X-Git-Url: http://git.megacz.com/?p=ghc-hetmet.git;a=blobdiff_plain;f=compiler%2Fparser%2FLexer.x;h=c6457a4e5f604d9ca108b693e84edb507371b8b2;hp=2f6b7329a9a54b75f7ffe80577df418e2973ced5;hb=fc5cd848af22c97dfde1953b9efcfc60e6ccf2bb;hpb=03d8585e0940e28e024548654fe3505685aca94f diff --git a/compiler/parser/Lexer.x b/compiler/parser/Lexer.x index 2f6b732..c6457a4 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 -- @@ -20,54 +19,67 @@ -- - 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 #-} --- 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 --- http://hackage.haskell.org/trac/ghc/wiki/Commentary/CodingStyle#Warnings --- for details +-- XXX The above flags turn off warnings in the generated code: +{-# OPTIONS_GHC -fno-warn-unused-matches #-} +{-# OPTIONS_GHC -fno-warn-unused-binds #-} +{-# OPTIONS_GHC -fno-warn-unused-imports #-} +{-# OPTIONS_GHC -fno-warn-missing-signatures #-} +-- But alex still generates some code that causes the "lazy unlifted bindings" +-- warning, and old compilers don't know about it so we can't easily turn +-- it off, so for now we use the sledge hammer: +{-# OPTIONS_GHC -w #-} + +{-# OPTIONS_GHC -funbox-strict-fields #-} module Lexer ( Token(..), lexer, pragState, mkPState, PState(..), P(..), ParseResult(..), getSrcLoc, + getPState, getDynFlags, withThisPackage, failLocMsgP, failSpanMsgP, srcParseFail, - getMessages, + getMessages, popContext, pushCurrentContext, setLastToken, setSrcLoc, getLexState, popLexState, pushLexState, extension, standaloneDerivingEnabled, bangPatEnabled, - addWarning + addWarning, + lexTokenStream ) where -#include "HsVersions.h" - import Bag import ErrUtils import Outputable import StringBuffer import FastString -import FastTypes import SrcLoc import UniqFM import DynFlags +import Module import Ctype -import Util ( maybePrefixMatch, readRational ) +import Util ( readRational ) import Control.Monad import Data.Bits -import Data.Char ( chr, isSpace ) +import Data.Char +import Data.List +import Data.Maybe +import Data.Map (Map) +import qualified Data.Map as Map 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 @@ -77,16 +89,16 @@ $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. @@ -98,6 +110,8 @@ $symchar = [$symbol \:] $nl = [\n\r] $idchar = [$small $large $digit \'] +$pragmachar = [$small $large $digit] + $docsym = [\| \^ \* \$] @varid = $small $idchar* @@ -149,12 +163,12 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- space followed by a Haddock comment symbol (docsym) (in which case we'd -- have a Haddock comment). The rules then munch the rest of the line. -"-- " ~[$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 @@ -162,17 +176,17 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } -- make sure that the first non-dash character isn't a symbol, and munch the -- rest of the line. -"---"\-* [^$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 @@ -212,7 +226,8 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } <0,option_prags> \n { begin bol } -"{-#" $whitechar* (line|LINE) { begin line_prag2 } +"{-#" $whitechar* $pragmachar+ / { known_pragma linePrags } + { dispatch_pragmas linePrags } -- single-line line pragmas, of the form -- # "" \n @@ -228,59 +243,42 @@ $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) - { token (ITinline_prag False) } - "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) - { token ITspec_prag } - "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) - $whitechar* (INLINE|inline) { token (ITspec_inline_prag True) } - "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e) - $whitechar* (NO(T?)INLINE|no(t?)inline) - { token (ITspec_inline_prag False) } - "{-#" $whitechar* (SOURCE|source) { token ITsource_prag } - "{-#" $whitechar* (DEPRECATED|deprecated) - { token ITdeprecated_prag } - "{-#" $whitechar* (SCC|scc) { token ITscc_prag } - "{-#" $whitechar* (GENERATED|generated) - { token ITgenerated_prag } - "{-#" $whitechar* (CORE|core) { token ITcore_prag } - "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag } - - "{-#" { nested_comment lexToken } + "{-#" $whitechar* $pragmachar+ + $whitechar+ $pragmachar+ / { known_pragma twoWordPrags } + { dispatch_pragmas twoWordPrags } + + "{-#" $whitechar* $pragmachar+ / { known_pragma oneWordPrags } + { dispatch_pragmas oneWordPrags } + + -- We ignore all these pragmas, but don't generate a warning for them + "{-#" $whitechar* $pragmachar+ / { known_pragma ignoredPrags } + { dispatch_pragmas ignoredPrags } -- 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) - { lex_string_prag IToptions_prag } - "{-#" $whitechar* (OPTIONS_HADDOCK|options_haddock) - { lex_string_prag ITdocOptions } + "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags } + { dispatch_pragmas fileHeaderPrags } + "-- #" { multiline_doc_comment } - "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag } - "{-#" $whitechar* (INCLUDE|include) { lex_string_prag ITinclude_prag } } <0> { - "-- #" .