X-Git-Url: http://git.megacz.com/?p=ghc-hetmet.git;a=blobdiff_plain;f=compiler%2Fparser%2FLexer.x;h=b3b48047e0ba0da605a8085e607341e99db6762c;hp=c935b2adb341b5dde4d9151e76ce9cd734731ebd;hb=25cead299c5857b9142a82c917080a654be44b83;hpb=6b4ab02f289394c82f27e46e44c017b4a0c88fb0 diff --git a/compiler/parser/Lexer.x b/compiler/parser/Lexer.x index c935b2a..b3b4804 100644 --- a/compiler/parser/Lexer.x +++ b/compiler/parser/Lexer.x @@ -12,7 +12,6 @@ ----------------------------------------------------------------------------- -- ToDo / known bugs: --- - Unicode -- - parsing integers is a bit slow -- - readRational is a bit slow -- @@ -20,8 +19,19 @@ -- - M... should be 3 tokens, not 1. -- - pragma-end should be only valid in a pragma +-- qualified operator NOTES. +-- +-- - If M.(+) is a single lexeme, then.. +-- - Probably (+) should be a single lexeme too, for consistency. +-- Otherwise ( + ) would be a prefix operator, but M.( + ) would not be. +-- - But we have to rule out reserved operators, otherwise (..) becomes +-- a different lexeme. +-- - Should we therefore also rule out reserved operators in the qualified +-- form? This is quite difficult to achieve. We don't do it for +-- qualified varids. + { -{-# OPTIONS -w #-} +{-# OPTIONS -Wwarn -w #-} -- The above warning supression flag is a temporary kludge. -- While working on this module you are encouraged to remove it and fix -- any warnings in the module. See @@ -31,6 +41,8 @@ -- Note that Alex itself generates code with with some unused bindings and -- without type signatures, so removing the flag might not be possible. +{-# OPTIONS_GHC -funbox-strict-fields #-} + module Lexer ( Token(..), lexer, pragState, mkPState, PState(..), P(..), ParseResult(..), getSrcLoc, @@ -39,7 +51,8 @@ module Lexer ( popContext, pushCurrentContext, setLastToken, setSrcLoc, getLexState, popLexState, pushLexState, extension, standaloneDerivingEnabled, bangPatEnabled, - addWarning + addWarning, + lexTokenStream ) where import Bag @@ -47,7 +60,6 @@ import ErrUtils import Outputable import StringBuffer import FastString -import FastTypes import SrcLoc import UniqFM import DynFlags @@ -56,19 +68,12 @@ import Util ( maybePrefixMatch, readRational ) import Control.Monad import Data.Bits -import Data.Char ( chr, ord, isSpace ) +import Data.Char import Data.Ratio -import Debug.Trace - -#if __GLASGOW_HASKELL__ >= 605 -import Data.Char ( GeneralCategory(..), generalCategory, isPrint, isUpper ) -#else -import Compat.Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper ) -#endif } $unispace = \x05 -- Trick Alex into handling Unicode. See alexGetChar. -$whitechar = [\ \n\r\f\v\xa0 $unispace] +$whitechar = [\ \n\r\f\v $unispace] $white_no_nl = $whitechar # \n $tab = \t @@ -78,16 +83,16 @@ $decdigit = $ascdigit -- for now, should really be $digit (ToDo) $digit = [$ascdigit $unidigit] $special = [\(\)\,\;\[\]\`\{\}] -$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~ \xa1-\xbf \xd7 \xf7] +$ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~] $unisymbol = \x04 -- Trick Alex into handling Unicode. See alexGetChar. $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\'] $unilarge = \x01 -- Trick Alex into handling Unicode. See alexGetChar. -$asclarge = [A-Z \xc0-\xd6 \xd8-\xde] +$asclarge = [A-Z] $large = [$asclarge $unilarge] $unismall = \x02 -- Trick Alex into handling Unicode. See alexGetChar. -$ascsmall = [a-z \xdf-\xf6 \xf8-\xff] +$ascsmall = [a-z] $small = [$ascsmall $unismall \_] $unigraphic = \x06 -- Trick Alex into handling Unicode. See alexGetChar. @@ -150,12 +155,12 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- space followed by a Haddock comment symbol (docsym) (in which case we'd -- have a Haddock comment). The rules then munch the rest of the line. -"-- " ~[$docsym \#] .* ; -"--" [^$symbol : \ ] .* ; +"-- " ~[$docsym \#] .* { lineCommentToken } +"--" [^$symbol : \ ] .* { lineCommentToken } -- Next, match Haddock comments if no -haddock flag -"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } ; +"-- " [$docsym \#] .* / { ifExtension (not . haddockEnabled) } { lineCommentToken } -- Now, when we've matched comments that begin with 2 dashes and continue -- with a different character, we need to match comments that begin with three @@ -163,17 +168,17 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- make sure that the first non-dash character isn't a symbol, and munch the -- rest of the line. -"---"\-* [^$symbol :] .