--- |Yet another parser combinator. This is mostly a subset of Parsec
--- but there are some differences:
+-- |Yet another parser combinator. This is mostly a subset of
+-- "Text.ParserCombinators.Parsec" but there are some differences:
--
--- * This parser works on ByteString instead of String.
+-- * This parser works on 'Data.ByteString.Base.LazyByteString'
+-- instead of 'Prelude.String'.
--
-- * Backtracking is the only possible behavior so there is no \"try\"
-- action.
( Parser
, ParserResult(..)
+ , failP
+
, parse
, parseStr
, char
, string
, (<|>)
+ , choice
, oneOf
, digit
, hexDigit
, notFollowedBy
, many
, many1
- , manyTill
- , many1Till
, count
, option
, sepBy
)
where
-import Control.Monad
-import Control.Monad.State
-import qualified Data.ByteString.Lazy.Char8 as B
-import Data.ByteString.Lazy.Char8 (ByteString)
+import Control.Monad.State.Strict
+import qualified Data.ByteString.Lazy as Lazy (ByteString)
+import qualified Data.ByteString.Lazy.Char8 as B hiding (ByteString)
+import qualified Data.Foldable as Fold
+import qualified Data.Sequence as Seq
+import Data.Sequence (Seq, (|>))
--- |@Parser a@ is obviously a parser which parses and returns @a@.
+-- |@'Parser' a@ is obviously a parser which parses and returns @a@.
newtype Parser a = Parser {
runParser :: State ParserState (ParserResult a)
}
-type ParserState = (ByteString, IsEOFFatal)
-type IsEOFFatal = Bool
+data ParserState
+ = PST {
+ pstInput :: Lazy.ByteString
+ , pstIsEOFFatal :: !Bool
+ }
+ deriving (Eq, Show)
+
data ParserResult a = Success !a
| IllegalInput -- 受理出來ない入力があった
-- (>>=) :: Parser a -> (a -> Parser b) -> Parser b
instance Monad Parser where
- p >>= f = Parser $! do saved@(_, isEOFFatal) <- get -- 失敗した時の爲に状態を保存
+ p >>= f = Parser $! do saved <- get -- 失敗した時の爲に状態を保存
result <- runParser p
case result of
- Success a -> a `seq` runParser (f a)
+ Success a -> runParser (f a)
IllegalInput -> do put saved -- 状態を復歸
return IllegalInput
- ReachedEOF -> do unless isEOFFatal
- $ put saved -- 状態を復歸
+ ReachedEOF -> do put saved -- 状態を復歸
return ReachedEOF
return x = x `seq` Parser $! return $! Success x
fail _ = Parser $! return $! IllegalInput
--- |@'parse' p bstr@ parses @bstr@ with @p@ and returns @(result,
--- remaining)@.
-parse :: Parser a -> ByteString -> (ParserResult a, ByteString)
+-- |@'failP'@ is just a synonym for @'Prelude.fail'
+-- 'Prelude.undefined'@.
+failP :: Parser a
+failP = fail undefined
+
+-- |@'parse' p bstr@ parses @bstr@ with @p@ and returns @(# result,
+-- remaining #)@.
+parse :: Parser a -> Lazy.ByteString -> (# ParserResult a, Lazy.ByteString #)
parse p input -- input は lazy である必要有り。
= p `seq`
- let (result, (input', _)) = runState (runParser p) (input, True)
+ let (result, state') = runState (runParser p) (PST input True)
in
- result `seq` (result, input') -- input' も lazy である必要有り。
+ result `seq` (# result, pstInput state' #) -- pstInput state' も lazy である必要有り。
-- |@'parseStr' p str@ packs @str@ and parses it.
-parseStr :: Parser a -> String -> (ParserResult a, ByteString)
+parseStr :: Parser a -> String -> (# ParserResult a, Lazy.ByteString #)
parseStr p input
= p `seq` -- input は lazy である必要有り。
- parse p $! B.pack input
+ parse p (B.pack input)
anyChar :: Parser Char
anyChar = Parser $!
- do (input, isEOFFatal) <- get
+ do state@(PST input _) <- get
if B.null input then
return ReachedEOF
else
- do let c = B.head input
- put (B.tail input, isEOFFatal)
- return (Success c)
+ do put $! state { pstInput = B.tail input }
+ return (Success $! B.head input)
eof :: Parser ()
eof = Parser $!
