variant using text internals in place of text streamDecodeUtf8

7 files changed, 550 insertions, 78 deletions

diff --git a/Pipes/Text.hs b/Pipes/Text.hs
index a5859a3..6845dd3 100644
--- a/Pipes/Text.hs
+++ b/Pipes/Text.hs
@@ -81,10 +81,6 @@ module Pipes.Text  (
     filter,
     scan,
     encodeUtf8,
-#if MIN_VERSION_text(0,11,4)
-    pipeDecodeUtf8,
-    pipeDecodeUtf8With,
-#endif
     pack,
     unpack,
     toCaseFold,
@@ -119,10 +115,8 @@ module Pipes.Text  (
     group,
     lines,
     words,
-#if MIN_VERSION_text(0,11,4)
     decodeUtf8,
     decodeUtf8With,
-#endif
     -- * Transformations
     intersperse,
     
@@ -167,6 +161,7 @@ import qualified GHC.IO.Exception as G
 import Pipes
 import qualified Pipes.ByteString as PB
 import qualified Pipes.ByteString.Parse as PBP
+import qualified Pipes.Text.Internal as PE
 import Pipes.Text.Parse (
     nextChar, drawChar, unDrawChar, peekChar, isEndOfChars )
 import Pipes.Core (respond, Server')
@@ -214,43 +209,60 @@ fromLazy  = foldrChunks (\e a -> yield e >> a) (return ())
 {-# INLINABLE fromLazy #-}
 
 -- | Stream text from 'stdin'
-stdin :: MonadIO m => Producer' Text m ()
+stdin :: MonadIO m => Producer' Text m (Producer ByteString m ())
 stdin = fromHandle IO.stdin
 {-# INLINABLE stdin #-}
 
 {-| Convert a 'IO.Handle' into a text stream using a text size 
     determined by the good sense of the text library. 
-
 -}
 
-fromHandle :: MonadIO m => IO.Handle -> Producer' Text m ()
-#if MIN_VERSION_text(0,11,4)
-fromHandle h = go TE.streamDecodeUtf8 where
+fromHandle :: MonadIO m => IO.Handle -> Producer' Text m (Producer ByteString m ())
+-- TODO: this should perhaps just be `decodeUtf8 (PB.fromHandle h)`
+-- if only so that mistakes can be concentrated in one place.
+-- This modifies something that was faster on an earlier iteration.
+-- Note also that the `text` replacement system is being ignored;
+-- with a replacement scheme one could have `Producer Text m ()`
+-- the relation to the replacement business needs to be thought out.
+-- The complicated type seems overmuch for the toy stdin above
+fromHandle h = go PE.streamDecodeUtf8 B.empty where
   act = B.hGetSome h defaultChunkSize
-  go dec = do chunk <- liftIO act
-              case dec chunk of 
-                TE.Some text _ dec' -> do yield text
-                                          unless (B.null chunk) (go dec')
+  go dec old = do chunk <- liftIO act
+                  if B.null chunk 
+                    then if B.null old then return (return ())
+                                       else return (yield old >> return ())
+                    else case dec chunk of 
+                           PE.Some text bs dec' -> 
+                              if T.null text then go dec' (B.append old bs) 
+                                             else do yield text
+                                                     go dec' B.empty
+                           PE.Other text bs ->
+                              if T.null text then return (do yield old
+                                                             yield bs
+                                                             PB.fromHandle h)
+                                             else do yield text
+                                                     return (do yield bs
+                                                                PB.fromHandle h)
 {-# INLINE fromHandle#-}
 -- bytestring fromHandle + streamDecodeUtf8 is 3 times as fast as
 -- the dedicated Text IO function 'hGetChunk' ;
 -- this way "runEffect $ PT.fromHandle hIn  >->  PT.toHandle hOut"
 -- runs the same as the conduit equivalent, only slightly slower 
 -- than "runEffect $ PB.fromHandle hIn  >->  PB.toHandle hOut"
-#else
-fromHandle h = go where
-    go = do txt <- liftIO (T.hGetChunk h)
-            unless (T.null txt) $ do yield txt