* ; + -- In the "0" mode we ignore these pragmas + "{-#" $whitechar* $pragmachar+ / { known_pragma fileHeaderPrags } + { 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 @@ -308,6 +306,9 @@ $tab+ { warn Opt_WarnTabs (text "Tab character") } "|]" / { ifExtension thEnabled } { token ITcloseQuote } \$ @varid / { ifExtension thEnabled } { skip_one_varid ITidEscape } "$(" / { ifExtension thEnabled } { token ITparenEscape } + + "[$" @varid "|" / { ifExtension qqEnabled } + { lex_quasiquote_tok } } <0> { @@ -359,38 +360,44 @@ $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 -- when trying to be close to Haskell98 <0> { -- Normal integral literals (:: Num a => a, from Integer) - @decimal { tok_num positive 0 0 decimal } - 0[oO] @octal { tok_num positive 2 2 octal } - 0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal } + @decimal { tok_num positive 0 0 decimal } + 0[oO] @octal { tok_num positive 2 2 octal } + 0[xX] @hexadecimal { tok_num positive 2 2 hexadecimal } -- Normal rational literals (:: Fractional a => a, from Rational) - @floating_point { strtoken tok_float } + @floating_point { strtoken tok_float } } <0> { - -- Unboxed ints (:: Int#) + -- Unboxed ints (:: Int#) and words (:: Word#) -- It's simpler (and faster?) to give separate cases to the negatives, -- especially considering octal/hexadecimal prefixes. - @decimal \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal } - 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal } - 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal } - @negative @decimal \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal } - @negative 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal } + @decimal \# / { ifExtension magicHashEnabled } { tok_primint positive 0 1 decimal } + 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 octal } + 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint positive 2 3 hexadecimal } + @negative @decimal \# / { ifExtension magicHashEnabled } { tok_primint negative 1 2 decimal } + @negative 0[oO] @octal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 octal } @negative 0[xX] @hexadecimal \# / { ifExtension magicHashEnabled } { tok_primint negative 3 4 hexadecimal } + @decimal \# \# / { ifExtension magicHashEnabled } { tok_primword 0 2 decimal } + 0[oO] @octal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 octal } + 0[xX] @hexadecimal \# \# / { ifExtension magicHashEnabled } { tok_primword 2 4 hexadecimal } + -- Unboxed floats and doubles (:: Float#, :: Double#) -- prim_{float,double} work with signed literals @signed @floating_point \# / { ifExtension magicHashEnabled } { init_strtoken 1 tok_primfloat } @@ -447,22 +454,28 @@ data Token | ITunsafe | ITstdcallconv | ITccallconv - | ITdotnet + | ITprimcallconv | ITmdo | ITfamily + | ITgroup + | ITby + | ITusing -- 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 @@ -493,8 +506,8 @@ 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 @@ -513,11 +526,11 @@ data Token | ITqconid (FastString,FastString) | ITqvarsym (FastString,FastString) | ITqconsym (FastString,FastString) + | ITprefixqvarsym (FastString,FastString) + | ITprefixqconsym (FastString,FastString) | ITdupipvarid FastString -- GHC extension: implicit param: ?x - | ITpragma StringBuffer - | ITchar Char | ITstring FastString | ITinteger Integer @@ -526,10 +539,11 @@ data Token | ITprimchar Char | ITprimstring FastString | ITprimint Integer + | ITprimword Integer | ITprimfloat Rational | ITprimdouble Rational - -- MetaHaskell extension tokens + -- Template Haskell extension tokens | ITopenExpQuote -- [| or [e| | ITopenPatQuote -- [p| | ITopenDecQuote -- [d| @@ -539,6 +553,7 @@ data Token | ITparenEscape -- $( | ITvarQuote -- ' | ITtyQuote -- '' + | ITquasiQuote (FastString,FastString,SrcSpan) -- [:...|...