X-Git-Url: http://git.cielonegro.org/gitweb.cgi?a=blobdiff_plain;f=Network%2FHTTP%2FLucu%2FHeaders.hs;h=a47f2ac9ea60c1869221c573238b2186f076618e;hb=a362be1c8664306b970c32e1df9b62081498feb1;hp=5eeab6feb699b8455b717ea920b81e60a94ece3c;hpb=3d017dd65ddede9a11c5b7a34a91e04340e67bc4;p=Lucu.git

diff --git a/Network/HTTP/Lucu/Headers.hs b/Network/HTTP/Lucu/Headers.hs
index 5eeab6f..a47f2ac 100644
--- a/Network/HTTP/Lucu/Headers.hs
+++ b/Network/HTTP/Lucu/Headers.hs
@@ -1,164 +1,109 @@
+{-# LANGUAGE
+    GeneralizedNewtypeDeriving
+  , OverloadedStrings
+  , UnicodeSyntax
+  #-}
 module Network.HTTP.Lucu.Headers
     ( Headers
     , HasHeaders(..)
 
-    , noCaseCmp
-    , noCaseEq
+    , singleton
 
-    , emptyHeaders
     , toHeaders
     , fromHeaders
 
-    , headersP
-    , hPutHeaders
+    , headers
+    , 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 Control.Monad
+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.List
+import Data.Map (Map)
 import qualified Data.Map as M
-import           Data.Word
-import           Foreign.ForeignPtr
-import           Foreign.Ptr
-import           Foreign.Storable
-import           Network.HTTP.Lucu.Parser
-import           Network.HTTP.Lucu.Parser.Http
-import           Network.HTTP.Lucu.Utils
-import           System.IO
-
-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 toLower (w2c c1) `compare` toLower (w2c 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 qualified Data.Map.Unicode as M
+import Data.Monoid
+import Data.Monoid.Unicode
+import Network.HTTP.Lucu.Parser.Http
+import Prelude.Unicode
 
+newtype Headers
+    = Headers (Map CIAscii Ascii)
+      deriving (Eq, Show, Monoid)
 
 class HasHeaders a where
-    getHeaders :: a -> Headers
-    setHeaders :: a -> Headers -> a
+    getHeaders ∷ a → Headers
+    setHeaders ∷ a → Headers → a
 
-    getHeader :: Strict.ByteString -> a -> Maybe Strict.ByteString
+    getHeader ∷ CIAscii → a → Maybe Ascii
     getHeader key a
-        = key `seq` a `seq`
-          M.lookup (toNCBS key) (getHeaders a)
-
-    deleteHeader :: Strict.ByteString -> a -> a
+        = case getHeaders a of
+            Headers m → M.lookup key m
+
+    hasHeader ∷ CIAscii → a → Bool
+    {-# INLINE hasHeader #-}
+    hasHeader key a
+        = case getHeaders a of
+            Headers m → key M.∈ m
+
+    getCIHeader ∷ CIAscii → a → Maybe CIAscii
+    {-# INLINE getCIHeader #-}
+    getCIHeader key a
+        = A.toCIAscii <$> getHeader key a
+
+    deleteHeader ∷ CIAscii → a → a
+    {-# INLINE deleteHeader #-}
     deleteHeader key a
-        = key `seq` a `seq`
-          setHeaders a $ M.delete (toNCBS key) (getHeaders a)
+        = case getHeaders a of
+            Headers m
+              → setHeaders a $ Headers $ M.delete key m
 
-    setHeader :: Strict.ByteString -> Strict.ByteString -> a -> a
+    setHeader ∷ CIAscii → Ascii → a → a
+    {-# INLINE setHeader #-}
     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
+        = case getHeaders a of
+            Headers m
+              → setHeaders a $ Headers $ M.insert key val m
+
+instance HasHeaders Headers where
+    getHeaders   = id
+    setHeaders _ = id
+
+singleton ∷ CIAscii → Ascii → Headers
+{-# INLINE singleton #-}
+singleton key val
+    = Headers $ M.singleton key val
+
+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 ]
@@ -171,49 +116,42 @@ fromHeaders hs = [(fromNCBS a, b) | (a, b) <- M.toList hs]
   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
+headers ∷ Parser Headers
+{-# INLINEABLE headers #-}
+headers = 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)
+      header ∷ Parser (CIAscii, Ascii)
+      header = do name ← A.toCIAscii <$> token
+                  void $ char ':'
+                  skipMany lws
+                  values ← content `sepBy` try lws
+                  skipMany (try lws)
                   crlf
-                  let value = foldr (++) "" contents
-                      norm  = normalize value
-                  return (C8.pack name, C8.pack norm)
-
-      normalize :: String -> String
-      normalize = trimBody . trim isWhiteSpace
-
-      trimBody = foldr (++) []
-                 . map (\ s -> if head s == ' ' then
-                                   " "
-                               else
-                                   s)
-                 . group
-                 . map (\ c -> if isWhiteSpace c
-                               then ' '
-                               else c)
-
-
-hPutHeaders :: Handle -> Headers -> IO ()
-hPutHeaders h hds
-    = h `seq` hds `seq`
-      mapM_ putH (M.toList hds) >> C8.hPut h (C8.pack "\r\n")
+                  return (name, joinValues values)
+
+      content ∷ Parser Ascii
+      {-# INLINE content #-}
+      content = A.unsafeFromByteString
+                <$>
+                takeWhile1 (\c → isText c ∧ c ≢ '\x20')
+
+      joinValues ∷ [Ascii] → Ascii
+      {-# INLINE joinValues #-}
+      joinValues = A.fromAsciiBuilder
+                   ∘ mconcat
+                   ∘ intersperse (A.toAsciiBuilder "\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 C8.hPut h (fromNCBS name)
-               C8.hPut h (C8.pack ": ")
-               C8.hPut h value
-               C8.hPut 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"