-                                     go
-{-# INLINABLE fromHandle#-}
-#endif
+-- #else
+-- fromHandle h = go where
+--     go = do txt <- liftIO (T.hGetChunk h)
+--             unless (T.null txt) $ do yield txt
+--                                      go
+-- {-# INLINABLE fromHandle#-}
+-- #endif
 {-| Stream text from a file using Pipes.Safe
 
 >>> runSafeT $ runEffect $ Text.readFile "hello.hs" >-> Text.map toUpper >-> hoist lift Text.stdout
 MAIN = PUTSTRLN "HELLO WORLD"
 -}
 
-readFile :: (MonadSafe m, Base m ~ IO) => FilePath -> Producer' Text m ()
+readFile :: (MonadSafe m, Base m ~ IO) => FilePath -> Producer' Text m (Producer ByteString m ())
 readFile file = Safe.withFile file IO.ReadMode fromHandle
 {-# INLINABLE readFile #-}
 
@@ -610,74 +622,44 @@ count :: (Monad m, Num n) => Text -> Producer Text m () -> m n
 count c p = P.fold (+) 0 id (p >-> P.map (fromIntegral . T.count c))
 {-# INLINABLE count #-}
 
-#if MIN_VERSION_text(0,11,4)
 -- | Transform a Pipe of 'ByteString's expected to be UTF-8 encoded
 -- into a Pipe of Text
 decodeUtf8
   :: Monad m
   => Producer ByteString m r -> Producer Text m (Producer ByteString m r)
-decodeUtf8 = go TE.streamDecodeUtf8
-  where go dec p = do
-            x <- lift (next p)
-            case x of
-                Left r -> return (return r)
-                Right (chunk, p') -> do
-                    let TE.Some text l dec' = dec chunk
-                    if B.null l
-                      then do
-                          yield text
-                          go dec' p'
-                      else return $ do
-                          yield l
-                          p'
+decodeUtf8 = decodeUtf8With Nothing
 {-# INLINEABLE decodeUtf8 #-}
 
 -- | Transform a Pipe of 'ByteString's expected to be UTF-8 encoded
 -- into a Pipe of Text with a replacement function of type @String -> Maybe Word8 -> Maybe Char@
 -- E.g. 'Data.Text.Encoding.Error.lenientDecode', which simply replaces bad bytes with \"�\"
-decodeUtf8With 
+decodeUtf8With
   :: Monad m  
-  => TE.OnDecodeError 
+  => Maybe TE.OnDecodeError 
   -> Producer ByteString m r -> Producer Text m (Producer ByteString m r)
-decodeUtf8With onErr = go (TE.streamDecodeUtf8With onErr)
-  where go dec p = do
-            x <- lift (next p)
-            case x of
-                Left r -> return (return r)
-                Right (chunk, p') -> do
-                    let TE.Some text l dec' = dec chunk
-                    if B.null l
-                      then do
-                          yield text
-                          go dec' p'
-                      else return $ do
-                          yield l
-                          p'
+decodeUtf8With onErr = go (PE.streamDecodeUtf8With onErr) B.empty where 
+  go dec old p = do
+    x <- lift (next p)
+    case x of
+      Left r -> if B.null old then return (return r)
+                              else return (do yield old 
+                                              return r)
+      Right (chunk, p') -> 
+        case dec chunk of 
+          PE.Some text l dec' -> 
+            if T.null text then go dec' (B.append old l) p'
+                           else do yield text
+                                   go dec' B.empty p'
+          PE.Other text bs ->
+            if T.null text then return (do yield old 
+                                           yield bs
+                                           p')
+                           else do yield text
+                                   return (do yield bs
+                                              p')
 {-# INLINEABLE decodeUtf8With #-}
 
--- | A simple pipe from 'ByteString' to 'Text'; a decoding error will arise
--- with any chunk that contains a sequence of bytes that is unreadable. Otherwise
--- only few bytes will only be moved from one chunk to the next before decoding.
