X-Git-Url: http://git.megacz.com/?a=blobdiff_plain;f=ghc%2Fcompiler%2Fparser%2FLexer.x;h=4c1b48efc0845f3009dc19f4d8e5e26eadb36103;hb=274a7b1ab7cf30fd1ceaddd02f991eaf694cc37c;hp=2d01a18967281e3e4089656edb9548b67b672945;hpb=42be1de4d0cc2c72945a1948a6489376cc24e692;p=ghc-hetmet.git diff --git a/ghc/compiler/parser/Lexer.x b/ghc/compiler/parser/Lexer.x index 2d01a18..4c1b48e 100644 --- a/ghc/compiler/parser/Lexer.x +++ b/ghc/compiler/parser/Lexer.x @@ -1,5 +1,5 @@ ----------------------------------------------------------------------------- --- (c) The University of Glasgow, 2003 +-- (c) The University of Glasgow, 2006 -- -- GHC's lexer. -- @@ -22,10 +22,12 @@ { module Lexer ( - Token(..), lexer, mkPState, + Token(..), lexer, pragState, mkPState, PState(..), P(..), ParseResult(..), getSrcLoc, - failMsgP, failLocMsgP, failSpanMsgP, srcParseFail, - popContext, pushCurrentContext, + failLocMsgP, failSpanMsgP, srcParseFail, + popContext, pushCurrentContext, setLastToken, setSrcLoc, + getLexState, popLexState, pushLexState, + extension, bangPatEnabled ) where #include "HsVersions.h" @@ -37,40 +39,49 @@ import FastString import FastTypes import SrcLoc import UniqFM -import CmdLineOpts +import DynFlags import Ctype -import Util ( maybePrefixMatch ) +import Util ( maybePrefixMatch, readRational ) import DATA_BITS -import Char +import Data.Char ( chr ) import Ratio --import TRACE + +#if __GLASGOW_HASKELL__ >= 605 +import Data.Char ( GeneralCategory(..), generalCategory, isPrint, isUpper ) +#else +import Compat.Unicode ( GeneralCategory(..), generalCategory, isPrint, isUpper ) +#endif } -$whitechar = [\ \t\n\r\f\v\xa0] +$unispace = \x05 +$whitechar = [\ \t\n\r\f\v\xa0 $unispace] $white_no_nl = $whitechar # \n $ascdigit = 0-9 -$unidigit = \x01 +$unidigit = \x03 +$decdigit = $ascdigit -- for now, should really be $digit (ToDo) $digit = [$ascdigit $unidigit] $special = [\(\)\,\;\[\]\`\{\}] $ascsymbol = [\!\#\$\%\&\*\+\.\/\<\=\>\?\@\\\^\|\-\~] -$unisymbol = \x02 +$unisymbol = \x04 $symbol = [$ascsymbol $unisymbol] # [$special \_\:\"\'] -$unilarge = \x03 +$unilarge = \x01 $asclarge = [A-Z \xc0-\xd6 \xd8-\xde] $large = [$asclarge $unilarge] -$unismall = \x04 +$unismall = \x02 $ascsmall = [a-z \xdf-\xf6 \xf8-\xff] $small = [$ascsmall $unismall \_] -$graphic = [$small $large $symbol $digit $special \:\"\'] +$unigraphic = \x06 +$graphic = [$small $large $symbol $digit $special $unigraphic \:\"\'] $octit = 0-7 -$hexit = [$digit A-F a-f] +$hexit = [$decdigit A-F a-f] $symchar = [$symbol \:] $nl = [\n\r] $idchar = [$small $large $digit \'] @@ -81,7 +92,7 @@ $idchar = [$small $large $digit \'] @varsym = $symbol $symchar* @consym = \: $symchar* -@decimal = $digit+ +@decimal = $decdigit+ @octal = $octit+ @hexadecimal = $hexit+ @exponent = [eE] [\-\+]? @decimal @@ -108,7 +119,7 @@ $white_no_nl+ ; -- have to exclude those. -- The regex says: "munch all the characters after the dashes, as long as -- the first one is not a symbol". -"--"\-* [^$symbol] .* ; +"--"\-* [^$symbol :] .* ; "--"\-* / { atEOL } ; -- 'bol' state: beginning of a line. Slurp up all the whitespace (including @@ -124,6 +135,7 @@ $white_no_nl+ ; \n ; ^\# (line)? { begin line_prag1 } ^\# pragma .* \n ; -- GCC 3.3 CPP generated, apparently + ^\# \! .* \n ; -- #!, for scripts () { do_bol } } @@ -146,44 +158,63 @@ $white_no_nl+ ; -- generate a matching '}' token. () { do_layout_left } -<0,glaexts> \n { begin bol } +<0,option_prags,glaexts> \n { begin bol } "{-#" $whitechar* (line|LINE) { begin line_prag2 } -- single-line line pragmas, of the form -- # "" \n - $digit+ { set_line line_prag1a } - \" [$graphic \ ]* \" { set_file line_prag1b } + $decdigit+ { setLine line_prag1a } + \" [$graphic \ ]* \" { setFile line_prag1b } .