-----------------------------------------------------------------------------
-- ToDo / known bugs:
--- - Unicode
-- - parsing integers is a bit slow
-- - readRational is a bit slow
--
-- qualified varids.
{
-{-# OPTIONS -Wwarn #-}
+{-# OPTIONS -Wwarn -w #-}
-- The above warning supression flag is a temporary kludge.
-- While working on this module you are encouraged to remove it and fix
-- any warnings in the module. See
import Outputable
import StringBuffer
import FastString
-import FastTypes
import SrcLoc
import UniqFM
import DynFlags
import Data.Bits
import Data.Char
import Data.Ratio
-import Debug.Trace
}
$unispace = \x05 -- Trick Alex into handling Unicode. See alexGetChar.
| ITunsafe
| ITstdcallconv
| ITccallconv
+ | ITprimcallconv
| ITdotnet
| ITmdo
| ITfamily
isSpecial ITunsafe = True
isSpecial ITccallconv = True
isSpecial ITstdcallconv = True
+isSpecial ITprimcallconv = True
isSpecial ITmdo = True
isSpecial ITfamily = True
isSpecial ITgroup = True
-- 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 ),
( "label", ITlabel, bit ffiBit),
( "dynamic", ITdynamic, bit ffiBit),
( "safe", ITsafe, bit ffiBit),
- ( "threadsafe", ITthreadsafe, bit ffiBit),
+ ( "threadsafe", ITthreadsafe, bit ffiBit), -- ToDo: remove
( "unsafe", ITunsafe, bit ffiBit),
( "stdcall", ITstdcallconv, bit ffiBit),
( "ccall", ITccallconv, bit ffiBit),
+ ( "prim", ITprimcallconv, bit ffiBit),
( "dotnet", ITdotnet, bit ffiBit),
( "rec", ITrec, bit arrowsBit),
pop_and act span buf len = do popLexState; act span buf len
{-# INLINE nextCharIs #-}
+nextCharIs :: StringBuffer -> (Char -> Bool) -> Bool
nextCharIs buf p = not (atEnd buf) && p (currentChar buf)
+notFollowedBy :: Char -> AlexAccPred Int
notFollowedBy char _ _ _ (AI _ _ buf)
= nextCharIs buf (/=char)
+notFollowedBySymbol :: AlexAccPred Int
notFollowedBySymbol _ _ _ (AI _ _ buf)
= nextCharIs buf (`notElem` "!#$%&*+./<=>?@\\^|-~")
+notFollowedByPragmaChar :: AlexAccPred Int
notFollowedByPragmaChar _ _ _ (AI _ _ buf)
= nextCharIs buf (\c -> not (isAlphaNum c || c == '_'))
-- 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 (/='#')
notFollowedByDocOrPragma
= not $ spaceAndP buf (`nextCharIs` (`elem` "|^*$#"))
+spaceAndP :: StringBuffer -> (StringBuffer -> Bool) -> Bool
spaceAndP buf p = p buf || nextCharIs buf (==' ') && p (snd (nextChar buf))
{-
= if haddockEnabled bits then False else (p buf)
-}
+atEOL :: AlexAccPred Int
atEOL _ _ _ (AI _ _ buf) = atEnd buf || currentChar buf == '\n'
+ifExtension :: (Int -> Bool) -> AlexAccPred Int
ifExtension pred bits _ _ _ = pred bits
multiline_doc_comment :: Action
Just (_,_) -> go ('\123':commentAcc) input docType False
Just (c,input) -> go (c:commentAcc) input docType False
+withLexedDocType :: (AlexInput -> (String -> Token) -> Bool -> P (Located Token))
+ -> P (Located Token)
withLexedDocType lexDocComment = do
input@(AI _ _ buf) <- getInput
case prevChar buf ' ' of
'$' -> 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
-- 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
+rulePrag span _ _ = do
setExts (.|. bit inRulePragBit)
return (L span ITrules_prag)
endPrag :: Action
-endPrag span buf len = do
+endPrag span _ _ = do
setExts (.&. complement (bit inRulePragBit))
return (L span ITclose_prag)
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
popContext
return (L span ITccurly)
+qvarid, qconid :: StringBuffer -> Int -> Token
qvarid buf len = ITqvarid $! splitQualName buf len False
qconid buf len = ITqconid $! splitQualName buf len False
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
where
fs = lexemeToFastString buf len
+conid :: StringBuffer -> Int -> Token
conid buf len = ITconid fs
where fs = lexemeToFastString 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
(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
-- certain keywords put us in the "layout" state, where we might
-- add an opening curly brace.