-pipeDecodeUtf8 :: Monad m => Pipe ByteString Text m r
-pipeDecodeUtf8 = go TE.streamDecodeUtf8
-  where go dec = do chunk <- await
-                    case dec chunk of 
-                      TE.Some text l dec' -> do yield text
-                                                go dec'
-{-# INLINEABLE pipeDecodeUtf8 #-}
-
--- | A simple pipe from 'ByteString' to 'Text' using a replacement function.
-pipeDecodeUtf8With 
-  :: Monad m  
-  => TE.OnDecodeError 
-  -> Pipe ByteString Text m r 
-pipeDecodeUtf8With onErr = go (TE.streamDecodeUtf8With onErr)
-  where go dec = do chunk <- await
-                    case dec chunk of 
-                      TE.Some text l dec' -> do yield text
-                                                go dec'
-{-# INLINEABLE pipeDecodeUtf8With #-}
-#endif
+
 
 -- | Splits a 'Producer' after the given number of characters
 splitAt
diff --git a/Pipes/Text/Internal.hs b/Pipes/Text/Internal.hs
new file mode 100644
index 0000000..05d9887
--- /dev/null
+++ b/Pipes/Text/Internal.hs
@@ -0,0 +1,157 @@
+{-# LANGUAGE BangPatterns, CPP, ForeignFunctionInterface, GeneralizedNewtypeDeriving, MagicHash,
+    UnliftedFFITypes #-}
+-- This module lifts material from Brian O'Sullivan's text package 
+-- especially Data.Text.Encoding in order to define a pipes-appropriate
+-- streamDecodeUtf8
+module Pipes.Text.Internal 
+    ( Decoding(..)
+    , streamDecodeUtf8With
+    , streamDecodeUtf8
+    ) where
+
+import Control.Exception (evaluate, try)
+#if __GLASGOW_HASKELL__ >= 702
+import Control.Monad.ST.Unsafe (unsafeIOToST, unsafeSTToIO)
+import Control.Monad.ST (ST, runST)
+#else
+import Control.Monad.ST (unsafeIOToST, unsafeSTToIO, ST, runST)
+#endif
+import Data.Bits ((.&.))
+import Data.ByteString as B
+import Data.ByteString.Internal as B
+import Data.Text ()
+import Data.Text.Encoding.Error (OnDecodeError, UnicodeException, strictDecode)
+import Data.Text.Internal (Text(..), safe, textP)
+import Data.Word (Word8, Word32)
+import Foreign.C.Types (CSize)
+import Foreign.ForeignPtr (withForeignPtr)
+import Foreign.Marshal.Utils (with)
+import Foreign.Ptr (Ptr, minusPtr, nullPtr, plusPtr)
+import Foreign.Storable (Storable, peek, poke)
+import GHC.Base hiding (ord)
+import GHC.Word
+import qualified Data.Text.Array as A
+import GHC.Exts (Char(..), Int(..), chr#, ord#, word2Int#)
+import GHC.Word (Word8(..), Word16(..), Word32(..))
+
+import Data.Text.Unsafe (unsafeDupablePerformIO)
+
+#include "pipes_text_cbits.h"
+
+-- | A stream oriented decoding result.
+data Decoding = Some Text ByteString (ByteString -> Decoding)
+              | Other Text ByteString
+instance Show Decoding where
+    showsPrec d (Some t bs _) = showParen (d > prec) $
+                                showString "Some " . showsPrec prec' t .
+                                showChar ' ' . showsPrec prec' bs .
+                                showString " _"
+      where prec = 10; prec' = prec + 1
+    showsPrec d (Other t bs)  = showParen (d > prec) $
+                                showString "Other " . showsPrec prec' t .
+                                showChar ' ' . showsPrec prec' bs .
+                                showString " _"
+      where prec = 10; prec' = prec + 1
+      
+newtype CodePoint = CodePoint Word32 deriving (Eq, Show, Num, Storable)
+newtype DecoderState = DecoderState Word32 deriving (Eq, Show, Num, Storable)
+
+-- | Decode, in a stream oriented way, a 'ByteString' containing UTF-8
+-- encoded text that is known to be valid.