* { pop } -- Haskell-style line pragmas, of the form -- {-# LINE "" #-} - $digit+ { set_line line_prag2a } - \" [$graphic \ ]* \" { set_file line_prag2b } + $decdigit+ { setLine line_prag2a } + \" [$graphic \ ]* \" { setFile line_prag2b } "#-}"|"-}" { pop } -- NOTE: accept -} at the end of a LINE pragma, for compatibility -- with older versions of GHC which generated these. -<0,glaexts> { +-- We only want RULES pragmas to be picked up when -fglasgow-exts +-- is on, because the contents of the pragma is always written using +-- glasgow-exts syntax (using forall etc.), so if glasgow exts are not +-- enabled, we're sure to get a parse error. +-- (ToDo: we should really emit a warning when ignoring pragmas) + + "{-#" $whitechar* (RULES|rules) { token ITrules_prag } + +<0,option_prags,glaexts> { + "{-#" $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 ITspecialise_prag } + { 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* (INLINE|inline) { token ITinline_prag } - "{-#" $whitechar* (NO(T?)INLINE|no(t?)inline) - { token ITnoinline_prag } - "{-#" $whitechar* (RULES|rules) { token ITrules_prag } "{-#" $whitechar* (DEPRECATED|deprecated) { token ITdeprecated_prag } "{-#" $whitechar* (SCC|scc) { token ITscc_prag } "{-#" $whitechar* (CORE|core) { token ITcore_prag } "{-#" $whitechar* (UNPACK|unpack) { token ITunpack_prag } - + "{-#" { nested_comment } -- ToDo: should only be valid inside a pragma: "#-}" { token ITclose_prag} } + { + "{-#" $whitechar* (OPTIONS|options) { lex_string_prag IToptions_prag } + "{-#" $whitechar* (OPTIONS_GHC|options_ghc) + { lex_string_prag IToptions_prag } + "{-#" $whitechar* (LANGUAGE|language) { token ITlanguage_prag } + "{-#" $whitechar* (INCLUDE|include) { lex_string_prag ITinclude_prag } +} -- '0' state: ordinary lexemes -- 'glaexts' state: glasgow extensions (postfix '#', etc.) @@ -224,7 +255,7 @@ $white_no_nl+ ; "|}" { token ITccurlybar } } -<0,glaexts> { +<0,option_prags,glaexts> { \( { special IToparen } \) { special ITcparen } \[ { special ITobrack } @@ -237,7 +268,7 @@ $white_no_nl+ ; \} { close_brace } } -<0,glaexts> { +<0,option_prags,glaexts> { @qual @varid { check_qvarid } @qual @conid { idtoken qconid } @varid { varid } @@ -341,10 +372,11 @@ data Token | ITdotnet | ITmdo - | ITspecialise_prag -- Pragmas + -- Pragmas + | ITinline_prag Bool -- True <=> INLINE, False <=> NOINLINE + | ITspec_prag -- SPECIALISE + | ITspec_inline_prag Bool -- SPECIALISE INLINE (or NOINLINE) | ITsource_prag - | ITinline_prag - | ITnoinline_prag | ITrules_prag | ITdeprecated_prag | ITline_prag @@ -352,6 +384,9 @@ data Token | ITcore_prag -- hdaume: core annotations | ITunpack_prag | ITclose_prag + | IToptions_prag String + | ITinclude_prag String + | ITlanguage_prag | ITdotdot -- reserved symbols | ITcolon @@ -416,25 +451,25 @@ data Token | ITprimdouble Rational -- MetaHaskell extension tokens - | ITopenExpQuote -- [| or [e| - | ITopenPatQuote -- [p| - | ITopenDecQuote -- [d| - | ITopenTypQuote -- [t| - | ITcloseQuote -- |] - | ITidEscape FastString -- $x - | ITparenEscape -- $( - | ITvarQuote -- ' - | ITtyQuote -- '' + | ITopenExpQuote -- [| or [e| + | ITopenPatQuote -- [p| + | ITopenDecQuote -- [d| + | ITopenTypQuote -- [t| + | ITcloseQuote -- |] + | ITidEscape FastString -- $x + | ITparenEscape -- $( + | ITvarQuote -- ' + | ITtyQuote -- '' -- Arrow notation extension | ITproc | ITrec - | IToparenbar -- (| - | ITcparenbar -- |) - | ITlarrowtail -- -< - | ITrarrowtail -- >- - | ITLarrowtail -- -<< - | ITRarrowtail -- >>- + | IToparenbar -- (| + | ITcparenbar -- |) + | ITlarrowtail -- -< + | ITrarrowtail -- >- + | ITLarrowtail -- -<< + | ITRarrowtail -- >>- | ITunknown String -- Used when the lexer can't make sense of it | ITeof -- end