|] -- Arrow notation extension | ITproc @@ -560,11 +575,14 @@ 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 #endif +{- isSpecial :: Token -> Bool -- If we see M.x, where x is a keyword, but -- is special, we treat is as just plain M.x, @@ -581,9 +599,14 @@ isSpecial ITthreadsafe = True isSpecial ITunsafe = True isSpecial ITccallconv = True isSpecial ITstdcallconv = True +isSpecial ITprimcallconv = True isSpecial ITmdo = True isSpecial ITfamily = True +isSpecial ITgroup = True +isSpecial ITby = True +isSpecial ITusing = True isSpecial _ = False +-} -- the bitmap provided as the third component indicates whether the -- corresponding extension keyword is valid under the extension options @@ -592,6 +615,7 @@ 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 ), @@ -621,22 +645,25 @@ 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), + ( "by", ITby, bit transformComprehensionsBit), + ( "using", ITusing, bit transformComprehensionsBit), ( "foreign", ITforeign, bit ffiBit), ( "export", ITexport, bit ffiBit), ( "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), - ( "dotnet", ITdotnet, bit ffiBit), + ( "prim", ITprimcallconv, bit ffiBit), - ( "rec", ITrec, bit arrowsBit), + ( "rec", ITrec, bit recBit), ( "proc", ITproc, bit arrowsBit) ] @@ -659,16 +686,15 @@ 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 && @@ -676,10 +702,17 @@ reservedSymsFM = listToUFM $ ,("→", ITrarrow, unicodeSyntaxEnabled) ,("←", ITlarrow, unicodeSyntaxEnabled) ,("⋯", ITdotdot, unicodeSyntaxEnabled) + + ,("⤙", ITlarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i) + ,("⤚", ITrarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i) + ,("⤛", ITLarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i) + ,("⤜", ITRarrowtail, \i -> unicodeSyntaxEnabled i && arrowsEnabled i) + + ,("★", ITstar, unicodeSyntaxEnabled) + -- 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 ] -- ----------------------------------------------------------------------------- @@ -688,11 +721,11 @@ reservedSymsFM = listToUFM $ type Action = SrcSpan -> StringBuffer -> Int -> P (Located Token) special :: Token -> Action -special tok span _buf len = return (L span tok) +special tok span _buf _len = return (L span tok) token, layout_token :: Token -> Action -token t span buf len = return (L span t) -layout_token t span buf len = pushLexState layout >> return (L span t) +token t span _buf _len = return (L span t) +layout_token t span _buf _len = pushLexState layout >> return (L span t) idtoken :: (StringBuffer -> Int -> Token) -> Action idtoken f span buf len = return (L span $! (f buf len)) @@ -714,17 +747,22 @@ begin :: Int -> Action begin code _span _str _len = do pushLexState code; lexToken pop :: Action -pop _span _buf _len = do popLexState; lexToken +pop _span _buf _len = do _ <- popLexState + lexToken pop_and :: Action -> Action -pop_and act span buf len = do popLexState; act span buf len +pop_and act span buf len = do _ <- popLexState + act span buf len {-# INLINE nextCharIs #-} +nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool nextCharIs buf p = not (atEnd buf) && p (currentChar buf) +notFollowedBy :: Char -> AlexAccPred Int notFollowedBy char _ _ _ (AI _ _ buf) = nextCharIs buf (/=char) +notFollowedBySymbol :: AlexAccPred Int notFollowedBySymbol _ _ _ (AI _ _ buf) = nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~") @@ -733,6 +771,7 @@ notFollowedBySymbol _ _ _ (AI _ _ buf) -- 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 (/='#') @@ -740,13 +779,18 @@ 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)) +{- 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 @@ -777,6 +821,11 @@ 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. @@ -784,20 +833,24 @@ 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 (c,_) -> 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 (c,_) -> go n input - Just (c,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 "") @@ -806,18 +859,20 @@ nested_doc_comment span buf _len = withLexedDocType (go "") Nothing -> errBrace input span Just ('-',input) -> case alexGetChar input of Nothing -> errBrace input span - Just ('\125',input@(AI end _ buf2)) -> + Just ('\125',input) -> docCommentEnd input commentAcc docType buf span - Just (c,_) -> go ('-':commentAcc) input docType False + Just (_,_) -> go ('-':commentAcc) input docType False Just ('\123', input) -> case alexGetChar input of Nothing -> errBrace input span Just ('-',input) -> do setInput input let cont = do input <- getInput; go commentAcc input docType False nested_comment cont span buf _len - Just (c,_) -> go ('\123':commentAcc) input docType False + 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 @@ -826,12 +881,25 @@ 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 (c, _) -> lexDocComment input (ITdocSection n) True + 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 _buf _len = do + setExts (.