+--
+-- If the input contains any invalid UTF-8 data, an exception will be
+-- thrown (either by this function or a continuation) that cannot be
+-- caught in pure code.  For more control over the handling of invalid
+-- data, use 'streamDecodeUtf8With'.
+streamDecodeUtf8 :: ByteString -> Decoding
+streamDecodeUtf8 = streamDecodeUtf8With (Just strictDecode)
+
+-- | Decode, in a stream oriented way, a 'ByteString' containing UTF-8
+-- encoded text.
+streamDecodeUtf8With :: Maybe OnDecodeError -> ByteString -> Decoding
+streamDecodeUtf8With mErr = case mErr of 
+    Nothing    -> decodeWith False strictDecode 
+    Just onErr -> decodeWith True onErr 
+ where
+  -- We create a slightly larger than necessary buffer to accommodate a
+  -- potential surrogate pair started in the last buffer
+ decodeWith replace onErr = decodeChunk 0 0
+  where
+  decodeChunk :: CodePoint -> DecoderState -> ByteString -> Decoding
+  decodeChunk codepoint0 state0 bs@(PS fp off len) =
+    runST $ (unsafeIOToST . decodeChunkToBuffer) =<< A.new (len+1)
+   where
+    decodeChunkToBuffer :: A.MArray s -> IO Decoding
+    decodeChunkToBuffer dest = withForeignPtr fp $ \ptr ->
+      with (0::CSize) $ \destOffPtr ->
+      with codepoint0 $ \codepointPtr ->
+      with state0 $ \statePtr ->
+      with nullPtr $ \curPtrPtr ->
+        let end = ptr `plusPtr` (off + len)
+            loop curPtr = do
+              poke curPtrPtr curPtr
+              curPtr' <- c_decode_utf8_with_state (A.maBA dest) destOffPtr
+                         curPtrPtr end codepointPtr statePtr
+              state <- peek statePtr
+              case state of
+                UTF8_REJECT ->  
+                  -- We encountered an encoding error
+                 if replace 
+                 then do 
+                  x <- peek curPtr'
+                  case onErr desc (Just x) of
+                    Nothing -> loop $ curPtr' `plusPtr` 1
+                    Just c -> do
+                      destOff <- peek destOffPtr
+                      w <- unsafeSTToIO $
+                           unsafeWrite dest (fromIntegral destOff) (safe c)
+                      poke destOffPtr (destOff + fromIntegral w)
+                      poke statePtr 0
+                      loop $ curPtr' `plusPtr` 1
+                 else do 
+                  n <- peek destOffPtr 
+                  chunkText <- unsafeSTToIO $ do
+                      arr <- A.unsafeFreeze dest
+                      return $! textP arr 0 (fromIntegral n)
+                  lastPtr <- peek curPtrPtr
+                  let left = lastPtr `minusPtr` curPtr
+                  return $ Other chunkText (B.drop left bs)
+                _ -> do
+                  -- We encountered the end of the buffer while decoding
+                  n <- peek destOffPtr
+                  codepoint <- peek codepointPtr
+                  chunkText <- unsafeSTToIO $ do
+                      arr <- A.unsafeFreeze dest
+                      return $! textP arr 0 (fromIntegral n)
+                  lastPtr <- peek curPtrPtr
+                  let left = lastPtr `minusPtr` curPtr
+                  return $ Some chunkText (B.drop left bs)
+                           (decodeChunk codepoint state)
+        in loop (ptr `plusPtr` off)
+  desc = "Data.Text.Encoding.streamDecodeUtf8With: Invalid UTF-8 stream"
+
+ord :: Char -> Int
