X-Git-Url: http://git.megacz.com/?p=ghc-hetmet.git;a=blobdiff_plain;f=compiler%2Fparser%2FLexer.x;h=b20d2c011cc46a36bd1977fc2f2dc190b96bfd08;hp=6cb2c3b7ca2c40f8f9c8c10b9f13be4d33432af9;hb=6ddfe9b18d4d280676aab2fa797ddbe6f8a09d6b;hpb=4bba92f93b88e15f0e0f23732d2cfa540acb737b diff --git a/compiler/parser/Lexer.x b/compiler/parser/Lexer.x index 6cb2c3b..b20d2c0 100644 --- a/compiler/parser/Lexer.x +++ b/compiler/parser/Lexer.x @@ -32,6 +32,7 @@ { -- XXX The above flags turn off warnings in the generated code: +{-# LANGUAGE BangPatterns #-} {-# OPTIONS_GHC -fno-warn-unused-matches #-} {-# OPTIONS_GHC -fno-warn-unused-binds #-} {-# OPTIONS_GHC -fno-warn-unused-imports #-} @@ -50,8 +51,9 @@ module Lexer ( failLocMsgP, failSpanMsgP, srcParseFail, getMessages, popContext, pushCurrentContext, setLastToken, setSrcLoc, + activeContext, nextIsEOF, getLexState, popLexState, pushLexState, - extension, standaloneDerivingEnabled, bangPatEnabled, + extension, bangPatEnabled, datatypeContextsEnabled, addWarning, lexTokenStream ) where @@ -66,6 +68,7 @@ import UniqFM import DynFlags import Module import Ctype +import BasicTypes ( InlineSpec(..), RuleMatchInfo(..), FractionalLit(..) ) import Util ( readRational ) import Control.Monad @@ -139,7 +142,7 @@ haskell :- -- everywhere: skip whitespace and comments $white_no_nl+ ; -$tab+ { warn Opt_WarnTabs (text "Tab character") } +$tab+ { warn Opt_WarnTabs (text "Warning: Tab character") } -- Everywhere: deal with nested comments. We explicitly rule out -- pragmas, "{-#", so that we don't accidentally treat them as comments. @@ -209,7 +212,7 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- context if the curly brace is missing. -- Careful! This stuff is quite delicate. { - \{ / { notFollowedBy '-' } { pop_and open_brace } + \{ / { notFollowedBy '-' } { hopefully_open_brace } -- we might encounter {-# here, but {- has been handled already \n ; ^\# (line)? { begin line_prag1 } @@ -285,7 +288,7 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- Haddock comments -<0> { +<0,option_prags> { "-- " $docsym / { ifExtension haddockEnabled } { multiline_doc_comment } "{-" \ ? $docsym / { ifExtension haddockEnabled } { nested_doc_comment } } @@ -307,6 +310,10 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape } "$(" / { ifExtension thEnabled } { token ITparenEscape } +-- For backward compatibility, accept the old dollar syntax + "[$" @varid "|" / { ifExtension qqEnabled } + { lex_quasiquote_tok } + "[" @varid "|" / { ifExtension qqEnabled } { lex_quasiquote_tok } } @@ -363,10 +370,8 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- ToDo: - move `var` and (sym) into lexical syntax? -- - remove backquote from $special? <0> { - @qual @varsym / { ifExtension oldQualOps } { idtoken qvarsym } - @qual @consym / { ifExtension oldQualOps } { idtoken qconsym } - @qual \( @varsym \) / { ifExtension newQualOps } { idtoken prefixqvarsym } - @qual \( @consym \) / { ifExtension newQualOps } { idtoken prefixqconsym } + @qual @varsym { idtoken qvarsym } + @qual @consym { idtoken qconsym } @varsym { varsym } @consym { consym } } @@ -451,6 +456,7 @@ data Token | ITdynamic | ITsafe | ITthreadsafe + | ITinterruptible | ITunsafe | ITstdcallconv | ITccallconv @@ -462,8 +468,7 @@ data Token | ITusing -- Pragmas - | ITinline_prag Bool -- True <=> INLINE, False <=> NOINLINE - | ITinline_conlike_prag Bool -- same + | ITinline_prag InlineSpec RuleMatchInfo | ITspec_prag -- SPECIALISE | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE) | ITsource_prag @@ -480,6 +485,8 @@ data Token | IToptions_prag String | ITinclude_prag String | ITlanguage_prag + | ITvect_prag + | ITvect_scalar_prag | ITdotdot -- reserved symbols | ITcolon @@ -506,8 +513,8 @@ data Token | ITvocurly | ITvccurly | ITobrack - | ITopabrack -- [:, for parallel arrays with -XParr - | ITcpabrack -- :], for parallel arrays with -XParr + | ITopabrack -- [:, for parallel arrays with -XParallelArrays + | ITcpabrack -- :], for parallel arrays with -XParallelArrays | ITcbrack | IToparen | ITcparen @@ -534,14 +541,14 @@ data Token | ITchar Char | ITstring FastString | ITinteger Integer - | ITrational Rational + | ITrational FractionalLit | ITprimchar Char | ITprimstring FastString | ITprimint Integer | ITprimword Integer - | ITprimfloat Rational - | ITprimdouble Rational + | ITprimfloat FractionalLit + | ITprimdouble FractionalLit -- Template Haskell extension tokens | ITopenExpQuote -- [| or [e| @@ -596,6 +603,7 @@ isSpecial ITlabel = True isSpecial ITdynamic = True isSpecial ITsafe = True isSpecial ITthreadsafe = True +isSpecial ITinterruptible = True isSpecial ITunsafe = True isSpecial ITccallconv = True isSpecial ITstdcallconv = True @@ -658,6 +666,7 @@ reservedWordsFM = listToUFM $ ( "dynamic", ITdynamic, bit ffiBit), ( "safe", ITsafe, bit ffiBit), ( "threadsafe", ITthreadsafe, bit ffiBit), -- ToDo: remove + ( "interruptible", ITinterruptible, bit ffiBit), ( "unsafe", ITunsafe, bit ffiBit), ( "stdcall", ITstdcallconv, bit ffiBit), ( "ccall", ITccallconv, bit ffiBit), @@ -701,7 +710,6 @@ reservedSymsFM = listToUFM $ explicitForallEnabled i) ,("→", ITrarrow, unicodeSyntaxEnabled) ,("←", ITlarrow, unicodeSyntaxEnabled) - ,("⋯", ITdotdot, unicodeSyntaxEnabled) ,("⤙", ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i) ,("⤚", ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i) @@ -750,6 +758,19 @@ pop :: Action pop _span _buf _len = do _ <- popLexState lexToken +hopefully_open_brace :: Action +hopefully_open_brace span buf len + = do relaxed <- extension relaxedLayout + ctx <- getContext + (AI l _) <- getInput + let offset = srcLocCol l + isOK = relaxed || + case ctx of + Layout prev_off : _ -> prev_off < offset + _ -> True + if isOK then pop_and open_brace span buf len + else failSpanMsgP span (text "Missing block") + pop_and :: Action -> Action pop_and act span buf len = do _ <- popLexState act span buf len @@ -1040,9 +1061,12 @@ 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 +tok_float str = ITrational $! readFractionalLit str +tok_primfloat str = ITprimfloat $! readFractionalLit str +tok_primdouble str = ITprimdouble $! readFractionalLit str + +readFractionalLit :: String -> FractionalLit +readFractionalLit str = (FL $! str) $! readRational str -- ----------------------------------------------------------------------------- -- Layout processing @@ -1100,10 +1124,12 @@ new_layout_context strict span _buf _len = do (AI l _) <- getInput let offset = srcLocCol l ctx <- getContext + nondecreasing <- extension nondecreasingIndentation + let strict' = strict || not nondecreasing case ctx of Layout prev_off : _ | - (strict && prev_off >= offset || - not strict && prev_off > offset) -> do + (strict' && prev_off >= offset || + not strict' && prev_off > offset) -> do -- token is indented to the left of the previous context. -- we must generate a {} sequence now. pushLexState layout_left @@ -1350,11 +1376,13 @@ readNum2 is_digit base conv i = do where read i input = do case