Pipes/Text/Internal/Codec.hs

   1
   2 {-# LANGUAGE DeriveDataTypeable, RankNTypes, BangPatterns #-}
   3 -- |
   4 -- Copyright: 2014 Michael Thompson, 2011 Michael Snoyman, 2010-2011 John Millikin
   5 -- License: MIT
   6 --  This Parts of this code were taken from enumerator and conduits, and adapted for pipes
   7
   8 -- This module follows the model of the enumerator and conduits libraries, and defines
   9 -- 'Codec' s for various encodings. Note that we do not export a 'Codec' for ascii and
  10 -- iso8859_1. A 'Lens' in the sense of the pipes library cannot be defined for these, so
  11 -- special functions appear in @Pipes.Text@
  12
  13
  14 module Pipes.Text.Internal.Codec
  15     ( Codec(..)
  16     , TextException(..)
  17     , utf8
  18     , utf16_le
  19     , utf16_be
  20     , utf32_le
  21     , utf32_be
  22     ) where
  23
  24 import Data.Bits ((.&.))
  25 import Data.Char (ord)
  26 import Data.ByteString as B
  27 import Data.ByteString (ByteString)
  28 import Data.ByteString.Internal as B
  29 import Data.ByteString.Char8 as B8
  30 import Data.Text (Text)
  31 import qualified Data.Text as T
  32 import qualified Data.Text.Encoding as TE
  33 import Data.Text.Encoding.Error ()
  34 import GHC.Word (Word8, Word32)
  35 import qualified Data.Text.Array as A
  36 import Data.Word (Word8, Word16)
  37 import System.IO.Unsafe (unsafePerformIO)
  38 import qualified Control.Exception as Exc
  39 import Data.Bits ((.&.), (.|.), shiftL)
  40 import Data.Typeable
  41 import Control.Arrow (first)
  42 import Data.Maybe (catMaybes)
  43 import Pipes.Text.Internal.Decoding
  44 import Pipes
  45 -- | A specific character encoding.
  46
  47 data Codec = Codec
  48   { codecName :: Text
  49   , codecEncode :: Text -> (ByteString, Maybe (TextException, Text))
  50   , codecDecode :: ByteString -> Decoding
  51   }
  52
  53 instance Show Codec where
  54     showsPrec d c = showParen (d > 10) $
  55                     showString "Codec " . shows (codecName c)
  56
  57 data TextException = DecodeException Codec Word8
  58                    | EncodeException Codec Char
  59                    | LengthExceeded Int
  60                    | TextException Exc.SomeException
  61     deriving (Show, Typeable)
  62 instance Exc.Exception TextException
  63
  64
  65 toDecoding :: (ByteString -> (Text, Either (TextException, ByteString) ByteString))
  66            -> (ByteString -> Decoding)
  67 toDecoding op = loop B.empty where
  68   loop !extra bs0 = case op (B.append extra bs0) of
  69                       (txt, Right bs) -> Some txt bs (loop bs)
  70                       (txt, Left (_,bs)) -> Other txt bs
  71 -- To do: toDecoding should be inlined in each of the 'Codec' definitions
  72 -- or else Codec changed to the conduit/enumerator definition.  We have
  73 -- altered it to use 'streamDecodeUtf8'
  74
  75 splitSlowly :: (ByteString -> Text)
  76             -> ByteString
  77             -> (Text, Either (TextException, ByteString) ByteString)
  78 splitSlowly dec bytes = valid where
  79     valid:_ = catMaybes $ Prelude.map decFirst $ splits (B.length bytes)
  80     splits 0 = [(B.empty, bytes)]
  81     splits n = B.splitAt n bytes : splits (n - 1)
  82     decFirst (a, b) = case tryEvaluate (dec a) of
  83         Left _ -> Nothing
  84         Right text -> let trouble = case tryEvaluate (dec b) of
  85                             Left exc -> Left (TextException exc, b)
  86                             Right _  -> Right B.empty
  87                       in Just (text, trouble) -- this case shouldn't occur,
  88                                       -- since splitSlowly is only called
  89                                       -- when parsing failed somewhere
  90
  91 utf8 :: Codec
  92 utf8 = Codec name enc (toDecoding dec) where
  93     name = T.pack "UTF-8"
  94     enc text = (TE.encodeUtf8 text, Nothing)
  95     dec bytes = case decodeSomeUtf8 bytes of (t,b) -> (t, Right b)
  96
  97 --     -- Whether the given byte is a continuation byte.
  98 --     isContinuation byte = byte .&. 0xC0 == 0x80
  99 --
 100 --     -- The number of continuation bytes needed by the given
 101 --     -- non-continuation byte. Returns -1 for an illegal UTF-8
 102 --     -- non-continuation byte and the whole split quickly must fail so
 103 --     -- as the input is passed to TE.decodeUtf8, which will issue a
 104 --     -- suitable error.