* ; +"---"\-* [^$symbol :] .* { lineCommentToken } -- Since the previous rules all match dashes followed by at least one -- character, we also need to match a whole line filled with just dashes. -"--"\-* / { atEOL } ; +"--"\-* / { atEOL } { lineCommentToken } -- We need this rule since none of the other single line comment rules -- actually match this case. -"-- " / { atEOL } ; +"-- " / { atEOL } { lineCommentToken } -- 'bol' state: beginning of a line. Slurp up all the whitespace (including -- blank lines) until we find a non-whitespace character, then do layout @@ -213,7 +218,8 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } <0,option_prags> \n { begin bol } -"{-#" $whitechar* (line|LINE) { begin line_prag2 } +"{-#" $whitechar* (line|LINE) / { notFollowedByPragmaChar } + { begin line_prag2 } -- single-line line pragmas, of the form -- # "" \n @@ -229,59 +235,80 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- NOTE: accept -} at the end of a LINE pragma, for compatibility -- with older versions of GHC which generated these. --- We only want RULES pragmas to be picked up when explicit forall --- syntax is enabled is on, because the contents of the pragma always --- uses it. If it's not on then we're sure to get a parse error. --- (ToDo: we should really emit a warning when ignoring pragmas) --- XXX Now that we can enable this without the -fglasgow-exts hammer, --- is it better just to let the parse error happen? -<0> - "{-#" $whitechar* (RULES|rules) / { ifExtension explicitForallEnabled } { token ITrules_prag } - <0,option_prags> { - "{-#" $whitechar* (INLINE|inline) { token (ITinline_prag True) } - "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline) + "{-#" $whitechar* (RULES|rules) / { notFollowedByPragmaChar } { rulePrag } + "{-#" $whitechar* (INLINE|inline) / { notFollowedByPragmaChar } + { token (ITinline_prag True) } + "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline) / { notFollowedByPragmaChar } { token (ITinline_prag False) } - "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) + "{-#" $whitechar* (INLINE|inline) + $whitechar+ (CONLIKE|conlike) / { notFollowedByPragmaChar } + { token (ITinline_conlike_prag True) } + "{-#" $whitechar* (NO(T)?INLINE|no(t?)inline) + $whitechar+ (CONLIKE|constructorlike) / { notFollowedByPragmaChar } + { token (ITinline_conlike_prag False) } + "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) / { notFollowedByPragmaChar } { token ITspec_prag } "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) - $whitechar* (INLINE|inline) { token (ITspec_inline_prag True) } + $whitechar+ (INLINE|inline) / { notFollowedByPragmaChar } + { token (ITspec_inline_prag True) } "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) - $whitechar* (NO(T?)INLINE|no(t?)inline) + $whitechar+ (NO(T?)INLINE|no(t?)inline) / { notFollowedByPragmaChar } { token (ITspec_inline_prag False) } - "{-#" $whitechar* (SOURCE|source) { token ITsource_prag } - "{-#" $whitechar* (DEPRECATED|deprecated) + "{-#" $whitechar* (SOURCE|source) / { notFollowedByPragmaChar } + { token ITsource_prag } + "{-#" $whitechar* (WARNING|warning) / { notFollowedByPragmaChar } + { token ITwarning_prag } + "{-#" $whitechar* (DEPRECATED|deprecated) / { notFollowedByPragmaChar } { token ITdeprecated_prag } - "{-#" $whitechar* (SCC|scc) { token ITscc_prag } - "{-#" $whitechar* (GENERATED|generated) + "{-#" $whitechar* (SCC|scc) / { notFollowedByPragmaChar } + { token ITscc_prag } + "{-#" $whitechar* (GENERATED|generated) / { notFollowedByPragmaChar } { token ITgenerated_prag } - "{-#" $whitechar* (CORE|core) { token ITcore_prag } - "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag } - - "{-#" { nested_comment lexToken } + "{-#" $whitechar* (CORE|core) / { notFollowedByPragmaChar } + { token ITcore_prag } + "{-#" $whitechar* (UNPACK|unpack) / { notFollowedByPragmaChar } + { token ITunpack_prag } + "{-#" $whitechar* (ANN|ann) / { notFollowedByPragmaChar } + { token ITann_prag } + + -- We ignore all these pragmas, but don't generate a warning for them + -- CFILES is a hugs-only