+{-# LANGUAGE
+ BangPatterns
+ , GeneralizedNewtypeDeriving
+ , OverloadedStrings
+ , UnicodeSyntax
+ #-}
module Network.HTTP.Lucu.Headers
( Headers
, HasHeaders(..)
- , noCaseCmp
- , noCaseEq
-
- , emptyHeaders
, toHeaders
, fromHeaders
, headersP
- , hPutHeaders
+ , printHeaders
)
where
-
-import qualified Data.ByteString as Strict (ByteString)
-import Data.ByteString.Internal (toForeignPtr, w2c, inlinePerformIO)
-import qualified Data.ByteString.Char8 as C8 hiding (ByteString)
-import Data.Char
-import Data.List
-import Data.Map (Map)
+import Control.Applicative
+import Data.Ascii (Ascii, AsciiBuilder, CIAscii)
+import qualified Data.Ascii as A
+import Data.Attoparsec.Char8 as P
+import qualified Data.ByteString as BS
+import Data.Map (Map)
import qualified Data.Map as M
-import Data.Ord
-import Data.Word
-import Foreign.ForeignPtr
-import Foreign.Ptr
-import Foreign.Storable
-import Network.HTTP.Lucu.HandleLike
-import Network.HTTP.Lucu.Parser
-import Network.HTTP.Lucu.Parser.Http
-import Network.HTTP.Lucu.Utils
-
-type Headers = Map NCBS Strict.ByteString
-newtype NCBS = NCBS Strict.ByteString
-
-toNCBS :: Strict.ByteString -> NCBS
-toNCBS = NCBS
-{-# INLINE toNCBS #-}
-
-fromNCBS :: NCBS -> Strict.ByteString
-fromNCBS (NCBS x) = x
-{-# INLINE fromNCBS #-}
-
-instance Eq NCBS where
- (NCBS a) == (NCBS b) = a == b
-
-instance Ord NCBS where
- (NCBS a) `compare` (NCBS b) = a `noCaseCmp` b
-
-instance Show NCBS where
- show (NCBS x) = show x
-
-noCaseCmp :: Strict.ByteString -> Strict.ByteString -> Ordering
-noCaseCmp a b = a `seq` b `seq`
- toForeignPtr a `cmp` toForeignPtr b
- where
- cmp :: (ForeignPtr Word8, Int, Int) -> (ForeignPtr Word8, Int, Int) -> Ordering
- cmp (x1, s1, l1) (x2, s2, l2)
- | x1 `seq` s1 `seq` l1 `seq` x2 `seq` s2 `seq` l2 `seq` False = undefined
- | l1 == 0 && l2 == 0 = EQ
- | x1 == x2 && s1 == s2 && l1 == l2 = EQ
- | otherwise
- = inlinePerformIO $
- withForeignPtr x1 $ \ p1 ->
- withForeignPtr x2 $ \ p2 ->
- noCaseCmp' (p1 `plusPtr` s1) l1 (p2 `plusPtr` s2) l2
-
-
--- もし先頭の文字列が等しければ、短い方が小さい。
-noCaseCmp' :: Ptr Word8 -> Int -> Ptr Word8 -> Int -> IO Ordering
-noCaseCmp' p1 l1 p2 l2
- | p1 `seq` l1 `seq` p2 `seq` l2 `seq` False = undefined
- | l1 == 0 && l2 == 0 = return EQ
- | l1 == 0 = return LT
- | l2 == 0 = return GT
- | otherwise
- = do c1 <- peek p1
- c2 <- peek p2
- case comparing (toLower . w2c) c1 c2 of
- EQ -> noCaseCmp' (p1 `plusPtr` 1) (l1 - 1) (p2 `plusPtr` 1) (l2 - 1)
- x -> return x
-
-
-noCaseEq :: Strict.ByteString -> Strict.ByteString -> Bool
-noCaseEq a b = a `seq` b `seq`
- toForeignPtr a `cmp` toForeignPtr b
- where
- cmp :: (ForeignPtr Word8, Int, Int) -> (ForeignPtr Word8, Int, Int) -> Bool
- cmp (x1, s1, l1) (x2, s2, l2)
- | x1 `seq` s1 `seq` l1 `seq` x2 `seq` s2 `seq` l2 `seq` False = undefined
- | l1 /= l2 = False