|. bit inRulePragBit) + return (L span ITrules_prag) + +endPrag :: Action +endPrag span _buf _len = 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 @@ -864,8 +932,9 @@ 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 @@ -875,14 +944,15 @@ 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 @@ -902,11 +972,15 @@ 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 maybe_layout keyword @@ -920,15 +994,22 @@ 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 @@ -950,15 +1031,27 @@ 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 @@ -978,21 +1071,31 @@ do_bol span _str _len = do return (L span ITvccurly) EQ -> do --trace "layout: inserting ';'" $ do - popLexState + _ <- popLexState return (L span ITsemi) GT -> do - popLexState + _ <- popLexState lexToken -- certain keywords put us in the "layout" state, where we might -- add an opening curly brace. -maybe_layout ITdo = pushLexState layout_do -maybe_layout ITmdo = pushLexState layout_do -maybe_layout ITof = pushLexState layout -maybe_layout ITlet = pushLexState layout -maybe_layout ITwhere = pushLexState layout -maybe_layout ITrec = pushLexState layout -maybe_layout _ = return () +maybe_layout :: Token -> P () +maybe_layout t = do -- If the alternative layout rule is enabled then + -- we never create an implicit layout context here. + -- Layout is handled XXX instead. + -- The code for closing implicit contexts, or + -- inserting implicit semi-colons, is therefore + -- irrelevant as it only applies in an implicit + -- context. + alr <- extension alternativeLayoutRule + unless alr $ f t + where f ITdo = pushLexState layout_do + f ITmdo = pushLexState layout_do + f ITof = pushLexState layout + f ITlet = pushLexState layout + f ITwhere = pushLexState layout + f ITrec = pushLexState layout + f _ = return () -- Pushing a new implicit layout context. If the indentation of the -- next token is not greater than the previous layout context, then @@ -1003,8 +1106,9 @@ 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 + _ <- popLexState (AI _ offset _) <- getInput ctx <- getContext case ctx of @@ -1015,12 +1119,13 @@ new_layout_context strict span _buf _len = do -- we must generate a {} sequence now. pushLexState layout_left return (L span ITvocurly) - other -> do + _ -> do setContext (Layout offset : ctx) return (L span ITvocurly) +do_layout_left :: Action do_layout_left span _buf _len = do - popLexState + _ <- popLexState pushLexState bol -- we must be at the start of a line return (L span ITvccurly) @@ -1030,17 +1135,18 @@ do_layout_left span _buf _len = do setLine :: Int -> Action setLine code span buf len = do let line = parseUnsignedInteger buf len 10 octDecDigit - setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0) + setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 1) -- subtract one: the line number refers to the *following* line - popLexState + _ <- popLexState pushLexState code lexToken setFile :: Int -> Action setFile code span buf len = do let file = lexemeToFastString (stepOn buf) (len-2) + setAlrLastLoc noSrcSpan setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span)) - popLexState + _ <- popLexState pushLexState code lexToken @@ -1049,7 +1155,7 @@ setFile code span buf len = do -- Options, includes and language pragmas. lex_string_prag :: (String -> Token) -> Action -lex_string_prag mkTok span buf len +lex_string_prag mkTok span _buf _len = do input <- getInput start <- getSrcLoc tok <- go [] input @@ -1062,7 +1168,7 @@ lex_string_prag mkTok span buf len else case alexGetChar input of Just (c,i) -> go (c:acc) i Nothing -> err input - isString i [] = True + isString _ [] = True isString i (x:xs) = case alexGetChar i of Just (c,i') | c == x -> isString i' xs @@ -1076,7 +1182,7 @@ lex_string_prag mkTok span buf len -- This stuff is horrible. I hates it. lex_string_tok :: Action -lex_string_tok span buf len = do +lex_string_tok span _buf _len = do tok <- lex_string "" end <- getSrcLoc return (L (mkSrcSpan (srcSpanStart span) end) tok) @@ -1118,6 +1224,7 @@ 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 @@ -1130,10 +1237,10 @@ lex_char_tok :: Action -- Here we are basically parsing character literals, such as 'x' or '\n' -- but, when Template Haskell is on, we additionally spot -- 'x and ''T, returning ITvarQuote and ITtyQuote respectively, --- but WIHTOUT CONSUMING the x or T part (the parser does that). +-- but WITHOUT CONSUMING the x or T part (the parser does that). -- So we have to do two characters of lookahead: when we see 'x we need to -- see if there's a trailing quote -lex_char_tok span buf len = do -- We've seen ' +lex_char_tok