thing. + "{-#" $whitechar* (OPTIONS_(HUGS|hugs|NHC98|nhc98|JHC|jhc|YHC|yhc|CATCH|catch|DERIVE|derive)|CFILES|cfiles|CONTRACT|contract) / { notFollowedByPragmaChar } + { nested_comment lexToken } -- ToDo: should only be valid inside a pragma: - "#-}" { token ITclose_prag} + "#-}" { endPrag } } { - "{-#" $whitechar* (OPTIONS|options) { lex_string_prag IToptions_prag } - "{-#" $whitechar* (OPTIONS_GHC|options_ghc) + "{-#" $whitechar* (OPTIONS|options) / { notFollowedByPragmaChar } + { lex_string_prag IToptions_prag } + "{-#" $whitechar* (OPTIONS_GHC|options_ghc) / { notFollowedByPragmaChar } { lex_string_prag IToptions_prag } "{-#" $whitechar* (OPTIONS_HADDOCK|options_haddock) + / { notFollowedByPragmaChar } { lex_string_prag ITdocOptions } "-- #" { multiline_doc_comment } - "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag } - "{-#" $whitechar* (INCLUDE|include) { lex_string_prag ITinclude_prag } + "{-#" $whitechar* (LANGUAGE|language) / { notFollowedByPragmaChar } + { token ITlanguage_prag } + "{-#" $whitechar* (INCLUDE|include) / { notFollowedByPragmaChar } + { lex_string_prag ITinclude_prag } +} + +<0> { + -- In the "0" mode we ignore these pragmas + "{-#" $whitechar* (OPTIONS|options|OPTIONS_GHC|options_ghc|OPTIONS_HADDOCK|options_haddock|LANGUAGE|language|INCLUDE|include) / { notFollowedByPragmaChar } + { nested_comment lexToken } } <0> { - "-- #" .* ; + "-- #" .* { lineCommentToken } } <0,option_prags> { - -- This is to catch things like {-# OPTIONS OPTIONS_HUGS ... - "{-#" $whitechar* $idchar+ { nested_comment lexToken } + "{-#" { warnThen Opt_WarnUnrecognisedPragmas (text "Unrecognised pragma") + (nested_comment lexToken) } } -- '0' state: ordinary lexemes @@ -363,13 +390,15 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } @conid "#"+ / { ifExtension magicHashEnabled } { idtoken conid } } --- ToDo: M.(,,,) - +-- ToDo: - move `var` and (sym) into lexical syntax? +-- - remove backquote from $special? <0> { - @qual @varsym { idtoken qvarsym } - @qual @consym { idtoken qconsym } - @varsym { varsym } - @consym { consym } + @qual @varsym / { ifExtension oldQualOps } { idtoken qvarsym } + @qual @consym / { ifExtension oldQualOps } { idtoken qconsym } + @qual \( @varsym \) / { ifExtension newQualOps } { idtoken prefixqvarsym } + @qual \( @consym \) / { ifExtension newQualOps } { idtoken prefixqconsym } + @varsym { varsym } + @consym { consym } } -- For the normal boxed literals we need to be careful @@ -455,6 +484,7 @@ data Token | ITunsafe | ITstdcallconv | ITccallconv + | ITprimcallconv | ITdotnet | ITmdo | ITfamily @@ -464,16 +494,19 @@ data Token -- Pragmas | ITinline_prag Bool -- True <=> INLINE, False <=> NOINLINE + | ITinline_conlike_prag Bool -- same | ITspec_prag -- SPECIALISE | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE) | ITsource_prag | ITrules_prag + | ITwarning_prag | ITdeprecated_prag | ITline_prag | ITscc_prag | ITgenerated_prag | ITcore_prag -- hdaume: core annotations | ITunpack_prag + | ITann_prag | ITclose_prag | IToptions_prag String | ITinclude_prag String @@ -504,8 +537,8 @@ data Token | ITvocurly | ITvccurly | ITobrack - | ITopabrack -- [:, for parallel arrays with -fparr - | ITcpabrack -- :], for parallel arrays with -fparr + | ITopabrack -- [:, for parallel arrays with -XParr + | ITcpabrack -- :], for parallel arrays with -XParr | ITcbrack | IToparen | ITcparen @@ -524,6 +557,8 @@ data Token | ITqconid (FastString,FastString) | ITqvarsym (FastString,FastString) | ITqconsym (FastString,FastString) + | ITprefixqvarsym (FastString,FastString) + | ITprefixqconsym (FastString,FastString) | ITdupipvarid FastString -- GHC extension: implicit param: ?x @@ -573,6 +608,8 @@ data Token | ITdocSection Int String -- a section heading | ITdocOptions String -- doc options (prune, ignore-exports, etc) | ITdocOptionsOld String -- doc options declared "-- # ..."-style + | ITlineComment String -- comment starting by "--" + | ITblockComment String -- comment in {- -} #ifdef DEBUG deriving Show -- debugging @@ -595,6 +632,7 @@ isSpecial ITthreadsafe = True isSpecial ITunsafe = True isSpecial ITccallconv = True isSpecial ITstdcallconv = True +isSpecial ITprimcallconv = True isSpecial ITmdo = True isSpecial ITfamily = True isSpecial ITgroup = True @@ -610,6 +648,7 @@ isSpecial _ = False -- facilitates using a keyword in two different extensions that can be -- activated independently) -- +reservedWordsFM :: UniqFM (Token, Int) reservedWordsFM = listToUFM $ map (\(x, y, z) -> (mkFastString x, (y, z))) [( "_", ITunderscore, 0 ), @@ -639,7 +678,7 @@ reservedWordsFM = listToUFM $ ( "where", ITwhere, 0 ), ( "_scc_", ITscc, 0 ), -- ToDo: remove - ( "forall", ITforall, bit explicitForallBit), + ( "forall", ITforall, bit explicitForallBit .