- | l1 == 0 && l2 == 0 = True
- | x1 == x2 && s1 == s2 && l1 == l2 = True
- | otherwise
- = inlinePerformIO $
- withForeignPtr x1 $ \ p1 ->
- withForeignPtr x2 $ \ p2 ->
- noCaseEq' (p1 `plusPtr` s1) (p2 `plusPtr` s2) l1
-
-
-noCaseEq' :: Ptr Word8 -> Ptr Word8 -> Int -> IO Bool
-noCaseEq' p1 p2 l
- | p1 `seq` p2 `seq` l `seq` False = undefined
- | l == 0 = return True
- | otherwise
- = do c1 <- peek p1
- c2 <- peek p2
- if toLower (w2c c1) == toLower (w2c c2) then
- noCaseEq' (p1 `plusPtr` 1) (p2 `plusPtr` 1) (l - 1)
- else
- return False
+import Data.Monoid
+import Data.Monoid.Unicode
+import Network.HTTP.Lucu.Parser.Http
+import Network.HTTP.Lucu.Utils
+import Prelude.Unicode
+newtype Headers
+ = Headers (Map CIAscii Ascii)
+ deriving (Eq, Show, Monoid)
class HasHeaders a where
- getHeaders :: a -> Headers
- setHeaders :: a -> Headers -> a
-
- getHeader :: Strict.ByteString -> a -> Maybe Strict.ByteString
- getHeader key a
- = key `seq` a `seq`
- M.lookup (toNCBS key) (getHeaders a)
-
- deleteHeader :: Strict.ByteString -> a -> a
- deleteHeader key a
- = key `seq` a `seq`
- setHeaders a $ M.delete (toNCBS key) (getHeaders a)
-
- setHeader :: Strict.ByteString -> Strict.ByteString -> a -> a
- setHeader key val a
- = key `seq` val `seq` a `seq`
- setHeaders a $ M.insert (toNCBS key) val (getHeaders a)
-
-
-emptyHeaders :: Headers
-emptyHeaders = M.empty
-
-
-toHeaders :: [(Strict.ByteString, Strict.ByteString)] -> Headers
-toHeaders xs = mkHeaders xs M.empty
-
-
-mkHeaders :: [(Strict.ByteString, Strict.ByteString)] -> Headers -> Headers
-mkHeaders [] m = m
-mkHeaders ((key, val):xs) m = mkHeaders xs $
- case M.lookup (toNCBS key) m of
- Nothing -> M.insert (toNCBS key) val m
- Just old -> M.insert (toNCBS key) (merge old val) m
+ getHeaders ∷ a → Headers
+ setHeaders ∷ a → Headers → a
+
+ getHeader ∷ CIAscii → a → Maybe Ascii
+ {-# INLINE getHeader #-}
+ getHeader !key !a
+ = case getHeaders a of
+ Headers m → M.lookup key m
+
+ deleteHeader ∷ CIAscii → a → a
+ {-# INLINE deleteHeader #-}
+ deleteHeader !key !a
+ = case getHeaders a of
+ Headers m
+ → setHeaders a $ Headers $ M.delete key m
+
+ setHeader ∷ CIAscii → Ascii → a → a
+ {-# INLINE setHeader #-}
+ setHeader !key !val !a
+ = case getHeaders a of
+ Headers m
+ → setHeaders a $ Headers $ M.insert key val m
+
+instance HasHeaders Headers where
+ getHeaders = id
+ setHeaders _ = id
+
+toHeaders ∷ [(CIAscii, Ascii)] → Headers
+{-# INLINE toHeaders #-}
+toHeaders = flip mkHeaders (∅)
+
+mkHeaders ∷ [(CIAscii, Ascii)] → Headers → Headers
+mkHeaders [] (Headers m) = Headers m
+mkHeaders ((key, val):xs) (Headers m)
+ = mkHeaders xs $ Headers $
+ case M.lookup key m of
+ Nothing → M.insert key val m
+ Just old → M.insert key (merge old val) m
where
- merge :: Strict.ByteString -> Strict.ByteString -> Strict.ByteString