+ord (C# c#) = I# (ord# c#)
+{-# INLINE ord #-}
+
+
+unsafeWrite :: A.MArray s -> Int -> Char -> ST s Int
+unsafeWrite marr i c
+    | n < 0x10000 = do 
+        A.unsafeWrite marr i (fromIntegral n)
+        return 1
+    | otherwise = do
+        A.unsafeWrite marr i lo
+        A.unsafeWrite marr (i+1) hi
+        return 2
+    where n = ord c
+          m = n - 0x10000
+          lo = fromIntegral $ (m `shiftR` 10) + 0xD800
+          hi = fromIntegral $ (m .&. 0x3FF) + 0xDC00
+          shiftR (I# x#) (I# i#) = I# (x# `iShiftRA#` i#)
+{-# INLINE unsafeWrite #-}
+
+foreign import ccall unsafe "_hs_pipes_text_decode_utf8_state" c_decode_utf8_with_state
+    :: MutableByteArray# s -> Ptr CSize
+    -> Ptr (Ptr Word8) -> Ptr Word8
+    -> Ptr CodePoint -> Ptr DecoderState -> IO (Ptr Word8)
\ No newline at end of file
diff --git a/cbits/cbits.c b/cbits/cbits.c
new file mode 100644
index 0000000..e0fdfd5
--- /dev/null
+++ b/cbits/cbits.c
@@ -0,0 +1,151 @@
+/*
+ * Copyright (c) 2011 Bryan O'Sullivan <bos@serpentine.com>.
+ *
+ * Portions copyright (c) 2008-2010 Björn Höhrmann <bjoern@hoehrmann.de>.
+ *
+ * See http://bjoern.hoehrmann.de/utf-8/decoder/dfa/ for details.
+ */
+
+#include <string.h>
+#include <stdint.h>
+#include <stdio.h>
+#include "pipes_text_cbits.h"
+
+
+
+#define UTF8_ACCEPT 0
+#define UTF8_REJECT 12
+
+static const uint8_t utf8d[] = {
+  /*
+   * The first part of the table maps bytes to character classes that
+   * to reduce the size of the transition table and create bitmasks.
+   */
+   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+   0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,  0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,
+   1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,  9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,9,
+   7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,  7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,7,
+   8,8,2,2,2,2,2,2,2,2,2,2,2,2,2,2,  2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,2,
+  10,3,3,3,3,3,3,3,3,3,3,3,3,4,3,3, 11,6,6,6,5,8,8,8,8,8,8,8,8,8,8,8,
+
+  /*
+   * The second part is a transition table that maps a combination of
+   * a state of the automaton and a character class to a state.
+   */
+   0,12,24,36,60,96,84,12,12,12,48,72, 12,12,12,12,12,12,12,12,12,12,12,12,
+  12, 0,12,12,12,12,12, 0,12, 0,12,12, 12,24,12,12,12,12,12,24,12,24,12,12,
+  12,12,12,12,12,12,12,24,12,12,12,12, 12,24,12,12,12,12,12,12,12,24,12,12,
+  12,12,12,12,12,12,12,36,12,36,12,12, 12,36,12,12,12,12,12,36,12,36,12,12,
+  12,36,12,12,12,12,12,12,12,12,12,12,
+};
+
+static inline uint32_t
+decode(uint32_t *state, uint32_t* codep, uint32_t byte) {
+  uint32_t type = utf8d[byte];
+
+  *codep = (*state != UTF8_ACCEPT) ?
+    (byte & 0x3fu) | (*codep << 6) :
+    (0xff >> type) & (byte);
+
+  return *state = utf8d[256 + *state + type];
+}
+
+/*
+ * A best-effort decoder. Runs until it hits either end of input or
+ * the start of an invalid byte sequence.
+ *
+ * At exit, we update *destoff with the next offset to write to, *src
+ * with the next source location past the last one successfully
+ * decoded, and return the next source location to read from.
+ *
+ * Moreover, we expose the internal decoder state (state0 and
+ * codepoint0), allowing one to restart the decoder after it
+ * terminates (say, due to a partial codepoint).
