc len = P $ \s -> POk s {
1588 data AlexInput = AI SrcLoc StringBuffer
1590 alexInputPrevChar :: AlexInput -> Char
1591 alexInputPrevChar (AI _ buf) = prevChar buf '\n'
1593 alexGetChar :: AlexInput -> Maybe (Char,AlexInput)
1594 alexGetChar (AI loc s)
1596 | otherwise = adj_c `seq` loc' `seq` s' `seq`
1597 --trace (show (ord c)) $
1598 Just (adj_c, (AI loc' s'))
1599 where (c,s') = nextChar s
1600 loc' = advanceSrcLoc loc c
1608 other_graphic = '\x6'
1611 | c <= '\x06' = non_graphic
1613 -- Alex doesn't handle Unicode, so when Unicode
1614 -- character is encountered we output these values
1615 -- with the actual character value hidden in the state.
1617 case generalCategory c of
1618 UppercaseLetter -> upper
1619 LowercaseLetter -> lower
1620 TitlecaseLetter -> upper
1621 ModifierLetter -> other_graphic
1622 OtherLetter -> lower -- see #1103
1623 NonSpacingMark -> other_graphic
1624 SpacingCombiningMark -> other_graphic
1625 EnclosingMark -> other_graphic
1626 DecimalNumber -> digit
1627 LetterNumber -> other_graphic
1628 OtherNumber -> other_graphic
1629 ConnectorPunctuation -> symbol
1630 DashPunctuation -> symbol
1631 OpenPunctuation -> other_graphic
1632 ClosePunctuation -> other_graphic
1633 InitialQuote -> other_graphic
1634 FinalQuote -> other_graphic
1635 OtherPunctuation -> symbol
1636 MathSymbol -> symbol
1637 CurrencySymbol -> symbol
1638 ModifierSymbol -> symbol
1639 OtherSymbol -> symbol
1641 _other -> non_graphic
1643 -- This version does not squash unicode characters, it is used when
1645 alexGetChar' :: AlexInput -> Maybe (Char,AlexInput)
1646 alexGetChar' (AI loc s)
1648 | otherwise = c `seq` loc' `seq` s' `seq`
1649 --trace (show (ord c)) $
1650 Just (c, (AI loc' s'))
1651 where (c,s') = nextChar s
1652 loc' = advanceSrcLoc loc c
1654 getInput :: P AlexInput
1655 getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (AI l b)
1657 setInput :: AlexInput -> P ()
1658 setInput (AI l b) = P $ \s -> POk s{ loc=l, buffer=b } ()
1660 pushLexState :: Int -> P ()
1661 pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} ()
1663 popLexState :: P Int
1664 popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls
1666 getLexState :: P Int
1667 getLexState = P $ \s@PState{ lex_state=ls:_ } -> POk s ls
1669 popNextToken :: P (Maybe (Located Token))
1671 = P $ \s@PState{ alr_next_token = m } ->
1672 POk (s {alr_next_token = Nothing}) m
1674 setAlrLastLoc :: SrcSpan -> P ()
1675 setAlrLastLoc l = P $ \s -> POk (s {alr_last_loc = l}) ()
1677 getAlrLastLoc :: P SrcSpan
1678 getAlrLastLoc = P $ \s@(PState {alr_last_loc = l}) -> POk s l
1680 getALRContext :: P [ALRContext]
1681 getALRContext = P $ \s@(PState {alr_context = cs}) -> POk s cs
1683 setALRContext :: [ALRContext] -> P ()
1684 setALRContext cs = P $ \s -> POk (s {alr_context = cs}) ()
1686 getJustClosedExplicitLetBlock :: P Bool
1687 getJustClosedExplicitLetBlock
1688 = P $ \s@(PState {alr_justClosedExplicitLetBlock = b}) -> POk s b