span _buf _len = do -- We've seen ' i1 <- getInput -- Look ahead to first character let loc = srcSpanStart span case alexGetChar' i1 of @@ -1146,25 +1253,25 @@ lex_char_tok span buf len = do -- We've seen ' return (L (mkSrcSpan loc end2) ITtyQuote) else lit_error - Just ('\\', i2@(AI end2 _ _)) -> do -- We've seen 'backslash + Just ('\\', i2@(AI _end2 _ _)) -> do -- We've seen 'backslash setInput i2 lit_ch <- lex_escape mc <- getCharOrFail -- Trailing quote if mc == '\'' then finish_char_tok loc lit_ch else do setInput i2; lit_error - Just (c, i2@(AI end2 _ _)) + Just (c, i2@(AI _end2 _ _)) | not (isAny c) -> lit_error | otherwise -> -- 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 @@ -1193,7 +1300,8 @@ 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 @@ -1217,7 +1325,7 @@ lex_escape = do 'x' -> readNum is_hexdigit 16 hexDigit 'o' -> readNum is_octdigit 8 octDecDigit - x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x) + x | is_decdigit x -> readNum2 is_decdigit 10 octDecDigit (octDecDigit x) c1 -> do i <- getInput @@ -1229,7 +1337,7 @@ lex_escape = do Just (c3,i3) -> let str = [c1,c2,c3] in case [ (c,rest) | (p,c) <- silly_escape_chars, - Just rest <- [maybePrefixMatch p str] ] of + Just rest <- [stripPrefix p str] ] of (escape_char,[]):_ -> do setInput i3 return escape_char @@ -1246,6 +1354,7 @@ 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 @@ -1258,6 +1367,7 @@ 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'), @@ -1299,6 +1409,7 @@ 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 @@ -1309,6 +1420,42 @@ getCharOrFail = do Just (c,i) -> do setInput i; return c -- ----------------------------------------------------------------------------- +-- QuasiQuote + +lex_quasiquote_tok :: Action +lex_quasiquote_tok span buf len = do + let quoter = reverse $ takeWhile (/= '$') + $ reverse $ lexemeToString buf (len - 1) + quoteStart <- getSrcLoc + quote <- lex_quasiquote "" + end <- getSrcLoc + return (L (mkSrcSpan (srcSpanStart span) end) + (ITquasiQuote (mkFastString quoter, + mkFastString (reverse quote), + mkSrcSpan quoteStart end))) + +lex_quasiquote :: String -> P String +lex_quasiquote s = do + i <- getInput + case alexGetChar' i of + Nothing -> lit_error + + Just ('\\',i) + | Just ('|',i) <- next -> do + setInput i; lex_quasiquote ('|' : s) + | Just (']',i) <- next -> do + setInput i; lex_quasiquote (']' : s) + where next = alexGetChar' i + + Just ('|',i) + | Just (']',i) <- next -> do + setInput i; return s + where next = alexGetChar' i + + Just (c, i) -> do + setInput i; lex_quasiquote (c : s) + +-- ----------------------------------------------------------------------------- -- Warnings warn :: DynFlag -> SDoc -> Action @@ -1316,6 +1463,11 @@ 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 @@ -1334,18 +1486,24 @@ 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], - lex_state :: [Int] + lex_state :: [Int], + -- Used in the alternative layout rule: + alr_pending_implicit_tokens :: [Located Token], + alr_next_token :: Maybe (Located Token), + alr_last_loc :: SrcSpan, + alr_context :: [ALRContext], + alr_expecting_ocurly :: Maybe ALRLayout } -- last_loc and last_len are used when generating error messages, -- and in pushCurrentContext only. Sigh, if only Happy passed the @@ -1353,6 +1511,13 @@ data PState = PState { -- Getting rid of last_loc would require finding another way to -- implement pushCurrentContext (which is only called from one place). +data ALRContext = ALRNoLayout Bool{- does it contain commas? -} + | ALRLayout ALRLayout Int +data ALRLayout = ALRLayoutLet + | ALRLayoutWhere + | ALRLayoutOf + | ALRLayoutDo + newtype P a = P { unP :: PState -> ParseResult a } instance Monad P where @@ -1361,7 +1526,7 @@ instance Monad P where fail = failP returnP :: a -> P a -returnP a = P $ \s -> POk s a +returnP a = a `seq` (P $ \s -> POk s a) thenP :: P a -> (a -> P b) -> P b (P m) `thenP` k = P $ \ s -> @@ -1376,10 +1541,21 @@ failMsgP :: String -> P a failMsgP msg = P $ \s -> PFailed (last_loc s) (text msg) failLocMsgP :: SrcLoc -> SrcLoc -> String -> P a -failLocMsgP loc1 loc2 str = P $ \s -> PFailed (mkSrcSpan loc1 loc2) (text str) +failLocMsgP loc1 loc2 str = P $ \_ -> PFailed (mkSrcSpan loc1 loc2) (text str) + +failSpanMsgP :: SrcSpan -> SDoc -> P a +failSpanMsgP span msg = P $ \_ -> PFailed span msg + +getPState :: P PState +getPState = P $ \s -> POk s s + +getDynFlags :: P DynFlags +getDynFlags = P $ \s -> POk s (dflags s) -failSpanMsgP :: SrcSpan -> String -> P a -failSpanMsgP span msg = P $ \s -> PFailed span (text