+ *
+ * In particular, there are a few possible outcomes,
+ *
+ *   1) We decoded the buffer entirely:
+ *      In this case we return srcend
+ *      state0 == UTF8_ACCEPT
+ *
+ *   2) We met an invalid encoding
+ *      In this case we return the address of the first invalid byte
+ *      state0 == UTF8_REJECT
+ *
+ *   3) We reached the end of the buffer while decoding a codepoint
+ *      In this case we return a pointer to the first byte of the partial codepoint
+ *      state0 != UTF8_ACCEPT, UTF8_REJECT
+ *
+ */
+const uint8_t *
+_hs_pipes_text_decode_utf8_state(uint16_t *const dest, size_t *destoff,
+                           const uint8_t **const src,
+                           const uint8_t *const srcend,
+                           uint32_t *codepoint0, uint32_t *state0)
+{
+  uint16_t *d = dest + *destoff;
+  const uint8_t *s = *src, *last = *src;
+  uint32_t state = *state0;
+  uint32_t codepoint = *codepoint0;
+
+  while (s < srcend) {
+#if defined(__i386__) || defined(__x86_64__)
+    /*
+     * This code will only work on a little-endian system that
+     * supports unaligned loads.
+     *
+     * It gives a substantial speed win on data that is purely or
+     * partly ASCII (e.g. HTML), at only a slight cost on purely
+     * non-ASCII text.
+     */
+
+    if (state == UTF8_ACCEPT) {
+      while (s < srcend - 4) {
+        codepoint = *((uint32_t *) s);
+        if ((codepoint & 0x80808080) != 0)
+          break;
+        s += 4;
+
+        /*
+         * Tried 32-bit stores here, but the extra bit-twiddling
+         * slowed the code down.
+         */
+
+        *d++ = (uint16_t) (codepoint & 0xff);
+        *d++ = (uint16_t) ((codepoint >> 8) & 0xff);
+        *d++ = (uint16_t) ((codepoint >> 16) & 0xff);
+        *d++ = (uint16_t) ((codepoint >> 24) & 0xff);
+      }
+      last = s;
+    }
+#endif
+
+    if (decode(&state, &codepoint, *s++) != UTF8_ACCEPT) {
+      if (state != UTF8_REJECT)
+        continue;
+      break;
+    }
+
+    if (codepoint <= 0xffff)
+      *d++ = (uint16_t) codepoint;
+    else {
+      *d++ = (uint16_t) (0xD7C0 + (codepoint >> 10));
+      *d++ = (uint16_t) (0xDC00 + (codepoint & 0x3FF));
+    }
+    last = s;
+  }
+
+  /* Invalid encoding, back up to the errant character */
+  if (state == UTF8_REJECT)
+    s -= 1;
+
+  *destoff = d - dest;
+  *codepoint0 = codepoint;
+  *state0 = state;
+  *src = last;
+
+  return s;
+}
+
diff --git a/include/pipes_text_cbits.h b/include/pipes_text_cbits.h
new file mode 100644
index 0000000..b9ab670
--- /dev/null
+++ b/include/pipes_text_cbits.h
@@ -0,0 +1,11 @@
+/*
+ * Copyright (c) 2013 Bryan O'Sullivan <bos@serpentine.com>.
+ */
+
+#ifndef _pipes_text_cbits_h
+#define _pipes_text_cbits_h
+
+#define UTF8_ACCEPT 0
+#define UTF8_REJECT 12
+
+#endif
diff --git a/pipes-text.cabal b/pipes-text.cabal
index e79f168..86fbab8 100644
--- a/pipes-text.cabal
+++ b/pipes-text.cabal
@@ -12,7 +12,9 @@ build-type:          Simple
 cabal-version:       >=1.10
 
 library
-  exposed-modules:     Pipes.Text, Pipes.Text.Parse
+  c-sources:    cbits/cbits.c
+  include-dirs: include
+  exposed-modules:     Pipes.Text, Pipes.Text.Parse, Pipes.Text.Internal
   -- other-modules:       
   other-extensions:    RankNTypes
   build-depends:       base         >= 4       && < 5  ,
diff --git a/test/Test.hs b/test/Test.hs
new file mode 100644
index 0000000..1579f2b
--- /dev/null
+++ b/test/Test.hs
@@ -0,0 +1,60 @@
+import Utils
+
+import Test.QuickCheck
+import Test.QuickCheck.Monadic
+import Test.Framework (Test, testGroup, defaultMain)
+import Test.Framework.Providers.QuickCheck2 (testProperty)
+
+import Control.Exception (catch)
+import Data.Char (chr, isDigit, isHexDigit, isLower, isSpace, isUpper, ord)
+import Data.Monoid (Monoid(..))