+maybe_layout :: Token -> P ()
maybe_layout ITdo = pushLexState layout_do
maybe_layout ITmdo = pushLexState layout_do
maybe_layout ITof = pushLexState layout
-- by a 'do', then we allow the new context to be at the same indentation as
-- the previous context. This is what the 'strict' argument is for.
--
+new_layout_context :: Bool -> Action
new_layout_context strict span _buf _len = do
popLexState
(AI _ offset _) <- getInput
setContext (Layout offset : ctx)
return (L span ITvocurly)
+do_layout_left :: Action
do_layout_left span _buf _len = do
popLexState
pushLexState bol -- we must be at the start of a line
c' <- lex_char c i
lex_string (c':s)
+lex_stringgap :: String -> P Token
lex_stringgap s = do
c <- getCharOrFail
case c of
c | isAny c -> do setInput inp; return c
_other -> lit_error
+isAny :: Char -> Bool
isAny c | c > '\x7f' = isPrint c
| otherwise = is_any c
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
then do setInput input; return (chr i)
else lit_error
+silly_escape_chars :: [(String, Char)]
silly_escape_chars = [
("NUL", '\NUL'),
("SOH", '\SOH'),
-- 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
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],
-- -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 :: Int
magicHashBit = 11 -- "#" in both functions and operators
+kindSigsBit :: Int
kindSigsBit = 12 -- Kind signatures on type variables
+recursiveDoBit :: Int
recursiveDoBit = 13 -- mdo
+unicodeSyntaxBit :: Int
unicodeSyntaxBit = 14 -- the forall symbol, arrow symbols, etc
+unboxedTuplesBit :: Int
unboxedTuplesBit = 15 -- (# and #)
+standaloneDerivingBit :: Int
standaloneDerivingBit = 16 -- standalone instance deriving declarations
+transformComprehensionsBit :: Int
transformComprehensionsBit = 17
+qqBit :: Int
qqBit = 18 -- enable quasiquoting
+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.(+)
-genericsEnabled, ffiEnabled, parrEnabled :: Int -> Bool
+always :: Int -> Bool
always _ = True
+genericsEnabled :: Int -> Bool
genericsEnabled flags = testBit flags genericsBit
-ffiEnabled flags = testBit flags ffiBit
+parrEnabled :: Int -> Bool
parrEnabled flags = testBit flags parrBit
+arrowsEnabled :: Int -> Bool
arrowsEnabled flags = testBit flags arrowsBit
+thEnabled :: Int -> Bool
thEnabled flags = testBit flags thBit
+ipEnabled :: Int -> Bool
ipEnabled flags = testBit flags ipBit
+explicitForallEnabled :: Int -> Bool
explicitForallEnabled flags = testBit flags explicitForallBit
+bangPatEnabled :: Int -> Bool
bangPatEnabled flags = testBit flags bangPatBit
-tyFamEnabled flags = testBit flags tyFamBit
+-- tyFamEnabled :: Int -> Bool
+-- tyFamEnabled flags = testBit flags tyFamBit
+haddockEnabled :: Int -> Bool
haddockEnabled flags = testBit flags haddockBit
+magicHashEnabled :: Int -> Bool
magicHashEnabled flags = testBit flags magicHashBit
-kindSigsEnabled flags = testBit flags kindSigsBit
-recursiveDoEnabled flags = testBit flags recursiveDoBit
+-- kindSigsEnabled :: Int -> Bool
+-- kindSigsEnabled flags = testBit flags kindSigsBit
+unicodeSyntaxEnabled :: Int -> Bool
unicodeSyntaxEnabled flags = testBit flags unicodeSyntaxBit
+unboxedTuplesEnabled :: Int -> Bool
unboxedTuplesEnabled flags = testBit flags unboxedTuplesBit
+standaloneDerivingEnabled :: Int -> Bool
standaloneDerivingEnabled flags = testBit flags standaloneDerivingBit
-transformComprehensionsEnabled flags = testBit flags transformComprehensionsBit
+qqEnabled :: Int -> Bool
qqEnabled flags = testBit flags qqBit
-inRulePrag flags = testBit flags inRulePragBit
+-- inRulePrag :: Int -> Bool
+-- inRulePrag flags = testBit flags inRulePragBit
+rawTokenStreamEnabled :: Int -> Bool
rawTokenStreamEnabled flags = testBit flags rawTokenStreamBit
+newQualOps :: Int -> Bool
newQualOps flags = testBit flags newQualOpsBit
+oldQualOps :: Int -> Bool
oldQualOps flags = not (newQualOps flags)
-- PState for parsing options pragmas
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 =
projects / ghc-hetmet.git / blobdiff