msg) +withThisPackage :: (PackageId -> a) -> P a +withThisPackage f + = do pkg <- liftM thisPackage getDynFlags + return $ f pkg extension :: (Int -> Bool) -> P Bool extension p = P $ \s -> POk s (p $! extsBitmap s) @@ -1387,6 +1563,9 @@ 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} () @@ -1425,9 +1604,9 @@ 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 + -- character is encountered we output these values -- with the actual character value hidden in the state. | otherwise = case generalCategory c of @@ -1435,20 +1614,20 @@ 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 @@ -1469,7 +1648,7 @@ alexGetChar' (AI loc ofs s) ofs' = advanceOffs c ofs advanceOffs :: Char -> Int -> Int -advanceOffs '\n' offs = 0 +advanceOffs '\n' _ = 0 advanceOffs '\t' offs = (offs `quot` 8 + 1) * 8 advanceOffs _ offs = offs + 1 @@ -1486,60 +1665,147 @@ popLexState :: P Int popLexState = P $ \s@PState{ lex_state=ls:l } -> POk s{ lex_state=l } ls getLexState :: P Int -getLexState = P $ \s@PState{ lex_state=ls:l } -> POk s ls +getLexState = P $ \s@PState{ lex_state=ls:_ } -> POk s ls + +popNextToken :: P (Maybe (Located Token)) +popNextToken + = P $ \s@PState{ alr_next_token = m } -> + POk (s {alr_next_token = Nothing}) m + +setAlrLastLoc :: SrcSpan -> P () +setAlrLastLoc l = P $ \s -> POk (s {alr_last_loc = l}) () + +getAlrLastLoc :: P SrcSpan +getAlrLastLoc = P $ \s@(PState {alr_last_loc = l}) -> POk s l + +getALRContext :: P [ALRContext] +getALRContext = P $ \s@(PState {alr_context = cs}) -> POk s cs + +setALRContext :: [ALRContext] -> P () +setALRContext cs = P $ \s -> POk (s {alr_context = cs}) () + +setNextToken :: Located Token -> P () +setNextToken t = P $ \s -> POk (s {alr_next_token = Just t}) () + +popPendingImplicitToken :: P (Maybe (Located Token)) +popPendingImplicitToken + = P $ \s@PState{ alr_pending_implicit_tokens = ts } -> + case ts of + [] -> POk s Nothing + (t : ts') -> POk (s {alr_pending_implicit_tokens = ts'}) (Just t) + +setPendingImplicitTokens :: [Located Token] -> P () +setPendingImplicitTokens ts = P $ \s -> POk (s {alr_pending_implicit_tokens = ts}) () + +getAlrExpectingOCurly :: P (Maybe ALRLayout) +getAlrExpectingOCurly = P $ \s@(PState {alr_expecting_ocurly = b}) -> POk s b + +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 -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 - -genericsEnabled, ffiEnabled, parrEnabled :: Int -> Bool +transformComprehensionsBit :: Int +transformComprehensionsBit = 17 +qqBit :: Int +qqBit = 18 -- enable quasiquoting +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 + +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 +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 -- 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, @@ -1547,7 +1813,12 @@ pragState buf loc = loc = loc, extsBitmap = 0, context = [], - lex_state = [bol, option_prags, 0] + lex_state = [bol, option_prags, 0], + alr_pending_implicit_tokens = [], + alr_next_token = Nothing, + alr_last_loc = noSrcSpan, + alr_context = [], + alr_expecting_ocurly = Nothing } @@ -1566,30 +1837,38 @@ mkPState buf loc flags = loc = loc, extsBitmap = fromIntegral bitmap, context = [], - lex_state = [bol, 0] + 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, + alr_context = [], + alr_expecting_ocurly = Nothing } 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 - .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags - .|. explicitForallBit `setBitIf` dopt Opt_ScopedTypeVariables flags - .|. explicitForallBit `setBitIf` dopt Opt_PolymorphicComponents flags - .|. explicitForallBit `setBitIf` dopt Opt_ExistentialQuantification flags - .|. explicitForallBit `setBitIf` dopt Opt_Rank2Types flags - .|. explicitForallBit `setBitIf` dopt Opt_RankNTypes 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 - .|. unicodeSyntaxBit `setBitIf` dopt Opt_UnicodeSyntax flags - .|. unboxedTuplesBit `setBitIf` dopt Opt_UnboxedTuples 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 + .|. rawTokenStreamBit `setBitIf` dopt Opt_KeepRawTokenStream flags + .|. newQualOpsBit `setBitIf` dopt Opt_NewQualifiedOperators flags + .