alexGetChar' input of Just (c,input') | is_digit c -> do - read (i*base + conv c) input' + let i' = i*base + conv c + if i' > 0x10ffff + then setInput input >> lexError "numeric escape sequence out of range" + else read i' input' _other -> do - if i >= 0 && i <= 0x10FFFF - then do setInput input; return (chr i) - else lit_error input + setInput input; return (chr i) + silly_escape_chars :: [(String, Char)] silly_escape_chars = [ @@ -1616,7 +1644,7 @@ alexGetChar (AI loc s) EnclosingMark -> other_graphic DecimalNumber -> digit LetterNumber -> other_graphic - OtherNumber -> other_graphic + OtherNumber -> digit -- see #4373 ConnectorPunctuation -> symbol DashPunctuation -> symbol OpenPunctuation -> other_graphic @@ -1648,6 +1676,11 @@ getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (AI l b) setInput :: AlexInput -> P () setInput (AI l b) = P $ \s -> POk s{ loc=l, buffer=b } () +nextIsEOF :: P Bool +nextIsEOF = do + AI _ s <- getInput + return $ atEnd s + pushLexState :: Int -> P () pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} () @@ -1662,6 +1695,15 @@ popNextToken = P $ \s@PState{ alr_next_token = m } -> POk (s {alr_next_token = Nothing}) m +activeContext :: P Bool +activeContext = do + ctxt <- getALRContext + expc <- getAlrExpectingOCurly + impt <- implicitTokenPending + case (ctxt,expc) of + ([],Nothing) -> return impt + _other -> return True + setAlrLastLoc :: SrcSpan -> P () setAlrLastLoc l = P $ \s -> POk (s {alr_last_loc = l}) () @@ -1685,6 +1727,13 @@ setJustClosedExplicitLetBlock b setNextToken :: Located Token -> P () setNextToken t = P $ \s -> POk (s {alr_next_token = Just t}) () +implicitTokenPending :: P Bool +implicitTokenPending + = P $ \s@PState{ alr_pending_implicit_tokens = ts } -> + case ts of + [] -> POk s False + _ -> POk s True + popPendingImplicitToken :: P (Maybe (Located Token)) popPendingImplicitToken = P $ \s@PState{ alr_pending_implicit_tokens = ts } -> @@ -1702,7 +1751,7 @@ setAlrExpectingOCurly :: Maybe ALRLayout -> P () setAlrExpectingOCurly b = P $ \s -> POk (s {alr_expecting_ocurly = b}) () -- for reasons of efficiency, flags indicating language extensions (eg, --- -fglasgow-exts or -XParr) are represented by a bitmap stored in an unboxed +-- -fglasgow-exts or -XParallelArrays) are represented by a bitmap stored in an unboxed -- integer genericsBit :: Int @@ -1736,8 +1785,8 @@ unicodeSyntaxBit :: Int unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc unboxedTuplesBit :: Int unboxedTuplesBit = 15 -- (# and #) -standaloneDerivingBit :: Int -standaloneDerivingBit = 16 -- standalone instance deriving declarations +datatypeContextsBit :: Int +datatypeContextsBit = 16 transformComprehensionsBit :: Int transformComprehensionsBit = 17 qqBit :: Int @@ -1746,12 +1795,14 @@ 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.(+) recBit :: Int recBit = 22 -- rec alternativeLayoutRuleBit :: Int alternativeLayoutRuleBit = 23 +relaxedLayoutBit :: Int +relaxedLayoutBit = 24 +nondecreasingIndentationBit :: Int +nondecreasingIndentationBit = 25 always :: Int -> Bool always _ = True @@ -1781,50 +1832,34 @@ unicodeSyntaxEnabled :: Int -> Bool unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit unboxedTuplesEnabled :: Int -> Bool unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit -standaloneDerivingEnabled :: Int -> Bool -standaloneDerivingEnabled flags = testBit flags standaloneDerivingBit +datatypeContextsEnabled :: Int -> Bool +datatypeContextsEnabled flags = testBit flags