of file token @@ -497,7 +532,7 @@ reservedWordsFM = listToUFM $ ( "where", ITwhere, 0 ), ( "_scc_", ITscc, 0 ), -- ToDo: remove - ( "forall", ITforall, bit glaExtsBit), + ( "forall", ITforall, bit tvBit), ( "mdo", ITmdo, bit glaExtsBit), ( "foreign", ITforeign, bit ffiBit), @@ -533,12 +568,22 @@ reservedSymsFM = listToUFM $ ,("!", ITbang, 0) ,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT - ,(".", ITdot, bit glaExtsBit) -- For 'forall a . t' + ,(".", ITdot, bit tvBit) -- For 'forall a . t' ,("-<", ITlarrowtail, bit arrowsBit) ,(">-", ITrarrowtail, bit arrowsBit) ,("-<<", ITLarrowtail, bit arrowsBit) ,(">>-", ITRarrowtail, bit arrowsBit) + +#if __GLASGOW_HASKELL__ >= 605 + ,("λ", ITlam, bit glaExtsBit) + ,("∷", ITdcolon, bit glaExtsBit) + ,("⇒", ITdarrow, bit glaExtsBit) + ,("∀", ITforall, bit glaExtsBit) + ,("→", ITrarrow, bit glaExtsBit) + ,("←", ITlarrow, bit glaExtsBit) + ,("⋯", ITdotdot, bit glaExtsBit) +#endif ] -- ----------------------------------------------------------------------------- @@ -578,12 +623,12 @@ pop _span _buf _len = do popLexState; lexToken pop_and :: Action -> Action pop_and act span buf len = do popLexState; act span buf len -notFollowedBy char _ _ _ (_,buf) = atEnd buf || currentChar buf /= char +notFollowedBy char _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf /= char -notFollowedBySymbol _ _ _ (_,buf) +notFollowedBySymbol _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf `notElem` "!#$%&*+./<=>?@\\^|-~" -atEOL _ _ _ (_,buf) = atEnd buf || currentChar buf == '\n' +atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n' ifExtension pred bits _ _ _ = pred bits @@ -613,8 +658,7 @@ nested_comment span _str _len = do Just (c,input) -> go n input c -> go n input - err input = do failLocMsgP (srcSpanStart span) (fst input) - "unterminated `{-'" + err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated `{-'" open_brace, close_brace :: Action open_brace span _str _len = do @@ -644,7 +688,8 @@ check_qvarid span buf len = do token = L span (ITqvarid (mod,var)) try_again = do - setInput (srcSpanStart span,buf) + (AI _ offs _) <- getInput + setInput (AI (srcSpanStart span) (offs-len) buf) pushLexState bad_qvarid lexToken @@ -655,23 +700,29 @@ splitQualName :: StringBuffer -> Int -> (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 0 0 +splitQualName orig_buf len = split orig_buf orig_buf where - split buf dot_off n - | n == len = done dot_off - | lookAhead buf n == '.' = split2 buf n (n+1) - | otherwise = split buf dot_off (n+1) + split buf dot_buf + | orig_buf `byteDiff` buf >= len = done dot_buf + | c == '.' = found_dot buf' + | otherwise = split buf' dot_buf + where + (c,buf') = nextChar buf -- careful, we might get names like M.... -- so, if the character after the dot is not upper-case, this is -- the end of the qualifier part. - split2 buf dot_off n - | isUpper (lookAhead buf n) = split buf dot_off (n+1) - | otherwise = done dot_off - - done dot_off = - (lexemeToFastString orig_buf dot_off, - lexemeToFastString (stepOnBy (dot_off+1) orig_buf) (len - dot_off -1)) + found_dot buf -- buf points after the '.' + | isUpper c = split buf' buf + | otherwise = done buf + where + (c,buf') = nextChar buf + + done dot_buf = + (lexemeToFastString orig_buf (qual_size - 1), + lexemeToFastString dot_buf (len - qual_size)) + where + qual_size = orig_buf `byteDiff` dot_buf varid span buf len = case lookupUFM reservedWordsFM fs of @@ -708,32 +759,26 @@ sym con span buf len = fs = lexemeToFastString buf len tok_decimal span buf len - = return (L span (ITinteger $! parseInteger buf len 10 oct_or_dec)) + = return (L span (ITinteger $! parseInteger buf len 10 octDecDigit)) tok_octal span buf len - = return (L span (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 8 oct_or_dec)) + = return (L span (ITinteger $! parseInteger (offsetBytes 2 buf) (len-2) 8 octDecDigit)) tok_hexadecimal span buf len - = return (L span (ITinteger $! parseInteger (stepOnBy 2 buf) (len-2) 16 hex)) + = return (L span (ITinteger $! parseInteger (offsetBytes 2 buf) (len-2) 16 hexDigit)) prim_decimal span buf len - = return (L span (ITprimint $! parseInteger buf (len-1) 10 oct_or_dec)) + = return (L span (ITprimint $! parseInteger buf (len-1) 10 octDecDigit)) prim_octal span buf len - = return (L span (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 8 oct_or_dec)) + = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 8 octDecDigit)) prim_hexadecimal span buf len - = return (L span (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 16 hex)) + = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 16 hexDigit)) -tok_float str = ITrational $! readRational__ str -prim_float str = ITprimfloat $! readRational__ str -prim_double str = ITprimdouble $! readRational__ str - -parseInteger :: StringBuffer -> Int -> Integer -> (Char->Int) -> Integer -parseInteger buf len radix to_int - = go 0 0 - where go i x | i == len = x - | otherwise = go (i+1) (x * radix + toInteger (to_int (lookAhead buf i))) +tok_float str = ITrational $! readRational str +prim_float str = ITprimfloat $! readRational str +prim_double str = ITprimdouble $! readRational str -- ----------------------------------------------------------------------------- -- Layout processing @@ -741,7 +786,7 @@ parseInteger buf len radix to_int -- we're at the first token on a line, insert layout tokens if necessary do_bol :: Action do_bol span _str _len = do - pos <- getOffside (srcSpanEnd span) + pos <- getOffside case pos of LT -> do --trace "layout: inserting '}'" $ do @@ -777,7 +822,7 @@ maybe_layout _ = return () -- new_layout_context strict span _buf _len = do popLexState - let offset = srcSpanStartCol span + (AI _ offset _) <- getInput ctx <- getContext case ctx of Layout prev_off : _ | @@ -799,23 +844,49 @@ do_layout_left span _buf _len = do -- ----------------------------------------------------------------------------- -- LINE pragmas -set_line :: Int -> Action -set_line code span buf len = do - let line = parseInteger buf len 10 oct_or_dec +setLine :: Int -> Action +setLine code span buf len = do + let line = parseInteger buf len 10 octDecDigit setSrcLoc (mkSrcLoc (srcSpanFile span) (fromIntegral line - 1) 0) -- subtract one: the line number refers to the *following* line popLexState pushLexState code lexToken -set_file :: Int -> Action -set_file code span buf len = do +setFile :: Int -> Action +setFile code span buf len = do let file = lexemeToFastString (stepOn buf) (len-2) setSrcLoc (mkSrcLoc file (srcSpanEndLine span) (srcSpanEndCol span)) popLexState pushLexState code lexToken + +-- ----------------------------------------------------------------------------- +-- Options, includes and language pragmas. + +lex_string_prag :: (String -> Token) -> Action +lex_string_prag mkTok span buf len + = do input <- getInput + start <- getSrcLoc + tok <- go [] input + end <- getSrcLoc + return (L (mkSrcSpan start end) tok) + where go acc input + = if isString input "#-}" + then do setInput input + return (mkTok (reverse acc)) + else case alexGetChar input of + Just (c,i) -> go (c:acc) i + Nothing -> err input + isString i [] = True + isString i (x:xs) + = case alexGetChar i of + Just (c,i') | c == x -> isString i' xs + _other -> False + err (AI end _ _) = failLocMsgP (srcSpanStart span) end "unterminated options pragma" + + -- ----------------------------------------------------------------------------- -- Strings & Chars @@ -830,7 +901,7 @@ lex_string_tok span buf len = do lex_string :: String -> P Token lex_string s = do i <- getInput - case alexGetChar i of + case alexGetChar' i of Nothing -> lit_error Just ('"',i) -> do @@ -839,14 +910,15 @@ lex_string s = do if glaexts then do i <- getInput - case alexGetChar i of + case alexGetChar' i of Just ('#',i) -> do setInput i if any (> '\xFF') s then failMsgP "primitive string literal must contain only characters <= \'\\xFF\'" - else let s' = mkFastStringNarrow (reverse s) in - -- always a narrow string/byte array + else let s' = mkZFastString (reverse s) in return (ITprimstring s') + -- mkZFastString is a hack to avoid encoding the + -- string in UTF-8. We just want the exact bytes. _other -> return (ITstring (mkFastString (reverse s))) else @@ -857,11 +929,11 @@ lex_string s = do setInput i; lex_string s | Just (c,i) <- next, is_space c -> do setInput i; lex_stringgap s - where next = alexGetChar i + where next = alexGetChar' i - Just _ -> do - c <- lex_char - lex_string (c:s) + Just (c, i) -> do + c' <- lex_char c i + lex_string (c':s) lex_stringgap s = do c <- getCharOrFail @@ -881,29 +953,30 @@ lex_char_tok :: Action 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 + case alexGetChar' i1 of Nothing -> lit_error - Just ('\'', i2@(end2,_)) -> do -- We've seen '' + Just ('\'', i2@(AI end2 _ _)) -> do -- We've seen '' th_exts <- extension thEnabled if th_exts then do setInput i2 return (L (mkSrcSpan loc end2) ITtyQuote) else lit_error - Just ('\\', i2@(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 lit_error + else do setInput i2; lit_error - Just (c, i2@(end2,_)) | not (is_any c) -> lit_error - | otherwise -> + 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 + case alexGetChar' i2 of -- Look ahead one more character Nothing -> lit_error Just ('\'', i3) -> do -- We've seen 'x' setInput i3 @@ -911,17 +984,18 @@ lex_char_tok span buf len = do -- We've seen ' _other -> do -- We've seen 'x not followed by quote -- If TH is on, just parse the quote only th_exts <- extension thEnabled - if th_exts then return (L (mkSrcSpan loc (fst i1)) ITvarQuote) - else lit_error + let (AI end _ _) = i1 + if th_exts then return (L (mkSrcSpan loc end) ITvarQuote) + else do setInput i2; lit_error finish_char_tok :: SrcLoc -> Char -> P (Located Token) finish_char_tok loc ch -- We've already seen the closing quote -- Just need to check for trailing # = do glaexts <- extension glaExtsEnabled - i@(end,_) <- getInput + i@(AI end _ _) <- getInput if glaexts then do - case alexGetChar i of - Just ('#',i@(end,_)) -> do + case alexGetChar' i of + Just ('#',i@(AI end _ _)) -> do setInput i return (L (mkSrcSpan loc end) (ITprimchar ch)) _other -> @@ -929,14 +1003,16 @@ finish_char_tok loc ch -- We've already seen the closing quote else do return (L (mkSrcSpan loc end) (ITchar ch)) -lex_char :: P Char -lex_char = do - mc <- getCharOrFail - case mc of - '\\' -> lex_escape - c | is_any c -> return c +lex_char :: Char -> AlexInput -> P Char +lex_char c inp = do + case c of + '\\' -> do setInput inp; lex_escape + c | isAny c -> do setInput inp; return c _other -> lit_error +isAny c | c > '\xff' = isPrint c + | otherwise = is_any c + lex_escape :: P Char lex_escape = do c <- getCharOrFail @@ -956,17 +1032,17 @@ lex_escape = do then return (chr (ord c - ord '@')) else lit_error - 'x' -> readNum is_hexdigit 16 hex - 'o' -> readNum is_octdigit 8 oct_or_dec - x | is_digit x -> readNum2 is_digit 10 oct_or_dec (oct_or_dec x) + 'x' -> readNum is_hexdigit 16 hexDigit + 'o' -> readNum is_octdigit 8 octDecDigit + x | is_digit x -> readNum2 is_digit 10 octDecDigit (octDecDigit x) c1 -> do i <- getInput - case alexGetChar i of + case alexGetChar' i of Nothing -> lit_error