1690 setJustClosedExplicitLetBlock :: Bool -> P ()
1691 setJustClosedExplicitLetBlock b
1692 = P $ \s -> POk (s {alr_justClosedExplicitLetBlock = b}) ()
1694 setNextToken :: Located Token -> P ()
1695 setNextToken t = P $ \s -> POk (s {alr_next_token = Just t}) ()
1697 popPendingImplicitToken :: P (Maybe (Located Token))
1698 popPendingImplicitToken
1699 = P $ \s@PState{ alr_pending_implicit_tokens = ts } ->
1702 (t : ts') -> POk (s {alr_pending_implicit_tokens = ts'}) (Just t)
1704 setPendingImplicitTokens :: [Located Token] -> P ()
1705 setPendingImplicitTokens ts = P $ \s -> POk (s {alr_pending_implicit_tokens = ts}) ()
1707 getAlrExpectingOCurly :: P (Maybe ALRLayout)
1708 getAlrExpectingOCurly = P $ \s@(PState {alr_expecting_ocurly = b}) -> POk s b
1710 setAlrExpectingOCurly :: Maybe ALRLayout -> P ()
1711 setAlrExpectingOCurly b = P $ \s -> POk (s {alr_expecting_ocurly = b}) ()
1713 -- for reasons of efficiency, flags indicating language extensions (eg,
1714 -- -fglasgow-exts or -XParr) are represented by a bitmap stored in an unboxed
1718 genericsBit = 0 -- {| and |}
1729 explicitForallBit :: Int
1730 explicitForallBit = 7 -- the 'forall' keyword and '.' symbol
1732 bangPatBit = 8 -- Tells the parser to understand bang-patterns
1733 -- (doesn't affect the lexer)
1735 tyFamBit = 9 -- indexed type families: 'family' keyword and kind sigs
1737 haddockBit = 10 -- Lex and parse Haddock comments
1739 magicHashBit = 11 -- "#" in both functions and operators
1741 kindSigsBit = 12 -- Kind signatures on type variables
1742 recursiveDoBit :: Int
1743 recursiveDoBit = 13 -- mdo
1744 unicodeSyntaxBit :: Int
1745 unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc
1746 unboxedTuplesBit :: Int
1747 unboxedTuplesBit = 15 -- (# and #)
1748 datatypeContextsBit :: Int
1749 datatypeContextsBit = 16
1750 transformComprehensionsBit :: Int
1751 transformComprehensionsBit = 17
1753 qqBit = 18 -- enable quasiquoting
1754 inRulePragBit :: Int
1756 rawTokenStreamBit :: Int
1757 rawTokenStreamBit = 20 -- producing a token stream with all comments included
1758 newQualOpsBit :: Int
1759 newQualOpsBit = 21 -- Haskell' qualified operator syntax, e.g. Prelude.(+)
1762 alternativeLayoutRuleBit :: Int
1763 alternativeLayoutRuleBit = 23
1765 always :: Int -> Bool
1767 genericsEnabled :: Int -> Bool
1768 genericsEnabled flags = testBit flags genericsBit
1769 parrEnabled :: Int -> Bool
1770 parrEnabled flags = testBit flags parrBit
1771 arrowsEnabled :: Int -> Bool
1772 arrowsEnabled flags = testBit flags arrowsBit
1773 thEnabled :: Int -> Bool
1774 thEnabled flags = testBit flags thBit
1775 ipEnabled :: Int -> Bool
1776 ipEnabled flags = testBit flags ipBit
1777 explicitForallEnabled :: Int -> Bool
1778 explicitForallEnabled flags = testBit flags explicitForallBit
1779 bangPatEnabled :: Int -> Bool
1780 bangPatEnabled flags = testBit flags bangPatBit
1781 -- tyFamEnabled :: Int -> Bool