- do (input, _) <- get
+ do PST input _ <- get
if B.null input then
- return $ Success ()
+ return $! Success ()
else
return IllegalInput
-- |@'allowEOF' p@ makes @p@ treat reaching EOF a normal failure.
allowEOF :: Parser a -> Parser a
allowEOF f = f `seq`
- Parser $! do (input, isEOFFatal) <- get
- put (input, False)
+ Parser $! do saved@(PST _ isEOFFatal) <- get
+ put $! saved { pstIsEOFFatal = False }
result <- runParser f
- (input', _) <- get
- put (input', isEOFFatal)
+ state <- get
+ put $! state { pstIsEOFFatal = isEOFFatal }
return result
satisfy :: (Char -> Bool) -> Parser Char
-satisfy f = f `seq`
- do c <- anyChar
- unless (f c) (fail "")
- return c
+satisfy !f
+ = do c <- anyChar
+ if f c then
+ return c
+ else
+ failP
char :: Char -> Parser Char
-char c = c `seq` satisfy (== c)
+char !c = satisfy (== c)
string :: String -> Parser String
-string str = str `seq`
- do mapM_ char str
- return str
+string !str = str `seq`
+ do mapM_ char str
+ return str
infixr 0 <|>
(<|>) :: Parser a -> Parser a -> Parser a
f <|> g
= f `seq` g `seq`
- Parser $! do saved@(_, isEOFFatal) <- get -- 状態を保存
+ Parser $! do saved <- get -- 状態を保存
result <- runParser f
case result of
- Success a -> return $ Success a
+ Success a -> return $! Success a
IllegalInput -> do put saved -- 状態を復歸
runParser g
- ReachedEOF -> if isEOFFatal then
- return ReachedEOF
+ ReachedEOF -> if pstIsEOFFatal saved then
+ do put saved
+ return ReachedEOF
else
do put saved
runParser g
+choice :: [Parser a] -> Parser a
+choice = foldl (<|>) failP
+
+
oneOf :: [Char] -> Parser Char
-oneOf = foldl (<|>) (fail "") . map char
+oneOf = foldl (<|>) failP . map char
notFollowedBy :: Parser a -> Parser ()
-notFollowedBy p = p `seq`
- p >>= fail "" <|> return ()
+notFollowedBy !p
+ = Parser $! do saved <- get -- 状態を保存
+ result <- runParser p
+ case result of
+ Success _ -> do put saved -- 状態を復歸
+ return IllegalInput
+ IllegalInput -> do put saved -- 状態を復歸
+ return $! Success ()
+ ReachedEOF -> do put saved -- 状態を復歸
+ return $! Success ()
digit :: Parser Char
if c >= '0' && c <= '9' then
return c
else
- fail ""
+ failP
hexDigit :: Parser Char
(c >= 'A' && c <= 'F') then
return c
else
- fail ""
+ failP
-many :: Parser a -> Parser [a]
-many p = p `seq`
- do x <- p
- xs <- many p
- return (x:xs)
- <|>
- return []
+many :: forall a. Parser a -> Parser [a]
+many !p = Parser $!
+ do state <- get
+ let (# result, state' #) = many' state Seq.empty
+ put state'
+ return result
+ where
+ many' :: ParserState -> Seq a -> (# ParserResult [a], ParserState #)
+ many' !st !soFar
+ = case runState (runParser p) st of
+ (Success a, st') -> many' st' (soFar |> a)
+ (IllegalInput, _) -> (# Success (Fold.toList soFar), st #)
+ (ReachedEOF , _) -> if pstIsEOFFatal st then
+ (# ReachedEOF, st #)
+ else
+ (# Success (Fold.toList soFar), st #)
many1 :: Parser a -> Parser [a]
-many1 p = p `seq`
- do ret <- many p
- case ret of
- [] -> fail ""
- xs -> return xs
-
-
-manyTill :: Parser a -> Parser end -> Parser [a]
-manyTill p end
- = p `seq` end `seq`
- many $! do x <- p
- end
- return x
-
-
-many1Till :: Parser a -> Parser end -> Parser [a]
-many1Till p end
- = p `seq` end `seq`
- many1 $! do x <- p
- end
- return x
+many1 !p = do x <- p
+ xs <- many p
+ return (x:xs)
count :: Int -> Parser a -> Parser [a]
-count 0 _ = return []
-count n p = n `seq` p `seq`
- do x <- p
- xs <- count (n-1) p
- return (x:xs)
+count !n !p = Parser $! count' n p Seq.empty
+
+-- This implementation is rather ugly but we need to make it
+-- tail-recursive to avoid stack overflow.
+count' :: Int -> Parser a -> Seq a -> State ParserState (ParserResult [a])
+count' 0 _ !soFar = return $! Success $! Fold.toList soFar
+count' !n !p !soFar = do saved <- get
+ result <- runParser p
+ case result of
+ Success a -> count' (n-1) p (soFar |> a)
+ IllegalInput -> do put saved
+ return IllegalInput
+ ReachedEOF -> do put saved
+ return ReachedEOF
+
-- def may be a _|_
option :: a -> Parser a -> Parser a
projects / Lucu.git / blobdiff