|. bit inRulePragBit), ( "mdo", ITmdo, bit recursiveDoBit), ( "family", ITfamily, bit tyFamBit), ( "group", ITgroup, bit transformComprehensionsBit), @@ -651,10 +690,11 @@ reservedWordsFM = listToUFM $ ( "label", ITlabel, bit ffiBit), ( "dynamic", ITdynamic, bit ffiBit), ( "safe", ITsafe, bit ffiBit), - ( "threadsafe", ITthreadsafe, bit ffiBit), + ( "threadsafe", ITthreadsafe, bit ffiBit), -- ToDo: remove ( "unsafe", ITunsafe, bit ffiBit), ( "stdcall", ITstdcallconv, bit ffiBit), ( "ccall", ITccallconv, bit ffiBit), + ( "prim", ITprimcallconv, bit ffiBit), ( "dotnet", ITdotnet, bit ffiBit), ( "rec", ITrec, bit arrowsBit), @@ -680,16 +720,15 @@ reservedSymsFM = listToUFM $ ,("!", ITbang, always) -- For data T (a::*) = MkT - ,("*", ITstar, \i -> kindSigsEnabled i || tyFamEnabled i) + ,("*", ITstar, always) -- \i -> kindSigsEnabled i || tyFamEnabled i) -- For 'forall a . t' - ,(".", ITdot, explicitForallEnabled) + ,(".", ITdot, always) -- \i -> explicitForallEnabled i || inRulePrag i) ,("-<", ITlarrowtail, arrowsEnabled) ,(">-", ITrarrowtail, arrowsEnabled) ,("-<<", ITLarrowtail, arrowsEnabled) ,(">>-", ITRarrowtail, arrowsEnabled) -#if __GLASGOW_HASKELL__ >= 605 ,("∷", ITdcolon, unicodeSyntaxEnabled) ,("⇒", ITdarrow, unicodeSyntaxEnabled) ,("∀", ITforall, \i -> unicodeSyntaxEnabled i && @@ -700,7 +739,6 @@ reservedSymsFM = listToUFM $ -- ToDo: ideally, → and ∷ should be "specials", so that they cannot -- form part of a large operator. This would let us have a better -- syntax for kinds: ɑ∷*→* would be a legal kind signature. (maybe). -#endif ] -- ----------------------------------------------------------------------------- @@ -741,19 +779,27 @@ pop_and :: Action -> Action pop_and act span buf len = do popLexState; act span buf len {-# INLINE nextCharIs #-} +nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool nextCharIs buf p = not (atEnd buf) && p (currentChar buf) +notFollowedBy :: Char -> AlexAccPred Int notFollowedBy char _ _ _ (AI _ _ buf) = nextCharIs buf (/=char) +notFollowedBySymbol :: AlexAccPred Int notFollowedBySymbol _ _ _ (AI _ _ buf) = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~") +notFollowedByPragmaChar :: AlexAccPred Int +notFollowedByPragmaChar _ _ _ (AI _ _ buf) + = nextCharIs buf (\c -> not (isAlphaNum c || c == '_')) + -- We must reject doc comments as being ordinary comments everywhere. -- In some cases the doc comment will be selected as the lexeme due to -- maximal munch, but not always, because the nested comment rule is -- valid in all states, but the doc-comment rules are only valid in -- the non-layout states. +isNormalComment :: AlexAccPred Int isNormalComment bits _ _ (AI _ _ buf) | haddockEnabled bits = notFollowedByDocOrPragma | otherwise = nextCharIs buf (/='#') @@ -761,6 +807,7 @@ isNormalComment bits _ _ (AI _ _ buf) notFollowedByDocOrPragma = not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#")) +spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf)) {- @@ -768,8 +815,10 @@ haddockDisabledAnd p bits _ _ (AI _ _ buf) = if haddockEnabled bits then False else (p buf) -} +atEOL :: AlexAccPred Int atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n' +ifExtension :: (Int -> Bool) -> AlexAccPred Int ifExtension pred bits _ _ _ = pred bits multiline_doc_comment :: Action @@ -800,6 +849,11 @@ multiline_doc_comment span buf _len = withLexedDocType (worker "") | otherwise -> input Nothing -> input +lineCommentToken :: Action +lineCommentToken span buf len = do + b <- extension rawTokenStreamEnabled + if b then strtoken ITlineComment span buf len else lexToken + {- nested comments require traversing by hand, they can't be parsed using regular expressions. @@ -807,20 +861,24 @@ multiline_doc_comment span buf _len = withLexedDocType (worker "") nested_comment :: P (Located Token) -> Action nested_comment cont span _str _len = do input <- getInput - go (1::Int) input + go "" (1::Int) input where - go 0 input = do setInput input; cont - go n input = case alexGetChar input of + go commentAcc 0 input = do setInput input + b <- extension rawTokenStreamEnabled + if b + then docCommentEnd input commentAcc ITblockComment _str span + else cont + go commentAcc n input = case alexGetChar input of Nothing -> errBrace input span Just ('-',input) -> case alexGetChar input of Nothing -> errBrace input span - Just ('\125',input) -> go (n-1) input - Just (_,_) -> go n input + Just ('\125',input) -> go commentAcc (n-1) input + Just (_,_) -> go ('-':commentAcc) n input Just ('\123',input) -> case alexGetChar input of Nothing -> errBrace input span - Just ('-',input) -> go (n+1) input - Just (_,_) -> go n input - Just (_,input) -> go n input + Just ('-',input) -> go ('-':'\123':commentAcc) (n+1) input + Just (_,_) -> go ('\123':commentAcc) n input + Just (c,input) -> go (c:commentAcc) n input nested_doc_comment :: Action nested_doc_comment span buf _len = withLexedDocType (go "") @@ -841,6 +899,8 @@ nested_doc_comment span buf _len = withLexedDocType (go "") Just (_,_) -> go ('\123':commentAcc) input docType False Just (c,input) -> go (c:commentAcc) input docType False +withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token)) + -> P (Located Token) withLexedDocType lexDocComment = do input@(AI _ _ buf) <- getInput case prevChar buf ' ' of @@ -849,12 +909,25 @@ withLexedDocType lexDocComment = do '$' -> lexDocComment input ITdocCommentNamed False '*' -> lexDocSection 1 input '#' -> lexDocComment input ITdocOptionsOld False + _ -> panic "withLexedDocType: Bad doc type" where lexDocSection n input = case alexGetChar input of Just ('*', input) -> lexDocSection (n+1) input Just (_, _) -> lexDocComment input (ITdocSection n) True Nothing -> do setInput input; lexToken -- eof reached, lex it normally +-- RULES pragmas turn on the forall and '.' keywords, and we turn them +-- off again at the end of the pragma. +rulePrag :: Action +rulePrag span _ _ = do + setExts (.|. bit inRulePragBit) + return (L span ITrules_prag) + +endPrag :: Action +endPrag span _ _ = do + setExts (.&. complement (bit inRulePragBit)) + return (L span ITclose_prag) + -- docCommentEnd ------------------------------------------------------------------------------- -- This function is quite tricky. We can't just return a new token, we also @@ -887,8 +960,9 @@ docCommentEnd input commentAcc docType buf span = do span `seq` setLastToken span' last_len last_line_len return (L span' (docType comment)) +errBrace :: AlexInput -> SrcSpan -> P a errBrace (AI end _ _) span = failLocMsgP (srcSpanStart span) end "unterminated `{-'" - + open_brace, close_brace :: Action open_brace span _str _len = do ctx <- getContext @@ -898,14 +972,15 @@ close_brace span _str _len = do popContext return (L span ITccurly) -qvarid buf len = ITqvarid $! splitQualName buf len -qconid buf len = ITqconid $! splitQualName buf len +qvarid, qconid :: StringBuffer -> Int -> Token +qvarid buf len = ITqvarid $! splitQualName buf len False +qconid buf len = ITqconid $! splitQualName buf len False -splitQualName :: StringBuffer -> Int -> (FastString,FastString) +splitQualName :: StringBuffer -> Int -> Bool -> (FastString,FastString) -- takes a StringBuffer and a length, and returns the module name -- and identifier parts of a qualified name. Splits at the *last* dot, -- because of hierarchical module names. -splitQualName orig_buf len = split orig_buf orig_buf +splitQualName orig_buf len parens = split orig_buf orig_buf where split buf dot_buf | orig_buf `byteDiff` buf >= len = done dot_buf @@ -925,11 +1000,14 @@ splitQualName orig_buf len = split orig_buf orig_buf done dot_buf = (lexemeToFastString orig_buf (qual_size - 1), - lexemeToFastString dot_buf (len - qual_size)) + if parens -- Prelude.