1782 -- tyFamEnabled flags = testBit flags tyFamBit
1783 haddockEnabled :: Int -> Bool
1784 haddockEnabled flags = testBit flags haddockBit
1785 magicHashEnabled :: Int -> Bool
1786 magicHashEnabled flags = testBit flags magicHashBit
1787 -- kindSigsEnabled :: Int -> Bool
1788 -- kindSigsEnabled flags = testBit flags kindSigsBit
1789 unicodeSyntaxEnabled :: Int -> Bool
1790 unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit
1791 unboxedTuplesEnabled :: Int -> Bool
1792 unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit
1793 datatypeContextsEnabled :: Int -> Bool
1794 datatypeContextsEnabled flags = testBit flags datatypeContextsBit
1795 qqEnabled :: Int -> Bool
1796 qqEnabled flags = testBit flags qqBit
1797 -- inRulePrag :: Int -> Bool
1798 -- inRulePrag flags = testBit flags inRulePragBit
1799 rawTokenStreamEnabled :: Int -> Bool
1800 rawTokenStreamEnabled flags = testBit flags rawTokenStreamBit
1801 newQualOps :: Int -> Bool
1802 newQualOps flags = testBit flags newQualOpsBit
1803 oldQualOps :: Int -> Bool
1804 oldQualOps flags = not (newQualOps flags)
1805 alternativeLayoutRule :: Int -> Bool
1806 alternativeLayoutRule flags = testBit flags alternativeLayoutRuleBit
1808 -- PState for parsing options pragmas
1810 pragState :: DynFlags -> StringBuffer -> SrcLoc -> PState
1811 pragState dynflags buf loc = (mkPState dynflags buf loc) {
1812 lex_state = [bol, option_prags, 0]
1815 -- create a parse state
1817 mkPState :: DynFlags -> StringBuffer -> SrcLoc -> PState
1818 mkPState flags buf loc =
1822 messages = emptyMessages,
1823 last_loc = mkSrcSpan loc loc,
1826 extsBitmap = fromIntegral bitmap,
1828 lex_state = [bol, 0],
1829 alr_pending_implicit_tokens = [],
1830 alr_next_token = Nothing,
1831 alr_last_loc = noSrcSpan,
1833 alr_expecting_ocurly = Nothing,
1834 alr_justClosedExplicitLetBlock = False
1837 bitmap = genericsBit `setBitIf` xopt Opt_Generics flags
1838 .|. ffiBit `setBitIf` xopt Opt_ForeignFunctionInterface flags
1839 .|. parrBit `setBitIf` xopt Opt_PArr flags
1840 .|. arrowsBit `setBitIf` xopt Opt_Arrows flags
1841 .|. thBit `setBitIf` xopt Opt_TemplateHaskell flags
1842 .|. qqBit `setBitIf` xopt Opt_QuasiQuotes flags
1843 .|. ipBit `setBitIf` xopt Opt_ImplicitParams flags
1844 .|. explicitForallBit `setBitIf` xopt Opt_ExplicitForAll flags
1845 .|. bangPatBit `setBitIf` xopt Opt_BangPatterns flags
1846 .|. tyFamBit `setBitIf` xopt Opt_TypeFamilies flags
1847 .|. haddockBit `setBitIf` dopt Opt_Haddock flags
1848 .|. magicHashBit `setBitIf` xopt Opt_MagicHash flags
1849 .|. kindSigsBit `setBitIf` xopt Opt_KindSignatures flags
1850 .|. recursiveDoBit `setBitIf` xopt Opt_RecursiveDo flags
1851 .|. recBit `setBitIf` xopt Opt_DoRec flags
1852 .|. recBit `setBitIf` xopt Opt_Arrows flags
1853 .|. unicodeSyntaxBit `setBitIf` xopt Opt_UnicodeSyntax flags
1854 .|. unboxedTuplesBit `setBitIf` xopt Opt_UnboxedTuples flags
1855 .|. datatypeContextsBit `setBitIf` xopt Opt_DatatypeContexts flags
