--- |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.
, char
, string
, (<|>)
+ , choice
, oneOf
, digit
, hexDigit
, notFollowedBy
, many
+ , manyChar
, many1
, count
, option
)
where
-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 Data.Int
+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
+ return !x = Parser $! return $! Success x
+ fail _ = Parser $! return $! IllegalInput
--- |@'failP'@ is just a synonym for @'Prelude.fail Prelude.undefined'@.
+-- |@'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 -> ByteString -> (ParserResult a, ByteString)
-parse p input -- input は lazy である必要有り。
- = p `seq`
- let (result, (input', _)) = runState (runParser p) (input, True)
+-- |@'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。
+ = 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, 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 p input
- = p `seq` -- input は lazy である必要有り。
- parse p $! B.pack input
+parseStr :: Parser a -> String -> (# ParserResult a, Lazy.ByteString #)
+parseStr !p input -- input は lazy である必要有り。
+ = 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 put (B.tail input, isEOFFatal)
+ 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)
+allowEOF !f
+ = Parser $! do saved@(PST _ isEOFFatal) <- get
+ put $! saved { pstIsEOFFatal = False }
- result <- runParser f
+ result <- runParser f
- (input', _) <- get
- put (input', isEOFFatal)
+ state <- get
+ put $! state { pstIsEOFFatal = isEOFFatal }
- return result
+ return result
satisfy :: (Char -> Bool) -> Parser Char
-satisfy f = f `seq`
- do c <- anyChar
- if f c then
- return c
- else
- failP
+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
+ = let bs = B.pack str
+ len = B.length bs
+ in
+ Parser $!
+ do st <- get
+ let (bs', rest) = B.splitAt len $ pstInput st
+ st' = st { pstInput = rest }
+ if B.length bs' < len then
+ return ReachedEOF
+ else
+ if bs == bs' then
+ do put st'
+ return $ Success str
+ else
+ return IllegalInput
infixr 0 <|>
-- |This is the backtracking alternation. There is no non-backtracking
-- equivalent.
(<|>) :: Parser a -> Parser a -> Parser a
-f <|> g
- = f `seq` g `seq`
- Parser $! do saved@(_, isEOFFatal) <- get -- 状態を保存
+(!f) <|> (!g)
+ = 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 (<|>) failP . map char
notFollowedBy :: Parser a -> Parser ()
-notFollowedBy p = p `seq`
- (p >> failP) <|> 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
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 #)
+
+manyChar :: Parser Char -> Parser Lazy.ByteString
+manyChar !p = Parser $!
+ do state <- get
+ case scan' state 0 of
+ Success len
+ -> do let (bs, rest) = B.splitAt len (pstInput state)
+ state' = state { pstInput = rest }
+ put state'
+ return $ Success bs
+ ReachedEOF
+ -> if pstIsEOFFatal state then
+ return ReachedEOF
+ else
+ error "internal error"
+ _ -> error "internal error"
+ where
+ scan' :: ParserState -> Int64 -> ParserResult Int64
+ scan' !st !soFar
+ = case runState (runParser p) st of
+ (Success _ , st') -> scan' st' (soFar + 1)
+ (IllegalInput, _ ) -> Success soFar
+ (ReachedEOF , _ ) -> if pstIsEOFFatal st then
+ ReachedEOF
+ else
+ Success soFar
many1 :: Parser a -> Parser [a]
-many1 p = p `seq`
- do x <- p
- xs <- many p
- return (x:xs)
+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
-option def p = p `seq`
- p <|> return def
+option def !p = p <|> return def
sepBy :: Parser a -> Parser sep -> Parser [a]
-sepBy p sep = p `seq` sep `seq`
- sepBy1 p sep <|> return []
+sepBy !p !sep = sepBy1 p sep <|> return []
sepBy1 :: Parser a -> Parser sep -> Parser [a]
-sepBy1 p sep = p `seq` sep `seq`
- do x <- p
- xs <- many $! sep >> p
- return (x:xs)
+sepBy1 !p !sep
+ = do x <- p
+ xs <- many $! sep >> p
+ return (x:xs)
sp :: Parser Char