(+) + then lexemeToFastString (stepOn dot_buf) (len - qual_size - 2) + else lexemeToFastString dot_buf (len - qual_size)) where qual_size = orig_buf `byteDiff` dot_buf -varid span buf len = +varid :: Action +varid span buf len = fs `seq` case lookupUFM reservedWordsFM fs of Just (keyword,0) -> do @@ -944,15 +1022,22 @@ varid span buf len = where fs = lexemeToFastString buf len +conid :: StringBuffer -> Int -> Token conid buf len = ITconid fs where fs = lexemeToFastString buf len -qvarsym buf len = ITqvarsym $! splitQualName buf len -qconsym buf len = ITqconsym $! splitQualName buf len +qvarsym, qconsym, prefixqvarsym, prefixqconsym :: StringBuffer -> Int -> Token +qvarsym buf len = ITqvarsym $! splitQualName buf len False +qconsym buf len = ITqconsym $! splitQualName buf len False +prefixqvarsym buf len = ITprefixqvarsym $! splitQualName buf len True +prefixqconsym buf len = ITprefixqconsym $! splitQualName buf len True +varsym, consym :: Action varsym = sym ITvarsym consym = sym ITconsym +sym :: (FastString -> Token) -> SrcSpan -> StringBuffer -> Int + -> P (Located Token) sym con span buf len = case lookupUFM reservedSymsFM fs of Just (keyword,exts) -> do @@ -974,16 +1059,27 @@ tok_integral itint transint transbuf translen (radix,char_to_int) span buf len = (offsetBytes transbuf buf) (subtract translen len) radix char_to_int -- some conveniences for use with tok_integral +tok_num :: (Integer -> Integer) + -> Int -> Int + -> (Integer, (Char->Int)) -> Action tok_num = tok_integral ITinteger +tok_primint :: (Integer -> Integer) + -> Int -> Int + -> (Integer, (Char->Int)) -> Action tok_primint = tok_integral ITprimint +tok_primword :: Int -> Int + -> (Integer, (Char->Int)) -> Action tok_primword = tok_integral ITprimword positive +positive, negative :: (Integer -> Integer) positive = id negative = negate +decimal, octal, hexadecimal :: (Integer, Char -> Int) decimal = (10,octDecDigit) octal = (8,octDecDigit) hexadecimal = (16,hexDigit) -- readRational can understand negative rationals, exponents, everything. +tok_float, tok_primfloat, tok_primdouble :: String -> Token tok_float str = ITrational $! readRational str tok_primfloat str = ITprimfloat $! readRational str tok_primdouble str = ITprimdouble $! readRational str @@ -1011,6 +1107,7 @@ do_bol span _str _len = do -- certain keywords put us in the "layout" state, where we might -- add an opening curly brace. +maybe_layout :: Token -> P () maybe_layout ITdo = pushLexState layout_do maybe_layout ITmdo = pushLexState layout_do maybe_layout ITof = pushLexState layout @@ -1028,6 +1125,7 @@ maybe_layout _ = return () -- by a 'do', then we allow the new context to be at the same indentation as -- the previous context. This is what the 'strict' argument is for. -- +new_layout_context :: Bool -> Action new_layout_context strict span _buf _len = do popLexState (AI _ offset _) <- getInput @@ -1044,6 +1142,7 @@ new_layout_context strict span _buf _len = do setContext (Layout offset : ctx) return (L span ITvocurly) +do_layout_left :: Action do_layout_left span _buf _len = do popLexState pushLexState bol -- we must be at the start of a line @@ -1143,6 +1242,7 @@ lex_string s = do c' <- lex_char c i lex_string (c':s) +lex_stringgap :: String -> P Token lex_stringgap s = do c <- getCharOrFail case c of @@ -1185,11 +1285,11 @@ lex_char_tok span _buf _len = do -- We've seen ' -- We've seen 'x, where x is a valid character -- (i.e. not newline etc) but not a quote or backslash case alexGetChar' i2 of -- Look ahead one more character - Nothing -> lit_error Just ('\'', i3) -> do -- We've seen 'x' setInput i3 finish_char_tok loc c _other -> do -- We've seen 'x not followed by quote + -- (including the possibility of EOF) -- If TH is on, just parse the quote only th_exts <- extension thEnabled let (AI end _ _) = i1 @@ -1218,7 +1318,8 @@ lex_char c inp = do c | isAny c -> do setInput inp; return c _other -> lit_error -isAny c | c > '\xff' = isPrint c +isAny :: Char -> Bool +isAny c | c > '\x7f' = isPrint c | otherwise = is_any c lex_escape :: P Char @@ -1271,6 +1372,7 @@ readNum is_digit base conv = do then readNum2 is_digit base conv (conv c) else do setInput i; lit_error +readNum2 :: (Char -> Bool) -> Int -> (Char -> Int) -> Int -> P Char readNum2 is_digit base conv i = do input <- getInput read i input @@ -1283,6 +1385,7 @@ readNum2 is_digit base conv i = do then do setInput input; return (chr i) else lit_error +silly_escape_chars :: [(String, Char)] silly_escape_chars = [ ("NUL", '\NUL'), ("SOH", '\SOH'), @@ -1324,6 +1427,7 @@ silly_escape_chars = [ -- the position of the error in the buffer. This is so that we can report -- a correct location to the user, but also so we can detect UTF-8 decoding -- errors if