Just (c2,i2) -> - case alexGetChar i2 of - Nothing -> lit_error + case alexGetChar' i2 of + Nothing -> do setInput i2; lit_error Just (c3,i3) -> let str = [c1,c2,c3] in case [ (c,rest) | (p,c) <- silly_escape_chars, @@ -981,40 +1057,24 @@ lex_escape = do readNum :: (Char -> Bool) -> Int -> (Char -> Int) -> P Char readNum is_digit base conv = do + i <- getInput c <- getCharOrFail if is_digit c then readNum2 is_digit base conv (conv c) - else lit_error + else do setInput i; lit_error readNum2 is_digit base conv i = do input <- getInput read i input where read i input = do - case alexGetChar input of + case alexGetChar' input of Just (c,input') | is_digit c -> do read (i*base + conv c) input' _other -> do - setInput input if i >= 0 && i <= 0x10FFFF - then return (chr i) + then do setInput input; return (chr i) else lit_error -is_hexdigit c - = is_digit c - || (c >= 'a' && c <= 'f') - || (c >= 'A' && c <= 'F') - -hex c | is_digit c = ord c - ord '0' - | otherwise = ord (to_lower c) - ord 'a' + 10 - -oct_or_dec c = ord c - ord '0' - -is_octdigit c = c >= '0' && c <= '7' - -to_lower c - | c >= 'A' && c <= 'Z' = chr (ord c - (ord 'A' - ord 'a')) - | otherwise = c - silly_escape_chars = [ ("NUL", '\NUL'), ("SOH", '\SOH'), @@ -1052,64 +1112,20 @@ silly_escape_chars = [ ("DEL", '\DEL') ] +-- before calling lit_error, ensure that the current input is pointing to +-- 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 = lexError "lexical error in string/character literal" getCharOrFail :: P Char getCharOrFail = do i <- getInput - case alexGetChar i of + case alexGetChar' i of Nothing -> lexError "unexpected end-of-file in string/character literal" Just (c,i) -> do setInput i; return c -- ----------------------------------------------------------------------------- --- Floats - -readRational :: ReadS Rational -- NB: doesn't handle leading "-" -readRational r = do - (n,d,s) <- readFix r - (k,t) <- readExp s - return ((n%1)*10^^(k-d), t) - where - readFix r = do - (ds,s) <- lexDecDigits r - (ds',t) <- lexDotDigits s - return (read (ds++ds'), length ds', t) - - readExp (e:s) | e `elem` "eE" = readExp' s - readExp s = return (0,s) - - readExp' ('+':s) = readDec s - readExp' ('-':s) = do - (k,t) <- readDec s - return (-k,t) - readExp' s = readDec s - - readDec s = do - (ds,r) <- nonnull isDigit s - return (foldl1 (\n d -> n * 10 + d) [ ord d - ord '0' | d <- ds ], - r) - - lexDecDigits = nonnull isDigit - - lexDotDigits ('.':s) = return (span isDigit s) - lexDotDigits s = return ("",s) - - nonnull p s = do (cs@(_:_),t) <- return (span p s) - return (cs,t) - -readRational__ :: String -> Rational -- NB: *does* handle a leading "-" -readRational__ top_s - = case top_s of - '-' : xs -> - (read_me xs) - xs -> read_me xs - where - read_me s - = case (do { (x,"") <- readRational s ; return x }) of - [x] -> x - [] -> error ("readRational__: no parse:" ++ top_s) - _ -> error ("readRational__: ambiguous parse:" ++ top_s) - --- ----------------------------------------------------------------------------- -- The Parse Monad data LayoutContext @@ -1127,6 +1143,9 @@ data ParseResult a data PState = PState { buffer :: StringBuffer, 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 loc :: SrcLoc, -- current loc (end of prev token + 1) extsBitmap :: !Int, -- bitmap that determines permitted extensions @@ -1134,7 +1153,10 @@ data PState = PState { lex_state :: [Int] } -- last_loc and last_len are used when generating error messages, - -- and in pushCurrentContext only. + -- and in pushCurrentContext only. Sigh, if only Happy passed the + -- current token to happyError, we could at least get rid of last_len. + -- Getting rid of last_loc would require finding another way to + -- implement pushCurrentContext (which is only called from one place). newtype P a = P { unP :: PState -> ParseResult a } @@ -1179,24 +1201,79 @@ getSrcLoc = P $ \s@(PState{ loc=loc }) -> POk s loc setLastToken :: SrcSpan -> Int -> P () setLastToken loc len = P $ \s -> POk s{ last_loc=loc, last_len=len } () -type AlexInput = (SrcLoc,StringBuffer) +data AlexInput = AI SrcLoc {-#UNPACK#-}!Int StringBuffer alexInputPrevChar :: AlexInput -> Char -alexInputPrevChar (_,s) = prevChar s '\n' +alexInputPrevChar (AI _ _ buf) = prevChar buf '\n' alexGetChar :: AlexInput -> Maybe (Char,AlexInput) -alexGetChar (loc,s) +alexGetChar (AI loc ofs s) + | atEnd s = Nothing + | otherwise = adj_c `seq` loc' `seq` ofs' `seq` s' `seq` + Just (adj_c, (AI loc' ofs' s')) + where (c,s') = nextChar s + loc' = advanceSrcLoc loc c + ofs' = advanceOffs c ofs + + non_graphic = '\x0' + upper = '\x1' + lower = '\x2' + digit = '\x3' + symbol = '\x4' + space = '\x5' + other_graphic = '\x6' + + adj_c + | c <= '\x06' = non_graphic + | c <= '\xff' = c + | otherwise = + case generalCategory c of + UppercaseLetter -> upper + LowercaseLetter -> lower + TitlecaseLetter -> upper + ModifierLetter -> other_graphic + OtherLetter -> other_graphic + NonSpacingMark -> other_graphic + SpacingCombiningMark -> other_graphic + EnclosingMark -> other_graphic + DecimalNumber -> digit + LetterNumber -> other_graphic + OtherNumber -> other_graphic + ConnectorPunctuation -> other_graphic + DashPunctuation -> other_graphic + OpenPunctuation -> other_graphic + ClosePunctuation -> other_graphic + InitialQuote -> other_graphic + FinalQuote -> other_graphic + OtherPunctuation -> other_graphic + MathSymbol -> symbol + CurrencySymbol -> symbol + ModifierSymbol -> symbol + OtherSymbol -> symbol + Space -> space + _other -> non_graphic + +-- This version does not squash unicode characters, it is used when +-- lexing strings. +alexGetChar' :: AlexInput -> Maybe (Char,AlexInput) +alexGetChar' (AI loc ofs s) | atEnd s = Nothing - | otherwise = c `seq` loc' `seq` s' `seq` Just (c, (loc', s')) - where c = currentChar s - loc' = advanceSrcLoc loc c - s' = stepOn s + | otherwise = c `seq` loc' `seq` ofs' `seq` s' `seq` + Just (c, (AI loc' ofs' s')) + where (c,s') = nextChar s + loc' = advanceSrcLoc loc c + ofs' = advanceOffs c ofs + +advanceOffs :: Char -> Int -> Int +advanceOffs '\n' offs = 0 +advanceOffs '\t' offs = (offs `quot` 8 + 1) * 8 +advanceOffs _ offs = offs + 1 getInput :: P AlexInput -getInput = P $ \s@PState{ loc=l, buffer=b } -> POk s (l,b) +getInput = P $ \s@PState{ loc=l, last_offs=o, buffer=b } -> POk s (AI l o b) setInput :: AlexInput -> P () -setInput (l,b) = P $ \s -> POk s{ loc=l, buffer=b } () +setInput (AI l o b) = P $ \s -> POk s{ loc=l, last_offs=o, buffer=b } () pushLexState :: Int -> P () pushLexState ls = P $ \s@PState{ lex_state=l } -> POk s{lex_state=ls:l} () @@ -1218,6 +1295,9 @@ parrBit = 2 arrowsBit = 4 thBit = 5 ipBit = 6 +tvBit = 7 -- Scoped type variables enables 'forall' keyword +bangPatBit = 8 -- Tells the parser to understand bang-patterns + -- (doesn't affect the lexer) glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool glaExtsEnabled flags = testBit flags glaExtsBit @@ -1226,6 +1306,24 @@ parrEnabled flags = testBit flags parrBit arrowsEnabled flags = testBit flags arrowsBit thEnabled flags = testBit flags thBit ipEnabled flags = testBit flags ipBit +tvEnabled flags = testBit flags tvBit +bangPatEnabled flags = testBit flags bangPatBit + +-- PState for parsing options pragmas +-- +pragState :: StringBuffer -> SrcLoc -> PState +pragState buf loc = + PState { + buffer = buf, + last_loc = mkSrcSpan loc loc, + last_offs = 0, + last_len = 0, + loc = loc, + extsBitmap = 0, + context = [], + lex_state = [bol, option_prags, 0] + } + -- create a parse state -- @@ -1234,6 +1332,7 @@ mkPState buf loc flags = PState { buffer = buf, last_loc = mkSrcSpan loc loc, + last_offs = 0, last_len = 0, loc = loc, extsBitmap = fromIntegral bitmap, @@ -1248,6 +1347,8 @@ mkPState buf loc flags = .