1856 .|. transformComprehensionsBit `setBitIf` xopt Opt_TransformListComp flags
1857 .|. rawTokenStreamBit `setBitIf` dopt Opt_KeepRawTokenStream flags
1858 .|. newQualOpsBit `setBitIf` xopt Opt_NewQualifiedOperators flags
1859 .|. alternativeLayoutRuleBit `setBitIf` xopt Opt_AlternativeLayoutRule flags
1861 setBitIf :: Int -> Bool -> Int
1862 b `setBitIf` cond | cond = bit b
1865 addWarning :: DynFlag -> SrcSpan -> SDoc -> P ()
1866 addWarning option srcspan warning
1867 = P $ \s@PState{messages=(ws,es), dflags=d} ->
1868 let warning' = mkWarnMsg srcspan alwaysQualify warning
1869 ws' = if dopt option d then ws `snocBag` warning' else ws
1870 in POk s{messages=(ws', es)} ()
1872 getMessages :: PState -> Messages
1873 getMessages PState{messages=ms} = ms
1875 getContext :: P [LayoutContext]
1876 getContext = P $ \s@PState{context=ctx} -> POk s ctx
1878 setContext :: [LayoutContext] -> P ()
1879 setContext ctx = P $ \s -> POk s{context=ctx} ()
1882 popContext = P $ \ s@(PState{ buffer = buf, context = ctx,
1883 last_len = len, last_loc = last_loc }) ->
1885 (_:tl) -> POk s{ context = tl } ()
1886 [] -> PFailed last_loc (srcParseErr buf len)
1888 -- Push a new layout context at the indentation of the last token read.
1889 -- This is only used at the outer level of a module when the 'module'
1890 -- keyword is missing.
1891 pushCurrentContext :: P ()
1892 pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } ->
1893 POk s{context = Layout (srcSpanStartCol loc) : ctx} ()
1895 getOffside :: P Ordering
1896 getOffside = P $ \s@PState{last_loc=loc, context=stk} ->
1897 let offs = srcSpanStartCol loc in
1898 let ord = case stk of
1899 (Layout n:_) -> --trace ("layout: " ++ show n ++ ", offs: " ++ show offs) $
1904 -- ---------------------------------------------------------------------------
1905 -- Construct a parse error
1908 :: StringBuffer -- current buffer (placed just after the last token)
1909 -> Int -- length of the previous token
1912 = hcat [ if null token
1913 then ptext (sLit "parse error (possibly incorrect indentation)")
1914 else hcat [ptext (sLit "parse error on input "),
1915 char '`', text token, char '\'']
1917 where token = lexemeToString (offsetBytes (-len) buf) len
1919 -- Report a parse failure, giving the span of the previous token as
1920 -- the location of the error. This is the entry point for errors
1921 -- detected during parsing.
1923 srcParseFail = P $ \PState{ buffer = buf, last_len = len,
1924 last_loc = last_loc } ->
1925 PFailed last_loc (srcParseErr buf len)
1927 -- A lexical error is reported at a particular position in the source file,
1928 -- not over a token range.
1929 lexError :: String -> P a
1932 (AI end buf) <- getInput
1933 reportLexError loc end buf str
1935 -- -----------------------------------------------------------------------------
1936 -- This is the top-level function: called from the parser each time a
1937 -- new token is to be read from the input.