they occur. +lit_error :: P a lit_error = lexError "lexical error in string/character literal" getCharOrFail :: P Char @@ -1377,6 +1481,11 @@ warn option warning srcspan _buf _len = do addWarning option srcspan warning lexToken +warnThen :: DynFlag -> SDoc -> Action -> Action +warnThen option warning action srcspan buf len = do + addWarning option srcspan warning + action srcspan buf len + -- ----------------------------------------------------------------------------- -- The Parse Monad @@ -1395,14 +1504,14 @@ data ParseResult a data PState = PState { buffer :: StringBuffer, - dflags :: DynFlags, - messages :: Messages, + dflags :: DynFlags, + messages :: Messages, last_loc :: SrcSpan, -- pos of previous token last_offs :: !Int, -- offset of the previous token from the -- beginning of the current line. -- \t is equal to 8 spaces. last_len :: !Int, -- len of previous token - last_line_len :: !Int, + last_line_len :: !Int, loc :: SrcLoc, -- current loc (end of prev token + 1) extsBitmap :: !Int, -- bitmap that determines permitted extensions context :: [LayoutContext], @@ -1448,6 +1557,9 @@ extension p = P $ \s -> POk s (p $! extsBitmap s) getExts :: P Int getExts = P $ \s -> POk s (extsBitmap s) +setExts :: (Int -> Int) -> P () +setExts f = P $ \s -> POk s{ extsBitmap = f (extsBitmap s) } () + setSrcLoc :: SrcLoc -> P () setSrcLoc new_loc = P $ \s -> POk s{loc=new_loc} () @@ -1486,7 +1598,7 @@ alexGetChar (AI loc ofs s) adj_c | c <= '\x06' = non_graphic - | c <= '\xff' = c + | c <= '\x7f' = c -- Alex doesn't handle Unicode, so when Unicode -- character is encoutered we output these values -- with the actual character value hidden in the state. @@ -1496,20 +1608,20 @@ alexGetChar (AI loc ofs s) LowercaseLetter -> lower TitlecaseLetter -> upper ModifierLetter -> other_graphic - OtherLetter -> other_graphic + OtherLetter -> lower -- see #1103 NonSpacingMark -> other_graphic SpacingCombiningMark -> other_graphic EnclosingMark -> other_graphic DecimalNumber -> digit LetterNumber -> other_graphic OtherNumber -> other_graphic - ConnectorPunctuation -> other_graphic - DashPunctuation -> other_graphic + ConnectorPunctuation -> symbol + DashPunctuation -> symbol OpenPunctuation -> other_graphic ClosePunctuation -> other_graphic InitialQuote -> other_graphic FinalQuote -> other_graphic - OtherPunctuation -> other_graphic + OtherPunctuation -> symbol MathSymbol -> symbol CurrencySymbol -> symbol ModifierSymbol -> symbol @@ -1550,61 +1662,102 @@ getLexState :: P Int getLexState = P $ \s@PState{ lex_state=ls:_ } -> POk s ls -- for reasons of efficiency, flags indicating language extensions (eg, --- -fglasgow-exts or -fparr) are represented by a bitmap stored in an unboxed +-- -fglasgow-exts or -XParr) are represented by a bitmap stored in an unboxed -- integer -genericsBit, ffiBit, parrBit :: Int +genericsBit :: Int genericsBit = 0 -- {| and |} +ffiBit :: Int ffiBit = 1 +parrBit :: Int parrBit = 2 +arrowsBit :: Int arrowsBit = 4 +thBit :: Int thBit = 5 +ipBit :: Int ipBit = 6 +explicitForallBit :: Int explicitForallBit = 7 -- the 'forall' keyword and '.' symbol +bangPatBit :: Int bangPatBit = 8 -- Tells the parser to understand bang-patterns -- (doesn't affect the lexer) +tyFamBit :: Int tyFamBit = 9 -- indexed type families: 'family' keyword and kind sigs +haddockBit :: Int haddockBit = 10 -- Lex and parse Haddock comments -magicHashBit = 11 -- # in both functions and operators +magicHashBit :: Int +magicHashBit = 11 -- "#" in both functions and operators +kindSigsBit :: Int kindSigsBit = 12 -- Kind signatures on type variables +recursiveDoBit :: Int recursiveDoBit = 13 -- mdo +unicodeSyntaxBit :: Int unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc +unboxedTuplesBit :: Int unboxedTuplesBit = 15 -- (# and #) +standaloneDerivingBit :: Int standaloneDerivingBit = 16 -- standalone instance deriving declarations +transformComprehensionsBit :: Int transformComprehensionsBit = 17 +qqBit :: Int qqBit = 18 -- enable quasiquoting - -genericsEnabled, ffiEnabled, parrEnabled :: Int -> Bool +inRulePragBit :: Int +inRulePragBit = 19 +rawTokenStreamBit :: Int +rawTokenStreamBit = 20 -- producing a token stream with all comments included +newQualOpsBit :: Int +newQualOpsBit = 21 -- Haskell' qualified operator syntax, e.g. Prelude.