|. arrowsBit `setBitIf` dopt Opt_Arrows flags .|. thBit `setBitIf` dopt Opt_TH flags .|. ipBit `setBitIf` dopt Opt_ImplicitParams flags + .|. tvBit `setBitIf` dopt Opt_ScopedTypeVariables flags + .|. bangPatBit `setBitIf` dopt Opt_BangPatterns flags -- setBitIf :: Int -> Bool -> Int b `setBitIf` cond | cond = bit b @@ -1270,13 +1371,13 @@ popContext = P $ \ s@(PState{ buffer = buf, context = ctx, -- This is only used at the outer level of a module when the 'module' -- keyword is missing. pushCurrentContext :: P () -pushCurrentContext = P $ \ s@PState{ last_loc=loc, context=ctx } -> - POk s{ context = Layout (srcSpanStartCol loc) : ctx} () +pushCurrentContext = P $ \ s@PState{ last_offs=offs, last_len=len, context=ctx } -> + POk s{context = Layout (offs-len) : ctx} () -getOffside :: SrcLoc -> P Ordering -getOffside loc = P $ \s@PState{context=stk} -> +getOffside :: P Ordering +getOffside = P $ \s@PState{last_offs=offs, context=stk} -> let ord = case stk of - (Layout n:_) -> compare (srcLocCol loc) n + (Layout n:_) -> compare offs n _ -> GT in POk s ord @@ -1293,25 +1394,23 @@ srcParseErr buf len else hcat [ptext SLIT("parse error on input "), char '`', text token, char '\''] ] - where token = lexemeToString (stepOnBy (-len) buf) len + where token = lexemeToString (offsetBytes (-len) buf) len -- Report a parse failure, giving the span of the previous token as -- the location of the error. This is the entry point for errors -- detected during parsing. srcParseFail :: P a srcParseFail = P $ \PState{ buffer = buf, last_len = len, - last_loc = last_loc, loc = loc } -> + last_loc = last_loc } -> PFailed last_loc (srcParseErr buf len) -- A lexical error is reported at a particular position in the source file, --- not over a token range. TODO: this is slightly wrong, because we record --- the error at the character position following the one which caused the --- error. We should somehow back up by one character. +-- not over a token range. lexError :: String -> P a lexError str = do loc <- getSrcLoc - i@(end,_) <- getInput - failLocMsgP loc end str + i@(AI end _ buf) <- getInput + reportLexError loc end buf str -- ----------------------------------------------------------------------------- -- This is the top-level function: called from the parser each time a @@ -1325,20 +1424,34 @@ lexer cont = do lexToken :: P (Located Token) lexToken = do - inp@(loc1,buf) <- getInput + 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 return (L span ITeof) - AlexError (loc2,_) -> do failLocMsgP loc1 loc2 "lexical error" + AlexError (AI loc2 _ buf) -> do + reportLexError loc1 loc2 buf "lexical error" AlexSkip inp2 _ -> do setInput inp2 lexToken - AlexToken inp2@(end,buf2) len t -> do + AlexToken inp2@(AI end _ buf2) len t -> do setInput inp2 let span = mkSrcSpan loc1 end - span `seq` setLastToken span len - t span buf len + let bytes = byteDiff buf buf2 + span `seq` setLastToken span bytes + t span buf bytes + +-- ToDo: Alex reports the buffer at the start of the erroneous lexeme, +-- but it would be more informative to report the location where the +-- error was actually discovered, especially if this is a decoding +-- error. +reportLexError loc1 loc2 buf str = + let + c = fst (nextChar buf) + in + if c == '\0' -- decoding errors are mapped to '\0', see utf8DecodeChar# + then failLocMsgP loc2 loc2 "UTF-8 decoding error" + else failLocMsgP loc1 loc2 (str ++ " at character " ++ show c) }