1939 lexer :: (Located Token -> P a) -> P a
1941 alr <- extension alternativeLayoutRule
1942 let lexTokenFun = if alr then lexTokenAlr else lexToken
1943 tok@(L _span _tok__) <- lexTokenFun
1944 --trace ("token: " ++ show _tok__) $ do
1947 lexTokenAlr :: P (Located Token)
1948 lexTokenAlr = do mPending <- popPendingImplicitToken
1949 t <- case mPending of
1951 do mNext <- popNextToken
1954 Just next -> return next
1955 alternativeLayoutRuleToken t
1958 setAlrLastLoc (getLoc t)
1960 ITwhere -> setAlrExpectingOCurly (Just ALRLayoutWhere)
1961 ITlet -> setAlrExpectingOCurly (Just ALRLayoutLet)
1962 ITof -> setAlrExpectingOCurly (Just ALRLayoutOf)
1963 ITdo -> setAlrExpectingOCurly (Just ALRLayoutDo)
1964 ITmdo -> setAlrExpectingOCurly (Just ALRLayoutDo)
1965 ITrec -> setAlrExpectingOCurly (Just ALRLayoutDo)
1969 alternativeLayoutRuleToken :: Located Token -> P (Located Token)
1970 alternativeLayoutRuleToken t
1971 = do context <- getALRContext
1972 lastLoc <- getAlrLastLoc
1973 mExpectingOCurly <- getAlrExpectingOCurly
1974 justClosedExplicitLetBlock <- getJustClosedExplicitLetBlock
1975 setJustClosedExplicitLetBlock False
1976 dflags <- getDynFlags
1977 let transitional = xopt Opt_AlternativeLayoutRuleTransitional dflags
1979 thisCol = srcSpanStartCol thisLoc
1980 newLine = (lastLoc == noSrcSpan)
1981 || (srcSpanStartLine thisLoc > srcSpanEndLine lastLoc)
1982 case (unLoc t, context, mExpectingOCurly) of
1983 -- This case handles a GHC extension to the original H98
1985 (ITocurly, _, Just alrLayout) ->
1986 do setAlrExpectingOCurly Nothing
1987 let isLet = case alrLayout of
1988 ALRLayoutLet -> True
1990 setALRContext (ALRNoLayout (containsCommas ITocurly) isLet : context)
1992 -- ...and makes this case unnecessary
1994 -- I think our implicit open-curly handling is slightly
1995 -- different to John's, in how it interacts with newlines
1997 (ITocurly, _, Just _) ->
1998 do setAlrExpectingOCurly Nothing
2002 (_, ALRLayout _ col : ls, Just expectingOCurly)
2003 | (thisCol > col) ||
2005 isNonDecreasingIntentation expectingOCurly) ->
2006 do setAlrExpectingOCurly Nothing
2007 setALRContext (ALRLayout expectingOCurly thisCol : context)
2009 return (L thisLoc ITocurly)
2011 do setAlrExpectingOCurly Nothing
2012 setPendingImplicitTokens [L lastLoc ITccurly]
2014 return (L lastLoc ITocurly)
2015 (_, _, Just expectingOCurly) ->
2016 do setAlrExpectingOCurly Nothing
2017 setALRContext (ALRLayout expectingOCurly thisCol : context)
2019 return (L thisLoc ITocurly)
2020 -- We do the [] cases earlier than in the spec, as we
2021 -- have an actual EOF token
2022 (ITeof, ALRLayout _ _ : ls, _) ->
2025 return (L thisLoc ITccurly)
2028 -- the other ITeof case omitted; general case below covers it
2030 | justClosedExplicitLetBlock ->
2032 (ITin, ALRLayout ALRLayoutLet _ : ls, _)
2034 do setPendingImplicitTokens [t]
2036 return (L thisLoc ITccurly)
2037 -- This next case is to handle a transitional issue:
2038 (ITwhere, ALRLayout _ col : ls, _)
2039 | newLine && thisCol == col && transitional ->
2040 do addWarning Opt_WarnAlternativeLayoutRuleTransitional
2042 (transitionalAlternativeLayoutWarning
2043 "`where' clause at the same depth as implicit layout block")