|. alternativeLayoutRuleBit `setBitIf` dopt Opt_AlternativeLayoutRule flags -- setBitIf :: Int -> Bool -> Int b `setBitIf` cond | cond = bit b @@ -1613,7 +1892,7 @@ setContext ctx = P $ \s -> POk s{context=ctx} () popContext :: P () popContext = P $ \ s@(PState{ buffer = buf, context = ctx, - loc = loc, last_len = len, last_loc = last_loc }) -> + last_len = len, last_loc = last_loc }) -> case ctx of (_:tl) -> POk s{ context = tl } () [] -> PFailed last_loc (srcParseErr buf len) @@ -1642,8 +1921,8 @@ srcParseErr -> Message srcParseErr buf len = hcat [ if null token - then ptext SLIT("parse error (possibly incorrect indentation)") - else hcat [ptext SLIT("parse error on input "), + then ptext (sLit "parse error (possibly incorrect indentation)") + else hcat [ptext (sLit "parse error on input "), char '`', text token, char '\''] ] where token = lexemeToString (offsetBytes (-len) buf) len @@ -1661,7 +1940,7 @@ srcParseFail = P $ \PState{ buffer = buf, last_len = len, lexError :: String -> P a lexError str = do loc <- getSrcLoc - i@(AI end _ buf) <- getInput + (AI end _ buf) <- getInput reportLexError loc end buf str -- ----------------------------------------------------------------------------- @@ -1670,31 +1949,193 @@ lexError str = do lexer :: (Located Token -> P a) -> P a lexer cont = do - tok@(L span tok__) <- lexToken + alr <- extension alternativeLayoutRule + let lexTokenFun = if alr then lexTokenAlr else lexToken + tok@(L _span _tok__) <- lexTokenFun -- trace ("token: " ++ show tok__) $ do cont tok +lexTokenAlr :: P (Located Token) +lexTokenAlr = do mPending <- popPendingImplicitToken + t <- case mPending of + Nothing -> + do mNext <- popNextToken + t <- case mNext of + Nothing -> lexToken + Just next -> return next + alternativeLayoutRuleToken t + Just t -> + return t + setAlrLastLoc (getLoc t) + case unLoc t of + ITwhere -> setAlrExpectingOCurly (Just ALRLayoutWhere) + ITlet -> setAlrExpectingOCurly (Just ALRLayoutLet) + ITof -> setAlrExpectingOCurly (Just ALRLayoutOf) + ITdo -> setAlrExpectingOCurly (Just ALRLayoutDo) + _ -> return () + return t + +alternativeLayoutRuleToken :: Located Token -> P (Located Token) +alternativeLayoutRuleToken t + = do context <- getALRContext + lastLoc <- getAlrLastLoc + mExpectingOCurly <- getAlrExpectingOCurly + let thisLoc = getLoc t + thisCol = srcSpanStartCol thisLoc + newLine = (lastLoc == noSrcSpan) + || (srcSpanStartLine thisLoc > srcSpanEndLine lastLoc) + case (unLoc t, context, mExpectingOCurly) of + -- I think our implicit open-curly handling is slightly + -- different to John's, in how it interacts with newlines + -- and "in" + (ITocurly, _, Just _) -> + do setAlrExpectingOCurly Nothing + setNextToken t + lexTokenAlr + (_, ALRLayout _ col : ls, Just expectingOCurly) + | (thisCol > col) || + (thisCol == col && + isNonDecreasingIntentation expectingOCurly) -> + do setAlrExpectingOCurly Nothing + setALRContext (ALRLayout expectingOCurly thisCol : context) + setNextToken t + return (L thisLoc ITocurly) + | otherwise -> + do setAlrExpectingOCurly Nothing + setPendingImplicitTokens [L lastLoc ITccurly] + setNextToken t + return (L lastLoc ITocurly) + (_, _, Just expectingOCurly) -> + do setAlrExpectingOCurly Nothing + setALRContext (ALRLayout expectingOCurly thisCol : context) + setNextToken t + return (L thisLoc ITocurly) + -- We do the [] cases earlier than in the spec, as we + -- have an actual EOF token + (ITeof, ALRLayout _ _ : ls, _) -> + do setALRContext ls + setNextToken t + return (L thisLoc ITccurly) + (ITeof, _, _) -> + return t + -- the other ITeof case omitted; general case below covers it + (ITin, ALRLayout ALRLayoutLet _ : ls, _) + | newLine -> + do setPendingImplicitTokens [t] + setALRContext ls + return (L thisLoc ITccurly) + (_, ALRLayout _ col : ls, _) + | newLine && thisCol == col -> + do setNextToken t + return (L thisLoc ITsemi) + | newLine && thisCol < col -> + do setALRContext ls + setNextToken t + -- Note that we use lastLoc, as we may need to close + -- more layouts, or give a semicolon + return (L lastLoc ITccurly) + (u, _, _) + | isALRopen u -> + do setALRContext (ALRNoLayout (containsCommas u) : context) + return t + (u, _, _) + | isALRclose u -> + case context of + ALRLayout _ _ : ls -> + do setALRContext ls + setNextToken t + return (L thisLoc ITccurly) + ALRNoLayout _ : ls -> + do setALRContext ls + return t + [] -> + -- XXX This is an error in John's code, but + -- it looks reachable to me at first glance + return t + (ITin, ALRLayout ALRLayoutLet _ : ls, _) -> + do setALRContext ls + setPendingImplicitTokens [t] + return (L thisLoc ITccurly) + (ITin, ALRLayout _ _ : ls, _) -> + do setALRContext ls + setNextToken t + return (L thisLoc ITccurly) + -- the other ITin case omitted; general case below covers it + (ITcomma, ALRLayout _ _ : ls, _) + | topNoLayoutContainsCommas ls -> + do setALRContext ls + setNextToken t + return (L thisLoc ITccurly) + (ITwhere, ALRLayout ALRLayoutDo _ : ls, _) -> + do setALRContext ls + setPendingImplicitTokens [t] + return (L thisLoc ITccurly) + -- the other ITwhere case omitted; general case below covers it + (_, _, _) -> return t + +isALRopen :: Token -> Bool +isALRopen ITcase = True +isALRopen ITif = True +isALRopen IToparen = True +isALRopen ITobrack = True +isALRopen ITocurly = True +-- GHC Extensions: +isALRopen IToubxparen = True +isALRopen _ = False + +isALRclose :: Token -> Bool +isALRclose ITof = True +isALRclose ITthen = True +isALRclose ITcparen = True +isALRclose ITcbrack = True +isALRclose ITccurly = True +-- GHC Extensions: +isALRclose ITcubxparen = True +isALRclose _ = False + +isNonDecreasingIntentation :: ALRLayout -> Bool +isNonDecreasingIntentation ALRLayoutDo = True +isNonDecreasingIntentation _ = False + +containsCommas :: Token -> Bool +containsCommas IToparen = True +containsCommas ITobrack = True +-- John doesn't have {} as containing commas, but records contain them, +-- which caused a problem parsing Cabal's Distribution.Simple.InstallDirs +-- (defaultInstallDirs). +containsCommas ITocurly = True +-- GHC Extensions: +containsCommas IToubxparen = True +containsCommas _ = False + +topNoLayoutContainsCommas :: [ALRContext] -> Bool +topNoLayoutContainsCommas [] = False +topNoLayoutContainsCommas (ALRLayout _ _ : ls) = topNoLayoutContainsCommas ls +topNoLayoutContainsCommas (ALRNoLayout b : _) = b + lexToken :: P (Located Token) lexToken = do inp@(AI loc1 _ buf) <- getInput sc <- getLexState exts <- getExts case alexScanUser exts inp sc of - AlexEOF -> do let span = mkSrcSpan loc1 loc1 - setLastToken span 0 0 - return (L span ITeof) - AlexError (AI loc2 _ buf) -> do - reportLexError loc1 loc2 buf "lexical error" + AlexEOF -> do + let span = mkSrcSpan loc1 loc1 + setLastToken span 0 0 + return (L span ITeof) + AlexError (AI loc2 _ buf) -> + reportLexError loc1 loc2 buf "lexical error" AlexSkip inp2 _ -> do - setInput inp2 - lexToken - AlexToken inp2@(AI end _ buf2) len t -> do - setInput inp2 - let span = mkSrcSpan loc1 end - let bytes = byteDiff buf buf2 - span `seq` setLastToken span bytes bytes - t span buf bytes - + setInput inp2 + lexToken + AlexToken inp2@(AI end _ buf2) _ t -> do + setInput inp2 + let span = mkSrcSpan loc1 end + let bytes = byteDiff buf buf2 + 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 = @@ -1704,4 +2145,68 @@ 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 + +linePrags = Map.singleton "line" (begin line_prag2) + +fileHeaderPrags = Map.fromList([("options", lex_string_prag IToptions_prag), + ("options_ghc", lex_string_prag IToptions_prag), + ("options_haddock", lex_string_prag ITdocOptions), + ("language", token ITlanguage_prag), + ("include", lex_string_prag ITinclude_prag)]) + +ignoredPrags = Map.fromList (map ignored pragmas) + where ignored opt = (opt, nested_comment lexToken) + impls = ["hugs", "nhc98", "jhc", "yhc", "catch", "derive"] + options_pragmas = map ("options_" ++) impls + -- CFILES is a hugs-only thing. + pragmas = options_pragmas ++ ["cfiles", "contract"] + +oneWordPrags = Map.fromList([("rules", rulePrag), + ("inline", token (ITinline_prag True)), + ("notinline", token (ITinline_prag False)), + ("specialize", token ITspec_prag), + ("source", token ITsource_prag), + ("warning", token ITwarning_prag), + ("deprecated", token ITdeprecated_prag), + ("scc", token ITscc_prag), + ("generated", token ITgenerated_prag), + ("core", token ITcore_prag), + ("unpack", token ITunpack_prag), + ("ann", token ITann_prag)]) + +twoWordPrags = Map.fromList([("inline conlike", token (ITinline_conlike_prag True)), + ("notinline conlike", token (ITinline_conlike_prag False)), + ("specialize inline", token (ITspec_inline_prag True)), + ("specialize notinline", token (ITspec_inline_prag False))]) + + +dispatch_pragmas :: Map String Action -> Action +dispatch_pragmas prags span buf len = case Map.lookup (clean_pragma (lexemeToString buf len)) prags of + Just found -> found span buf len + Nothing -> lexError "unknown pragma" + +known_pragma :: Map String Action -> AlexAccPred Int +known_pragma prags _ _ len (AI _ _ buf) = (isJust $ Map.lookup (clean_pragma (lexemeToString (offsetBytes (- len) buf) len)) prags) + && (nextCharIs buf (\c -> not (isAlphaNum c || c == '_'))) + +clean_pragma :: String -> String +clean_pragma prag = canon_ws (map toLower (unprefix prag)) + where unprefix prag' = case stripPrefix "{-#" prag' of + Just rest -> rest + Nothing -> prag' + canonical prag' = case prag' of + "noinline" -> "notinline" + "specialise" -> "specialize" + "constructorlike" -> "conlike" + _ -> prag' + canon_ws s = unwords (map canonical (words s)) }