Reimplement MultipartForm

[Lucu.git] / Network / HTTP / Lucu / Utils.hs

diff --git a/Network/HTTP/Lucu/Utils.hs b/Network/HTTP/Lucu/Utils.hs
index dbc65ac17ad8a52f553af26e5506a33df9aab137..3d38b8b3aec36c9dff990cb4c7e66d05995ad27e 100644 (file)
--- a/Network/HTTP/Lucu/Utils.hs
+++ b/Network/HTTP/Lucu/Utils.hs
@@ -1,74 +1,68 @@
 {-# LANGUAGE
-    BangPatterns
+    OverloadedStrings
   , UnicodeSyntax
   #-}
 -- |Utility functions used internally in the Lucu httpd. These
 -- functions may be useful too for something else.
 module Network.HTTP.Lucu.Utils
     ( splitBy
-    , joinWith
-    , trim
-    , isWhiteSpace
     , quoteStr
     , parseWWWFormURLEncoded
+    , splitPathInfo
+    , show3
     )
     where
+import Blaze.ByteString.Builder.ByteString as B
+import Blaze.Text.Int as BT
 import Control.Monad
-import Data.List     hiding (last)
+import Data.Ascii (Ascii, AsciiBuilder)
+import qualified Data.Ascii as A
+import Data.ByteString (ByteString)
+import qualified Data.ByteString.Char8 as BS
+import Data.List hiding (last)
+import Data.Monoid.Unicode
 import Network.URI
-import Prelude       hiding (last)
+import Prelude hiding (last)
 import Prelude.Unicode
 
 -- |> splitBy (== ':') "ab:c:def"
 --  > ==> ["ab", "c", "def"]
-splitBy :: (a -> Bool) -> [a] -> [[a]]
+splitBy ∷ (a → Bool) → [a] → [[a]]
+{-# INLINEABLE splitBy #-}
 splitBy isSep src
-    = case break isSep src
-      of (last , []       ) -> [last]
-         (first, _sep:rest) -> first : splitBy isSep rest
-
--- |> joinWith ":" ["ab", "c", "def"]
---  > ==> "ab:c:def"
-joinWith :: [a] -> [[a]] -> [a]
-joinWith = (join .) . intersperse
-
--- |> trim (== '_') "__ab_c__def___"
---  > ==> "ab_c__def"
-trim :: (a -> Bool) -> [a] -> [a]
-trim !p = trimTail . trimHead
-    where
-      trimHead = dropWhile p
-      trimTail = reverse . trimHead . reverse
-
--- |@'isWhiteSpace' c@ is 'Prelude.True' iff c is one of SP, HT, CR
--- and LF.
-isWhiteSpace :: Char -> Bool
-isWhiteSpace ' '  = True
-isWhiteSpace '\t' = True
-isWhiteSpace '\r' = True
-isWhiteSpace '\n' = True
-isWhiteSpace _    = False
-{-# INLINE isWhiteSpace #-}
+    = case break isSep src of
+        (last , []       ) → [last]
+        (first, _sep:rest) → first : splitBy isSep rest
 
 -- |> quoteStr "abc"
 --  > ==> "\"abc\""
 --
 --  > quoteStr "ab\"c"
 --  > ==> "\"ab\\\"c\""
-quoteStr :: String -> String
-quoteStr !str = concat (["\""] ++ map quote str ++ ["\""])
+quoteStr ∷ Ascii → AsciiBuilder
+quoteStr str = A.toAsciiBuilder "\"" ⊕
+               go (A.toByteString str) (∅) ⊕
+               A.toAsciiBuilder "\""
     where
-      quote :: Char -> String
-      quote '"' = "\\\""
-      quote c   = [c]
+      go ∷ BS.ByteString → AsciiBuilder → AsciiBuilder
+      go bs ab
+          = case BS.break (≡ '"') bs of
+              (x, y)
+                  | BS.null y → ab ⊕ b2ab x
+                  | otherwise → go (BS.tail y) (ab ⊕ b2ab x
+                                                   ⊕ A.toAsciiBuilder "\\\"")
 
+      b2ab ∷ BS.ByteString → AsciiBuilder
+      b2ab = A.toAsciiBuilder ∘ A.unsafeFromByteString
 
 -- |> parseWWWFormURLEncoded "aaa=bbb&ccc=ddd"
 --  > ==> [("aaa", "bbb"), ("ccc", "ddd")]
-parseWWWFormURLEncoded ∷ String → [(String, String)]
+parseWWWFormURLEncoded ∷ Ascii → [(ByteString, ByteString)]
 parseWWWFormURLEncoded src
-    | null src  = []
-    | otherwise = do pairStr ← splitBy (\ c → c ≡ ';' ∨ c ≡ '&') src
+    -- THINKME: We could gain some performance by using attoparsec
+    -- here.
+    | src ≡ ""  = []
+    | otherwise = do pairStr ← splitBy (\ c → c ≡ ';' ∨ c ≡ '&') (A.toString src)
                      let (key, value) = break (≡ '=') pairStr
                      return ( unescape key
                             , unescape $ case value of
@@ -76,9 +70,31 @@ parseWWWFormURLEncoded src
                                            val       → val
                             )
     where
-      unescape ∷ String → String
-      unescape = unEscapeString ∘ map plusToSpace
+      unescape ∷ String → ByteString
+      unescape = BS.pack ∘ unEscapeString ∘ map plusToSpace
 
       plusToSpace ∷ Char → Char
       plusToSpace '+' = ' '
       plusToSpace c   = c
+
+-- |> splitPathInfo "http://example.com/foo/bar"
+--  > ==> ["foo", "bar"]
+splitPathInfo ∷ URI → [ByteString]
+splitPathInfo uri
+    = let reqPathStr = uriPath uri
+          reqPath    = [unEscapeString x | x ← splitBy (≡ '/') reqPathStr, (¬) (null x)]
+      in
+        map BS.pack reqPath
+
+-- |> show3 5
+--  > ==> "005"
+show3 ∷ Integral n ⇒ n → AsciiBuilder
+{-# INLINEABLE show3 #-}
+show3 = A.unsafeFromBuilder ∘ go
+    where
+      go i | i ≥ 0 ∧ i < 10   = B.fromByteString "00" ⊕ BT.digit    i
+           | i ≥ 0 ∧ i < 100  = B.fromByteString "0"  ⊕ BT.integral i
+           | i ≥ 0 ∧ i < 1000 =                         BT.integral i
+           | otherwise        = error ("show3: the integer i must satisfy 0 <= i < 1000: " ⧺ show i)
+-- FIXME: Drop this function as soon as possible, to eliminate the
+-- dependency on blaze-textual.