- -- カンマ區切りである事を假定する。RFC ではカンマ區切りに出來ない
- -- ヘッダは複數個あってはならない事になってゐる。
+ merge ∷ Ascii → Ascii → Ascii
+ {-# INLINE merge #-}
merge a b
- | C8.null a && C8.null b = C8.empty
- | C8.null a = b
- | C8.null b = a
- | otherwise = C8.concat [a, C8.pack ", ", b]
-
+ | nullA a ∧ nullA b = (∅)
+ | nullA a = b
+ | nullA b = a
+ | otherwise = a ⊕ ", " ⊕ b
-fromHeaders :: Headers -> [(Strict.ByteString, Strict.ByteString)]
-fromHeaders hs = [(fromNCBS a, b) | (a, b) <- M.toList hs]
+ nullA ∷ Ascii → Bool
+ {-# INLINE nullA #-}
+ nullA = BS.null ∘ A.toByteString
+fromHeaders ∷ Headers → [(CIAscii, Ascii)]
+fromHeaders (Headers m) = M.toList m
{-
message-header = field-name ":" [ field-value ]
field-value の先頭および末尾にある LWS は全て削除され、それ以外の
LWS は單一の SP に變換される。
-}
-headersP :: Parser Headers
-headersP = do xs <- many header
- _ <- crlf
- return $! toHeaders xs
+headersP ∷ Parser Headers
+{-# INLINEABLE headersP #-}
+headersP = do xs ← P.many header
+ crlf
+ return $ toHeaders xs
where
- header :: Parser (Strict.ByteString, Strict.ByteString)
- header = do name <- token
- _ <- char ':'
- -- FIXME: これは多少インチキだが、RFC 2616 のこの部分
- -- の記述はひどく曖昧であり、この動作が本當に間違って
- -- ゐるのかどうかも良く分からない。例へば
- -- quoted-string の内部にある空白は纏めていいのか惡い
- -- のか?直勸的には駄目さうに思へるが、そんな記述は見
- -- 付からない。
- contents <- many (lws <|> many1 text)
- _ <- crlf
- let value = foldr (++) "" contents
- norm = normalize value
- return (C8.pack name, C8.pack norm)
-
- normalize :: String -> String
- normalize = trimBody . trim isWhiteSpace
-
- trimBody = concat
- . map (\ s -> if head s == ' ' then
- " "
- else
- s)
- . group
- . map (\ c -> if isWhiteSpace c
- then ' '
- else c)
-
-
-hPutHeaders :: HandleLike h => h -> Headers -> IO ()
-hPutHeaders h hds
- = h `seq` hds `seq`
- mapM_ putH (M.toList hds) >> hPutBS h (C8.pack "\r\n")
+ header ∷ Parser (CIAscii, Ascii)
+ header = do name ← A.toCIAscii <$> token
+ _ ← char ':'
+ skipMany lws
+ values ← sepBy content (try lws)
+ skipMany (try lws)
+ crlf
+ return (name, joinValues values)
+
+ content ∷ Parser Ascii
+ {-# INLINE content #-}
+ content = A.unsafeFromByteString
+ <$>
+ takeWhile1 (\c → ((¬) (isSPHT c)) ∧ isText c)
+
+ joinValues ∷ [Ascii] → Ascii
+ {-# INLINE joinValues #-}
+ joinValues = A.fromAsciiBuilder ∘ joinWith "\x20" ∘ map A.toAsciiBuilder
+
+printHeaders ∷ Headers → AsciiBuilder
+printHeaders (Headers m)
+ = mconcat (map printHeader (M.toList m)) ⊕
+ A.toAsciiBuilder "\x0D\x0A"
where
- putH :: (NCBS, Strict.ByteString) -> IO ()
- putH (name, value)
- = name `seq` value `seq`
- do hPutBS h (fromNCBS name)
- hPutBS h (C8.pack ": ")
- hPutBS h value
- hPutBS h (C8.pack "\r\n")
+ printHeader ∷ (CIAscii, Ascii) → AsciiBuilder
+ printHeader (name, value)
+ = A.toAsciiBuilder (A.fromCIAscii name) ⊕
+ A.toAsciiBuilder ": " ⊕
+ A.toAsciiBuilder value ⊕
+ A.toAsciiBuilder "\x0D\x0A"