(+) + +always :: Int -> Bool always _ = True +genericsEnabled :: Int -> Bool genericsEnabled flags = testBit flags genericsBit -ffiEnabled flags = testBit flags ffiBit +parrEnabled :: Int -> Bool parrEnabled flags = testBit flags parrBit +arrowsEnabled :: Int -> Bool arrowsEnabled flags = testBit flags arrowsBit +thEnabled :: Int -> Bool thEnabled flags = testBit flags thBit +ipEnabled :: Int -> Bool ipEnabled flags = testBit flags ipBit +explicitForallEnabled :: Int -> Bool explicitForallEnabled flags = testBit flags explicitForallBit +bangPatEnabled :: Int -> Bool bangPatEnabled flags = testBit flags bangPatBit -tyFamEnabled flags = testBit flags tyFamBit +-- tyFamEnabled :: Int -> Bool +-- tyFamEnabled flags = testBit flags tyFamBit +haddockEnabled :: Int -> Bool haddockEnabled flags = testBit flags haddockBit +magicHashEnabled :: Int -> Bool magicHashEnabled flags = testBit flags magicHashBit -kindSigsEnabled flags = testBit flags kindSigsBit -recursiveDoEnabled flags = testBit flags recursiveDoBit +-- kindSigsEnabled :: Int -> Bool +-- kindSigsEnabled flags = testBit flags kindSigsBit +unicodeSyntaxEnabled :: Int -> Bool unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit +unboxedTuplesEnabled :: Int -> Bool unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit +standaloneDerivingEnabled :: Int -> Bool standaloneDerivingEnabled flags = testBit flags standaloneDerivingBit -transformComprehensionsEnabled flags = testBit flags transformComprehensionsBit +qqEnabled :: Int -> Bool qqEnabled flags = testBit flags qqBit +-- inRulePrag :: Int -> Bool +-- inRulePrag flags = testBit flags inRulePragBit +rawTokenStreamEnabled :: Int -> Bool +rawTokenStreamEnabled flags = testBit flags rawTokenStreamBit +newQualOps :: Int -> Bool +newQualOps flags = testBit flags newQualOpsBit +oldQualOps :: Int -> Bool +oldQualOps flags = not (newQualOps flags) -- PState for parsing options pragmas -- -pragState :: StringBuffer -> SrcLoc -> PState -pragState buf loc = +pragState :: DynFlags -> StringBuffer -> SrcLoc -> PState +pragState dynflags buf loc = PState { - buffer = buf, + buffer = buf, messages = emptyMessages, - -- XXX defaultDynFlags is not right, but we don't have a real - -- dflags handy - dflags = defaultDynFlags, + dflags = dynflags, last_loc = mkSrcSpan loc loc, last_offs = 0, last_len = 0, @@ -1643,6 +1796,7 @@ mkPState buf loc flags = .|. qqBit `setBitIf` dopt Opt_QuasiQuotes flags .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags .|. explicitForallBit `setBitIf` dopt Opt_ScopedTypeVariables flags + .|. explicitForallBit `setBitIf` dopt Opt_LiberalTypeSynonyms flags .|. explicitForallBit `setBitIf` dopt Opt_PolymorphicComponents flags .|. explicitForallBit `setBitIf` dopt Opt_ExistentialQuantification flags .|. explicitForallBit `setBitIf` dopt Opt_Rank2Types flags @@ -1656,7 +1810,9 @@ mkPState buf loc flags = .|. unicodeSyntaxBit `setBitIf` dopt Opt_UnicodeSyntax flags .|. unboxedTuplesBit `setBitIf` dopt Opt_UnboxedTuples flags .|. standaloneDerivingBit `setBitIf` dopt Opt_StandaloneDeriving flags - .|. transformComprehensionsBit `setBitIf` dopt Opt_TransformListComp flags + .|. transformComprehensionsBit `setBitIf` dopt Opt_TransformListComp flags + .|. rawTokenStreamBit `setBitIf` dopt Opt_KeepRawTokenStream flags + .|. newQualOpsBit `setBitIf` dopt Opt_NewQualifiedOperators flags -- setBitIf :: Int -> Bool -> Int b `setBitIf` cond | cond = bit b @@ -1763,6 +1919,7 @@ lexToken = do span `seq` setLastToken span bytes bytes t span buf bytes +reportLexError :: SrcLoc -> SrcLoc -> StringBuffer -> [Char] -> P a reportLexError loc1 loc2 buf str | atEnd buf = failLocMsgP loc1 loc2 (str ++ " at end of input") | otherwise = @@ -1772,4 +1929,13 @@ reportLexError loc1 loc2 buf str if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar# then failLocMsgP loc2 loc2 (str ++ " (UTF-8 decoding error)") else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c) + +lexTokenStream :: StringBuffer -> SrcLoc -> DynFlags -> ParseResult [Located Token] +lexTokenStream buf loc dflags = unP go initState + where initState = mkPState buf loc (dopt_set (dopt_unset dflags Opt_Haddock) Opt_KeepRawTokenStream) + go = do + ltok <- lexer return + case ltok of + L _ ITeof -> return [] + _ -> liftM (ltok:) go }