datatypeContextsBit 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) alternativeLayoutRule :: Int -> Bool alternativeLayoutRule flags = testBit flags alternativeLayoutRuleBit +relaxedLayout :: Int -> Bool +relaxedLayout flags = testBit flags relaxedLayoutBit +nondecreasingIndentation :: Int -> Bool +nondecreasingIndentation flags = testBit flags nondecreasingIndentationBit -- PState for parsing options pragmas -- pragState :: DynFlags -> StringBuffer -> SrcLoc -> PState -pragState dynflags buf loc = - PState { - buffer = buf, - messages = emptyMessages, - dflags = dynflags, - last_loc = mkSrcSpan loc loc, - last_len = 0, - loc = loc, - extsBitmap = 0, - context = [], - lex_state = [bol, option_prags, 0], - alr_pending_implicit_tokens = [], - alr_next_token = Nothing, - alr_last_loc = noSrcSpan, - alr_context = [], - alr_expecting_ocurly = Nothing, - alr_justClosedExplicitLetBlock = False - } - +pragState dynflags buf loc = (mkPState dynflags buf loc) { + lex_state = [bol, option_prags, 0] + } -- create a parse state -- -mkPState :: StringBuffer -> SrcLoc -> DynFlags -> PState -mkPState buf loc flags = +mkPState :: DynFlags -> StringBuffer -> SrcLoc -> PState +mkPState flags buf loc = PState { - buffer = buf, + buffer = buf, dflags = flags, messages = emptyMessages, last_loc = mkSrcSpan loc loc, @@ -1833,7 +1868,6 @@ mkPState buf loc flags = extsBitmap = fromIntegral bitmap, context = [], lex_state = [bol, 0], - -- we begin in the layout state if toplev_layout is set alr_pending_implicit_tokens = [], alr_next_token = Nothing, alr_last_loc = noSrcSpan, @@ -1842,33 +1876,35 @@ mkPState buf loc flags = alr_justClosedExplicitLetBlock = False } where - bitmap = genericsBit `setBitIf` dopt Opt_Generics flags - .|. ffiBit `setBitIf` dopt Opt_ForeignFunctionInterface flags - .|. parrBit `setBitIf` dopt Opt_PArr flags - .|. arrowsBit `setBitIf` dopt Opt_Arrows flags - .|. thBit `setBitIf` dopt Opt_TemplateHaskell flags - .|. qqBit `setBitIf` dopt Opt_QuasiQuotes flags - .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags - .|. explicitForallBit `setBitIf` dopt Opt_ExplicitForAll flags - .|. bangPatBit `setBitIf` dopt Opt_BangPatterns flags - .|. tyFamBit `setBitIf` dopt Opt_TypeFamilies flags - .|. haddockBit `setBitIf` dopt Opt_Haddock flags - .|. magicHashBit `setBitIf` dopt Opt_MagicHash flags - .|. kindSigsBit `setBitIf` dopt Opt_KindSignatures flags - .|. recursiveDoBit `setBitIf` dopt Opt_RecursiveDo flags - .|. recBit `setBitIf` dopt Opt_DoRec flags - .|. recBit `setBitIf` dopt Opt_Arrows 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 + bitmap = genericsBit `setBitIf` xopt Opt_Generics flags + .|. ffiBit `setBitIf` xopt Opt_ForeignFunctionInterface flags + .|. parrBit `setBitIf` xopt Opt_ParallelArrays flags + .|. arrowsBit `setBitIf` xopt Opt_Arrows flags + .|. thBit `setBitIf` xopt Opt_TemplateHaskell flags + .|. qqBit `setBitIf` xopt Opt_QuasiQuotes flags + .|. ipBit `setBitIf` xopt Opt_ImplicitParams flags + .|. explicitForallBit `setBitIf` xopt Opt_ExplicitForAll flags + .|. bangPatBit `setBitIf` xopt Opt_BangPatterns flags + .|. tyFamBit `setBitIf` xopt Opt_TypeFamilies flags + .|. haddockBit `setBitIf` dopt Opt_Haddock flags + .|. magicHashBit `setBitIf` xopt Opt_MagicHash flags + .|. kindSigsBit `setBitIf` xopt Opt_KindSignatures flags + .|. recursiveDoBit `setBitIf` xopt Opt_RecursiveDo flags + .|. recBit `setBitIf` xopt Opt_DoRec flags + .|. recBit `setBitIf` xopt Opt_Arrows flags + .