+import Data.String (fromString)
+import Data.Text.Encoding.Error
+import qualified Data.List as L
+
+import qualified Data.Bits as Bits (shiftL, shiftR)
+import qualified Data.ByteString as B
+import qualified Data.ByteString.Lazy as BL
+import qualified Data.Text as T
+import qualified Data.Text.Lazy as TL
+import qualified Data.Text.Encoding as E
+import qualified Pipes.Text.Internal as PE
+
+main :: IO ()
+main = defaultMain [tests]
+-- >>> :main  -a 10000
+
+tests = testGroup "stream_decode" [
+
+  testProperty "t_utf8_incr_valid" t_utf8_incr_valid,
+  testProperty "t_utf8_incr_mixed" t_utf8_incr_mixed]
+
+t_utf8_incr_valid  = do
+        Positive n <- arbitrary
+        forAll genUnicode $ recode n `eq` id
+    where recode n = T.concat . feedChunksOf n PE.streamDecodeUtf8 . E.encodeUtf8
+          feedChunksOf :: Int -> (B.ByteString -> PE.Decoding) -> B.ByteString
+                       -> [T.Text]
+          feedChunksOf n f bs
+            | B.null bs  = []
+            | otherwise  = let (a,b) = B.splitAt n bs
+                               PE.Some t _ f' = f a
+                           in case f a of 
+                              PE.Some t _ f' -> t : feedChunksOf n f' b
+                              _             -> []
+
+t_utf8_incr_mixed  = do    
+       Positive n <- arbitrary  
+       txt <- genUnicode
+       forAll (vector 9) $ (roundtrip . chunk (mod n 7 + 1) . appendBytes txt) `eq` appendBytes txt
+    where 
+    roundtrip :: [B.ByteString] -> B.ByteString
+    roundtrip bss = go (PE.streamDecodeUtf8With Nothing) B.empty B.empty bss where                                                      
+       go dec acc old [] = acc <> old
+       go dec acc old (bs:bss) = case dec bs of 
+         PE.Some t new dec' -> if T.null t then go dec' (acc <> E.encodeUtf8 t) (old <> new) bss
+                                           else go dec' (acc <> E.encodeUtf8 t) new bss
+         PE.Other t bs' -> if T.null t then acc <> old <> bs <> B.concat bss 
+                                       else acc <> E.encodeUtf8 t <> bs' <> B.concat bss 
+    chunk n bs = let (a,b) = B.splitAt n bs in if B.null a then [] else a : chunk n b
+    appendBytes txt bts = E.encodeUtf8 txt <> B.pack bts ; (<>) = B.append
diff --git a/test/Utils.hs b/test/Utils.hs
new file mode 100644
index 0000000..75cd1db
--- /dev/null
+++ b/test/Utils.hs
@@ -0,0 +1,109 @@
+{-#LANGUAGE ScopedTypeVariables#-}
+module Utils where
+import Control.Exception (SomeException, bracket, bracket_, evaluate, try)
+import System.IO.Unsafe (unsafePerformIO)
+import Debug.Trace (trace)
+import Data.Bits ((.&.))
+import Data.Char (chr)
+import Data.String (IsString, fromString)
+import System.Random (Random (..), RandomGen)
+import Test.QuickCheck hiding ((.&.))
+import Test.QuickCheck.Monadic (assert, monadicIO, run)
+import qualified Data.ByteString as B
+import Pipes.Text.Internal
+
+
+
+
+
+-- Ensure that two potentially bottom values (in the sense of crashing
+-- for some inputs, not looping infinitely) either both crash, or both
+-- give comparable results for some input.