2046 -- Note that we use lastLoc, as we may need to close
2047 -- more layouts, or give a semicolon
2048 return (L lastLoc ITccurly)
2049 -- This next case is to handle a transitional issue:
2050 (ITvbar, ALRLayout _ col : ls, _)
2051 | newLine && thisCol == col && transitional ->
2052 do addWarning Opt_WarnAlternativeLayoutRuleTransitional
2054 (transitionalAlternativeLayoutWarning
2055 "`|' at the same depth as implicit layout block")
2058 -- Note that we use lastLoc, as we may need to close
2059 -- more layouts, or give a semicolon
2060 return (L lastLoc ITccurly)
2061 (_, ALRLayout _ col : ls, _)
2062 | newLine && thisCol == col ->
2064 return (L thisLoc ITsemi)
2065 | newLine && thisCol < col ->
2068 -- Note that we use lastLoc, as we may need to close
2069 -- more layouts, or give a semicolon
2070 return (L lastLoc ITccurly)
2071 -- We need to handle close before open, as 'then' is both
2072 -- an open and a close
2076 ALRLayout _ _ : ls ->
2079 return (L thisLoc ITccurly)
2080 ALRNoLayout _ isLet : ls ->
2081 do let ls' = if isALRopen u
2082 then ALRNoLayout (containsCommas u) False : ls
2085 when isLet $ setJustClosedExplicitLetBlock True
2088 do let ls = if isALRopen u
2089 then [ALRNoLayout (containsCommas u) False]
2092 -- XXX This is an error in John's code, but
2093 -- it looks reachable to me at first glance
2097 do setALRContext (ALRNoLayout (containsCommas u) False : context)
2099 (ITin, ALRLayout ALRLayoutLet _ : ls, _) ->
2101 setPendingImplicitTokens [t]
2102 return (L thisLoc ITccurly)
2103 (ITin, ALRLayout _ _ : ls, _) ->
2106 return (L thisLoc ITccurly)
2107 -- the other ITin case omitted; general case below covers it
2108 (ITcomma, ALRLayout _ _ : ls, _)
2109 | topNoLayoutContainsCommas ls ->
2112 return (L thisLoc ITccurly)
2113 (ITwhere, ALRLayout ALRLayoutDo _ : ls, _) ->
2115 setPendingImplicitTokens [t]
2116 return (L thisLoc ITccurly)
2117 -- the other ITwhere case omitted; general case below covers it
2118 (_, _, _) -> return t
2120 transitionalAlternativeLayoutWarning :: String -> SDoc
2121 transitionalAlternativeLayoutWarning msg
2122 = text "transitional layout will not be accepted in the future:"
2125 isALRopen :: Token -> Bool
2126 isALRopen ITcase = True
2127 isALRopen ITif = True
2128 isALRopen ITthen = True
2129 isALRopen IToparen = True
2130 isALRopen ITobrack = True
2131 isALRopen ITocurly = True
2133 isALRopen IToubxparen = True
2134 isALRopen ITparenEscape = True
2137 isALRclose :: Token -> Bool
2138 isALRclose ITof = True
2139 isALRclose ITthen = True
2140 isALRclose ITelse = True
2141 isALRclose ITcparen = True
2142 isALRclose ITcbrack = True
2143 isALRclose ITccurly = True
2145 isALRclose ITcubxparen = True
2146 isALRclose _ = False
2148 isNonDecreasingIntentation :: ALRLayout -> Bool
2149 isNonDecreasingIntentation ALRLayoutDo = True
2150 isNonDecreasingIntentation _ = False
2152 containsCommas :: Token -> Bool
2153 containsCommas IToparen = True
2154 containsCommas ITobrack = True
2155 -- John doesn't have {} as containing commas, but records contain them,
2156 -- which caused a problem parsing Cabal's Distribution.Simple.InstallDirs
2157 -- (defaultInstallDirs).