|. unicodeSyntaxBit `setBitIf` xopt Opt_UnicodeSyntax flags + .|. unboxedTuplesBit `setBitIf` xopt Opt_UnboxedTuples flags + .|. datatypeContextsBit `setBitIf` xopt Opt_DatatypeContexts flags + .|. transformComprehensionsBit `setBitIf` xopt Opt_TransformListComp flags + .|. transformComprehensionsBit `setBitIf` xopt Opt_MonadComprehensions flags .|. rawTokenStreamBit `setBitIf` dopt Opt_KeepRawTokenStream flags - .|. newQualOpsBit `setBitIf` dopt Opt_NewQualifiedOperators flags - .|. alternativeLayoutRuleBit `setBitIf` dopt Opt_AlternativeLayoutRule flags + .|. alternativeLayoutRuleBit `setBitIf` xopt Opt_AlternativeLayoutRule flags + .|. relaxedLayoutBit `setBitIf` xopt Opt_RelaxedLayout flags + .|. nondecreasingIndentationBit `setBitIf` xopt Opt_NondecreasingIndentation flags -- setBitIf :: Int -> Bool -> Int b `setBitIf` cond | cond = bit b - | otherwise = 0 + | otherwise = 0 addWarning :: DynFlag -> SrcSpan -> SDoc -> P () addWarning option srcspan warning @@ -1982,7 +2018,7 @@ alternativeLayoutRuleToken t justClosedExplicitLetBlock <- getJustClosedExplicitLetBlock setJustClosedExplicitLetBlock False dflags <- getDynFlags - let transitional = dopt Opt_AlternativeLayoutRuleTransitional dflags + let transitional = xopt Opt_AlternativeLayoutRuleTransitional dflags thisLoc = getLoc t thisCol = srcSpanStartCol thisLoc newLine = (lastLoc == noSrcSpan) @@ -2208,7 +2244,8 @@ reportLexError loc1 loc2 buf str 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) + where dflags' = dopt_set (dopt_unset dflags Opt_Haddock) Opt_KeepRawTokenStream + initState = mkPState dflags' buf loc go = do ltok <- lexer return case ltok of @@ -2231,8 +2268,11 @@ ignoredPrags = Map.fromList (map ignored pragmas) pragmas = options_pragmas ++ ["cfiles", "contract"] oneWordPrags = Map.fromList([("rules", rulePrag), - ("inline", token (ITinline_prag True)), - ("notinline", token (ITinline_prag False)), + ("inline", token (ITinline_prag Inline FunLike)), + ("inlinable", token (ITinline_prag Inlinable FunLike)), + ("inlineable", token (ITinline_prag Inlinable FunLike)), + -- Spelling variant + ("notinline", token (ITinline_prag NoInline FunLike)), ("specialize", token ITspec_prag), ("source", token ITsource_prag), ("warning", token ITwarning_prag), @@ -2241,13 +2281,14 @@ oneWordPrags = Map.fromList([("rules", rulePrag), ("generated", token ITgenerated_prag), ("core", token ITcore_prag), ("unpack", token ITunpack_prag), - ("ann", token ITann_prag)]) + ("ann", token ITann_prag), + ("vectorize", token ITvect_prag)]) -twoWordPrags = Map.fromList([("inline conlike", token (ITinline_conlike_prag True)), - ("notinline conlike", token (ITinline_conlike_prag False)), +twoWordPrags = Map.fromList([("inline conlike", token (ITinline_prag Inline ConLike)), + ("notinline conlike", token (ITinline_prag NoInline ConLike)), ("specialize inline", token (ITspec_inline_prag True)), - ("specialize notinline", token (ITspec_inline_prag False))]) - + ("specialize notinline", token (ITspec_inline_prag False)), + ("vectorize scalar", token ITvect_scalar_prag)]) dispatch_pragmas :: Map String Action -> Action dispatch_pragmas prags span buf len = case Map.lookup (clean_pragma (lexemeToString buf len)) prags of @@ -2266,6 +2307,7 @@ clean_pragma prag = canon_ws (map toLower (unprefix prag)) canonical prag' = case prag' of "noinline" -> "notinline" "specialise" -> "specialize" + "vectorise" -> "vectorize" "constructorlike" -> "conlike" _ -> prag' canon_ws s = unwords (map canonical (words s))