-----------------------------------------------------------------------------
--- (c) The University of Glasgow, 2003
+-- (c) The University of Glasgow, 2006
--
-- GHC's lexer.
--
import FastTypes
import SrcLoc
import UniqFM
-import CmdLineOpts
+import DynFlags
import Ctype
import Util ( maybePrefixMatch, readRational )
import DATA_BITS
-import Char
+import Data.Char
import Ratio
--import TRACE
}
-$whitechar = [\ \t\n\r\f\v\xa0]
+$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 \']
@varsym = $symbol $symchar*
@consym = \: $symchar*
-@decimal = $digit+
+@decimal = $decdigit+
@octal = $octit+
@hexadecimal = $hexit+
@exponent = [eE] [\-\+]? @decimal
-- have to exclude those.
-- The regex says: "munch all the characters after the dashes, as long as
-- the first one is not a symbol".
-"--"\-* [^$symbol] .* ;
+"--"\-* [^$symbol :] .* ;
"--"\-* / { atEOL } ;
-- 'bol' state: beginning of a line. Slurp up all the whitespace (including
-- single-line line pragmas, of the form
-- # <line> "<file>" <extra-stuff> \n
-<line_prag1> $digit+ { setLine line_prag1a }
+<line_prag1> $decdigit+ { setLine line_prag1a }
<line_prag1a> \" [$graphic \ ]* \" { setFile line_prag1b }
<line_prag1b> .* { pop }
-- Haskell-style line pragmas, of the form
-- {-# LINE <line> "<file>" #-}
-<line_prag2> $digit+ { setLine line_prag2a }
+<line_prag2> $decdigit+ { setLine line_prag2a }
<line_prag2a> \" [$graphic \ ]* \" { setFile line_prag2b }
<line_prag2b> "#-}"|"-}" { pop }
-- 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 -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)
+<glaexts>
+ "{-#" $whitechar* (RULES|rules) { token ITrules_prag }
+
<0,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 ITspec_prag }
+ "{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
+ $whitechar* (INLINE|inline) { token (ITspec_inline_prag True) }
"{-#" $whitechar* (SPECIALI[SZ]E|speciali[sz]e)
- { token ITspecialise_prag }
+ $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 }
| 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
| 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 -- $(
+ | ITvarQuote -- '
+ | ITtyQuote -- ''
-- Arrow notation extension
| ITproc
| ITrec
- | IToparenbar -- (|
- | ITcparenbar -- |)
- | ITlarrowtail -- -<
- | ITrarrowtail -- >-
- | ITLarrowtail -- -<<
- | ITRarrowtail -- >>-
+ | IToparenbar -- (|
+ | ITcparenbar -- |)
+ | ITlarrowtail -- -<
+ | ITrarrowtail -- >-
+ | ITLarrowtail -- -<<
+ | ITRarrowtail -- >>-
| ITunknown String -- Used when the lexer can't make sense of it
| ITeof -- end of file token
( "where", ITwhere, 0 ),
( "_scc_", ITscc, 0 ), -- ToDo: remove
- ( "forall", ITforall, bit glaExtsBit),
+ ( "forall", ITforall, bit tvBit),
( "mdo", ITmdo, bit glaExtsBit),
( "foreign", ITforeign, bit ffiBit),
,("!", ITbang, 0)
,("*", ITstar, bit glaExtsBit) -- For data T (a::*) = MkT
- ,(".", ITdot, bit glaExtsBit) -- For 'forall a . t'
+ ,(".", ITdot, bit tvBit) -- For 'forall a . t'
,("-<", ITlarrowtail, bit arrowsBit)
,(">-", ITrarrowtail, bit arrowsBit)
,("-<<", ITLarrowtail, bit arrowsBit)
,(">>-", ITRarrowtail, bit arrowsBit)
+
+#if __GLASGOW_HASKELL__ >= 605
+ ,("λ", ITlam, 0)
+ ,("∀", ITforall, bit tvBit)
+ ,("→", ITrarrow, 0)
+ ,("←", ITlarrow, 0)
+ ,("⋯", ITdotdot, 0)
+#endif
]
-- -----------------------------------------------------------------------------
pop_and :: Action -> Action
pop_and act span buf len = do popLexState; act span buf len
-notFollowedBy char _ _ _ (_,buf) = atEnd buf || currentChar buf /= char
+notFollowedBy char _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf /= char
-notFollowedBySymbol _ _ _ (_,buf)
+notFollowedBySymbol _ _ _ (AI _ _ buf)
= atEnd buf || currentChar buf `notElem` "!#$%&*+./<=>?@\\^|-~"
-atEOL _ _ _ (_,buf) = atEnd buf || currentChar buf == '\n'
+atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
ifExtension pred bits _ _ _ = pred bits
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
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
-- 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
= 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 octDecDigit))
+ = 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 hexDigit))
+ = 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 octDecDigit))
prim_octal span buf len
- = return (L span (ITprimint $! parseInteger (stepOnBy 2 buf) (len-3) 8 octDecDigit))
+ = 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 hexDigit))
+ = return (L span (ITprimint $! parseInteger (offsetBytes 2 buf) (len-3) 16 hexDigit))
tok_float str = ITrational $! readRational str
prim_float str = ITprimfloat $! readRational str
-- 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
--
new_layout_context strict span _buf _len = do
popLexState
- let offset = srcSpanStartCol span
+ (AI _ offset _) <- getInput
ctx <- getContext
case ctx of
Layout prev_off : _ |
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
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
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
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
_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 ->
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
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,
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
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
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
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 = c `seq` loc' `seq` s' `seq` Just (c, (loc', s'))
- where c = currentChar s
- loc' = advanceSrcLoc loc c
- s' = stepOn s
+ | 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
+#if __GLASGOW_HASKELL__ < 605
+ = c -- no Unicode support
+#else
+ | 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
+#endif
+
+-- 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` 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} ()
arrowsBit = 4
thBit = 5
ipBit = 6
+tvBit = 7 -- Scoped type variables enables 'forall' keyword
glaExtsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
glaExtsEnabled flags = testBit flags glaExtsBit
arrowsEnabled flags = testBit flags arrowsBit
thEnabled flags = testBit flags thBit
ipEnabled flags = testBit flags ipBit
+tvEnabled flags = testBit flags tvBit
-- create a parse state
--
PState {
buffer = buf,
last_loc = mkSrcSpan loc loc,
+ last_offs = 0,
last_len = 0,
loc = loc,
extsBitmap = fromIntegral bitmap,
.|. arrowsBit `setBitIf` dopt Opt_Arrows flags
.|. thBit `setBitIf` dopt Opt_TH flags
.|. ipBit `setBitIf` dopt Opt_ImplicitParams flags
+ .|. tvBit `setBitIf` dopt Opt_ScopedTypeVariables flags
--
setBitIf :: Int -> Bool -> Int
b `setBitIf` cond | cond = bit b
-- 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
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
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
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)
}