+(=^=) :: (Eq a, Show a) => a -> a -> Bool
+i =^= j = unsafePerformIO $ do
+  x <- try (evaluate i)
+  y <- try (evaluate j)
+  case (x,y) of
+    (Left (_ :: SomeException), Left (_ :: SomeException))
+                       -> return True
+    (Right a, Right b) -> return (a == b)
+    e                  -> trace ("*** Divergence: " ++ show e) return False
+infix 4 =^=
+{-# NOINLINE (=^=) #-}
+
+-- Do two functions give the same answer?
+eq :: (Eq a, Show a) => (t -> a) -> (t -> a) -> t -> Bool
+eq a b s  = a s =^= b s
+
+-- What about with the RHS packed?
+-- eqP :: (Eq a, Show a, Stringy s) =>
+--        (String -> a) -> (s -> a) -> String -> Word8 -> Bool
+-- eqP f g s w  = eql "orig" (f s) (g t) &&
+--                eql "mini" (f s) (g mini) &&
+--                eql "head" (f sa) (g ta) &&
+--                eql "tail" (f sb) (g tb)
+--     where t             = packS s
+--           mini          = packSChunkSize 10 s
+--           (sa,sb)       = splitAt m s
+--           (ta,tb)       = splitAtS m t
+--           l             = length s
+--           m | l == 0    = n
+--             | otherwise = n `mod` l
+--           n             = fromIntegral w
+--           eql d a b
+--             | a =^= b   = True
+--             | otherwise = trace (d ++ ": " ++ show a ++ " /= " ++ show b) False
+
+
+instance Arbitrary B.ByteString where
+    arbitrary     = B.pack `fmap` arbitrary
+
+genUnicode :: IsString a => Gen a
+genUnicode = fmap fromString string where
+    string = sized $ \n ->
+        do k <- choose (0,n)
+           sequence [ char | _ <- [1..k] ]
+
+    excluding :: [a -> Bool] -> Gen a -> Gen a
+    excluding bad gen = loop
+      where
+        loop = do
+          x <- gen
+          if or (map ($ x) bad)
+            then loop
+            else return x
+
+    reserved = [lowSurrogate, highSurrogate, noncharacter]
+    lowSurrogate c = c >= 0xDC00 && c <= 0xDFFF
+    highSurrogate c = c >= 0xD800 && c <= 0xDBFF
+    noncharacter c = masked == 0xFFFE || masked == 0xFFFF
+      where
+        masked = c .&. 0xFFFF
+
+    ascii = choose (0,0x7F)
+    plane0 = choose (0xF0, 0xFFFF)
+    plane1 = oneof [ choose (0x10000, 0x10FFF)
+                   , choose (0x11000, 0x11FFF)
+                   , choose (0x12000, 0x12FFF)
+                   , choose (0x13000, 0x13FFF)
+                   , choose (0x1D000, 0x1DFFF)
+                   , choose (0x1F000, 0x1FFFF)
+                   ]
+    plane2 = oneof [ choose (0x20000, 0x20FFF)
+                   , choose (0x21000, 0x21FFF)
+                   , choose (0x22000, 0x22FFF)
+                   , choose (0x23000, 0x23FFF)
+                   , choose (0x24000, 0x24FFF)
+                   , choose (0x25000, 0x25FFF)
+                   , choose (0x26000, 0x26FFF)
+                   , choose (0x27000, 0x27FFF)
+                   , choose (0x28000, 0x28FFF)
+                   , choose (0x29000, 0x29FFF)
+                   , choose (0x2A000, 0x2AFFF)
+                   , choose (0x2B000, 0x2BFFF)
+                   , choose (0x2F000, 0x2FFFF)
+                   ]
+    plane14 = choose (0xE0000, 0xE0FFF)
+    planes = [ascii, plane0, plane1, plane2, plane14]
+
+    char = chr `fmap` excluding reserved (oneof planes)