2158 containsCommas ITocurly = True
2160 containsCommas IToubxparen = True
2161 containsCommas _ = False
2163 topNoLayoutContainsCommas :: [ALRContext] -> Bool
2164 topNoLayoutContainsCommas [] = False
2165 topNoLayoutContainsCommas (ALRLayout _ _ : ls) = topNoLayoutContainsCommas ls
2166 topNoLayoutContainsCommas (ALRNoLayout b _ : _) = b
2168 lexToken :: P (Located Token)
2170 inp@(AI loc1 buf) <- getInput
2173 case alexScanUser exts inp sc of
2175 let span = mkSrcSpan loc1 loc1
2177 return (L span ITeof)
2178 AlexError (AI loc2 buf) ->
2179 reportLexError loc1 loc2 buf "lexical error"
2180 AlexSkip inp2 _ -> do
2183 AlexToken inp2@(AI end buf2) _ t -> do
2185 let span = mkSrcSpan loc1 end
2186 let bytes = byteDiff buf buf2
2187 span `seq` setLastToken span bytes
2190 reportLexError :: SrcLoc -> SrcLoc -> StringBuffer -> [Char] -> P a
2191 reportLexError loc1 loc2 buf str
2192 | atEnd buf = failLocMsgP loc1 loc2 (str ++ " at end of input")
2195 c = fst (nextChar buf)
2197 if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar#
2198 then failLocMsgP loc2 loc2 (str ++ " (UTF-8 decoding error)")
2199 else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c)
2201 lexTokenStream :: StringBuffer -> SrcLoc -> DynFlags -> ParseResult [Located Token]
2202 lexTokenStream buf loc dflags = unP go initState
2203 where dflags' = dopt_set (dopt_unset dflags Opt_Haddock) Opt_KeepRawTokenStream
2204 initState = mkPState dflags' buf loc
2206 ltok <- lexer return
2208 L _ ITeof -> return []
2209 _ -> liftM (ltok:) go
2211 linePrags = Map.singleton "line" (begin line_prag2)
2213 fileHeaderPrags = Map.fromList([("options", lex_string_prag IToptions_prag),
2214 ("options_ghc", lex_string_prag IToptions_prag),
2215 ("options_haddock", lex_string_prag ITdocOptions),
2216 ("language", token ITlanguage_prag),
2217 ("include", lex_string_prag ITinclude_prag)])
2219 ignoredPrags = Map.fromList (map ignored pragmas)
2220 where ignored opt = (opt, nested_comment lexToken)
2221 impls = ["hugs", "nhc98", "jhc", "yhc", "catch", "derive"]
2222 options_pragmas = map ("options_" ++) impls
2223 -- CFILES is a hugs-only thing.
2224 pragmas = options_pragmas ++ ["cfiles", "contract"]
2226 oneWordPrags = Map.fromList([("rules", rulePrag),
2227 ("inline", token (ITinline_prag Inline FunLike)),
2228 ("inlinable", token (ITinline_prag Inlinable FunLike)),
2229 ("inlineable", token (ITinline_prag Inlinable FunLike)),
2231 ("notinline", token (ITinline_prag NoInline FunLike)),
2232 ("specialize", token ITspec_prag),
2233 ("source", token ITsource_prag),
2234 ("warning", token ITwarning_prag),
2235 ("deprecated", token ITdeprecated_prag),
2236 ("scc", token ITscc_prag),
2237 ("generated", token ITgenerated_prag),
2238 ("core", token ITcore_prag),
2239 ("unpack", token ITunpack_prag),
2240 ("ann", token ITann_prag)])
2242 twoWordPrags = Map.fromList([("inline conlike", token (ITinline_prag Inline ConLike)),
2243 ("notinline conlike", token (ITinline_prag NoInline ConLike)),
2244 ("specialize inline", token (ITspec_inline_prag True)),
2245 ("specialize notinline", token (ITspec_inline_prag False))])
2248 dispatch_pragmas :: Map String Action -> Action
2249 dispatch_pragmas prags span buf len = case Map.lookup (clean_pragma (lexemeToString buf len)) prags of
2250 Just found -> found span buf len
2251 Nothing -> lexError "unknown pragma"
2253 known_pragma :: Map String Action -> AlexAccPred Int
2254 known_pragma prags _ _ len (AI _ buf) = (isJust $ Map.lookup (clean_pragma (lexemeToString (offsetBytes (- len) buf) len)) prags)
2255 && (nextCharIs buf (\c -> not (isAlphaNum c || c == '_')))
2257 clean_pragma :: String -> String
2258 clean_pragma prag = canon_ws (map toLower (unprefix prag))
2259 where unprefix prag' = case stripPrefix "{-#" prag' of
2262 canonical prag' = case prag' of
2263 "noinline" -> "notinline"
2264 "specialise" -> "specialize"
2265 "constructorlike" -> "conlike"
2267 canon_ws s = unwords (map canonical (words s))