 105 --     required x0
 106 --         | x0 .&. 0x80 == 0x00 = 0
 107 --         | x0 .&. 0xE0 == 0xC0 = 1
 108 --         | x0 .&. 0xF0 == 0xE0 = 2
 109 --         | x0 .&. 0xF8 == 0xF0 = 3
 110 --         | otherwise           = -1
 111 --
 112 --     splitQuickly bytes
 113 --         | B.null l || req == -1 = Nothing
 114 --         | req == B.length r = Just (TE.decodeUtf8 bytes, B.empty)
 115 --         | otherwise = Just (TE.decodeUtf8 l', r')
 116 --       where
 117 --         (l, r) = B.spanEnd isContinuation bytes
 118 --         req = required (B.last l)
 119 --         l' = B.init l
 120 --         r' = B.cons (B.last l) r
 121
 122
 123 utf16_le :: Codec
 124 utf16_le = Codec name enc (toDecoding dec) where
 125     name = T.pack "UTF-16-LE"
 126     enc text = (TE.encodeUtf16LE text, Nothing)
 127     dec bytes = case splitQuickly bytes of
 128         Just (text, extra) -> (text, Right extra)
 129         Nothing -> splitSlowly TE.decodeUtf16LE bytes
 130
 131     splitQuickly bytes = maybeDecode (loop 0) where
 132         maxN = B.length bytes
 133
 134         loop n |  n      == maxN = decodeAll
 135                | (n + 1) == maxN = decodeTo n
 136         loop n = let
 137             req = utf16Required
 138                 (B.index bytes n)
 139                 (B.index bytes (n + 1))
 140             decodeMore = loop $! n + req
 141             in if n + req > maxN
 142                 then decodeTo n
 143                 else decodeMore
 144
 145         decodeTo n = first TE.decodeUtf16LE (B.splitAt n bytes)
 146         decodeAll = (TE.decodeUtf16LE bytes, B.empty)
 147
 148 utf16_be :: Codec
 149 utf16_be = Codec name enc (toDecoding dec) where
 150     name = T.pack "UTF-16-BE"
 151     enc text = (TE.encodeUtf16BE text, Nothing)
 152     dec bytes = case splitQuickly bytes of
 153         Just (text, extra) -> (text, Right extra)
 154         Nothing -> splitSlowly TE.decodeUtf16BE bytes
 155
 156     splitQuickly bytes = maybeDecode (loop 0) where
 157         maxN = B.length bytes
 158
 159         loop n |  n      == maxN = decodeAll
 160                | (n + 1) == maxN = decodeTo n
 161         loop n = let
 162             req = utf16Required
 163                 (B.index bytes (n + 1))
 164                 (B.index bytes n)
 165             decodeMore = loop $! n + req
 166             in if n + req > maxN
 167                 then decodeTo n
 168                 else decodeMore
 169
 170         decodeTo n = first TE.decodeUtf16BE (B.splitAt n bytes)
 171         decodeAll = (TE.decodeUtf16BE bytes, B.empty)
 172
 173 utf16Required :: Word8 -> Word8 -> Int
 174 utf16Required x0 x1 = if x >= 0xD800 && x <= 0xDBFF then 4 else 2 where
 175     x :: Word16
 176     x = (fromIntegral x1 `shiftL` 8) .|. fromIntegral x0
 177
 178
 179 utf32_le :: Codec
 180 utf32_le = Codec name enc (toDecoding dec) where
 181     name = T.pack "UTF-32-LE"
 182     enc text = (TE.encodeUtf32LE text, Nothing)
 183     dec bs = case utf32SplitBytes TE.decodeUtf32LE bs of
 184         Just (text, extra) -> (text, Right extra)
 185         Nothing -> splitSlowly TE.decodeUtf32LE bs
 186
 187
 188 utf32_be :: Codec
 189 utf32_be = Codec name enc (toDecoding dec) where
 190     name = T.pack "UTF-32-BE"
 191     enc text = (TE.encodeUtf32BE text, Nothing)
 192     dec bs = case utf32SplitBytes TE.decodeUtf32BE bs of
 193         Just (text, extra) -> (text, Right extra)
 194         Nothing -> splitSlowly TE.decodeUtf32BE bs
 195
 196 utf32SplitBytes :: (ByteString -> Text)
 197                 -> ByteString
 198                 -> Maybe (Text, ByteString)
 199 utf32SplitBytes dec bytes = split where
 200     split = maybeDecode (dec toDecode, extra)
 201     len = B.length bytes
 202     lenExtra = mod len 4
 203
 204     lenToDecode = len - lenExtra
 205     (toDecode, extra) = if lenExtra == 0
 206         then (bytes, B.empty)
 207         else B.splitAt lenToDecode bytes
 208
 209
 210 tryEvaluate :: a -> Either Exc.SomeException a
 211 tryEvaluate = unsafePerformIO . Exc.try . Exc.evaluate
 212
 213 maybeDecode :: (a, b) -> Maybe (a, b)
 214 maybeDecode (a, b) = case tryEvaluate a of
 215     Left _ -> Nothing
 216     Right _ -> Just (a, b)