diff --git a/benchmark/FileIO.hs b/benchmark/FileIO.hs
index b36166492..1bfc845a5 100644
--- a/benchmark/FileIO.hs
+++ b/benchmark/FileIO.hs
@@ -18,12 +18,12 @@ import Gauge
 import qualified Streamly.FileSystem.Handle as FH
 import qualified Streamly.Mem.Array as A
 import qualified Streamly.Prelude as S
-import qualified Streamly.String as SS
 
 #ifdef DEVBUILD
 import Data.Char (ord, chr)
 import System.IO (hSeek, SeekMode(..))
 import qualified Streamly.Fold as FL
+import qualified Streamly.String as SS
 #endif
 
 -- Input and output file handles
@@ -94,8 +94,8 @@ main = do
                 Handles inh _ <- readIORef href
                 let s = FH.readArrays inh
                 larr <- S.last s
-                return $ case larr of
-                    Nothing -> Nothing
+                case larr of
+                    Nothing -> return Nothing
                     Just arr -> A.readIndex arr (A.length arr - 1)
             -- Note: this cannot be fairly compared with GNU wc -c or wc -m as
             -- wc uses lseek to just determine the file size rather than reading
diff --git a/examples/HandleIO.hs b/examples/HandleIO.hs
index 52e198411..2ddd724d2 100644
--- a/examples/HandleIO.hs
+++ b/examples/HandleIO.hs
@@ -2,40 +2,52 @@ import qualified Streamly.Prelude as S
 import qualified Streamly.Fold as FL
 import qualified Streamly.Mem.Array as A
 import qualified Streamly.FileSystem.Handle as FH
+import qualified System.IO as FH
+-- import qualified Streamly.FileSystem.FD as FH
 import qualified Streamly.String as SS
 
 import Data.Char (ord)
 import System.Environment (getArgs)
-import System.IO (openFile, IOMode(..), stdout, Handle, hSeek, SeekMode(..))
+import System.IO (IOMode(..), hSeek, SeekMode(..))
 
-cat :: Handle -> IO ()
-cat src = FH.writeArrays stdout $ FH.readArraysUpto (256*1024) src
+cat :: FH.Handle -> IO ()
+cat src = FH.writeArrays FH.stdout $ FH.readArraysOfUpto (256*1024) src
+{-
+cat src =
+      FH.writeByChunksOf (1024*1024) FH.stdout
+    $ FH.readByChunksUpto (16*1024) src
+-}
 
-cp :: Handle -> Handle -> IO ()
-cp src dst = FH.writeArrays dst $ FH.readArraysUpto (256*1024) src
+cp :: FH.Handle -> FH.Handle -> IO ()
+cp src dst = FH.writeArrays dst $ FH.readArraysOfUpto (256*1024) src
+{-
+cp src dst =
+      FH.writeByChunksOf (1024*1024) dst
+    $ FH.readByChunksUpto (16*1024) src
+-}
 
 ord' :: Num a => Char -> a
 ord' = (fromIntegral . ord)
 
-wcl :: Handle -> IO ()
+wcl :: FH.Handle -> IO ()
 wcl src = print =<< (S.length
     $ flip SS.foldLines FL.drain
     $ SS.decodeChar8
     $ FH.read src)
 
-grepc :: String -> Handle -> IO ()
+grepc :: String -> FH.Handle -> IO ()
 grepc pat src = print . (subtract 1) =<< (S.length
     $ FL.splitOn (A.fromList (map ord' pat)) FL.drain
     $ FH.read src)
 
-avgll :: Handle -> IO ()
+avgll :: FH.Handle -> IO ()
 avgll src = print =<< (FL.foldl' avg
     $ FL.splitSuffixBy (== ord' '\n') FL.length
     $ FH.read src)
     where avg = (/) <$> toDouble FL.sum <*> toDouble FL.length
           toDouble = fmap (fromIntegral :: Int -> Double)
 
-llhisto :: Handle -> IO ()
+llhisto :: FH.Handle -> IO ()
 llhisto src = print =<< (FL.foldl' (FL.classify FL.length)
     $ S.map bucket
     $ FL.splitSuffixBy (== ord' '\n') FL.length
@@ -46,7 +58,7 @@ llhisto src = print =<< (FL.foldl' (FL.classify FL.length)
 main :: IO ()
 main = do
     name <- fmap head getArgs
-    src <- openFile name ReadMode
+    src <- FH.openFile name ReadMode
     let rewind = hSeek src AbsoluteSeek 0
 
     rewind >> putStrLn "cat"    >> cat src          -- Unix cat program
@@ -55,5 +67,5 @@ main = do
     rewind >> putStr "avgll "   >> avgll src        -- get average line length
     rewind >> putStr "llhisto " >> llhisto src      -- get line length histogram
 
-    dst <- openFile "dst-xyz.txt" WriteMode
+    dst <- FH.openFile "dst-xyz.txt" WriteMode
     rewind >> putStr "cp " >> cp src dst       -- Unix cp program
diff --git a/src/Streamly/FileSystem/FD.hs b/src/Streamly/FileSystem/FD.hs
new file mode 100644
index 000000000..1351cb501
--- /dev/null
+++ b/src/Streamly/FileSystem/FD.hs
@@ -0,0 +1,561 @@
+{-# LANGUAGE CPP #-}
+{-# LANGUAGE BangPatterns #-}
+{-# LANGUAGE MagicHash #-}
+{-# LANGUAGE RecordWildCards #-}
+{-# LANGUAGE UnboxedTuples #-}
+
+#include "inline.hs"
+
+-- |
+-- Module      : Streamly.FileSystem.FD
+-- Copyright   : (c) 2019 Composewell Technologies
+--
+-- License     : BSD3
+-- Maintainer  : harendra.kumar@gmail.com
+-- Stability   : experimental
+-- Portability : GHC
+--
+-- This module is a an experimental replacement for
+-- "Streamly.FileSystem.Handle". The former module provides IO facilities based
+-- on the GHC Handle type. The APIs in this module avoid the GHC handle layer
+-- and provide more explicit control over buffering.
+--
+-- Read and write data as streams and arrays to and from files.
+--
+-- This module provides read and write APIs based on handles. Before reading or
+-- writing, a file must be opened first using 'openFile'. The 'Handle' returned
+-- by 'openFile' is then used to access the file. A 'Handle' is backed by an
+-- operating system file descriptor. When the 'Handle' is garbage collected the
+-- underlying file descriptor is automatically closed. A handle can be
+-- explicitly closed using 'closeFile'.
+--
+-- Reading and writing APIs are divided into two categories, sequential
+-- streaming APIs and random or seekable access APIs.  File IO APIs are quite
+-- similar to "Streamly.Mem.Array" read write APIs. In that regard, arrays can
+-- be considered as in-memory files or files can be considered as on-disk
+-- arrays.
+--
+-- > import qualified Streamly.FileSystem.FD as FD
+--
+
+module Streamly.FileSystem.FD
+    (
+    -- * File Handles
+      Handle
+    , stdin
+    , stdout
+    , stderr
+    , openFile
+
+    -- TODO file path based APIs
+    -- , readFile
+    -- , writeFile
+
+    -- * Streaming IO
+    -- | Streaming APIs read or write data to or from a file or device
+    -- sequentially, they never perform a seek to a random location.  When
+    -- reading, the stream is lazy and generated on-demand as the consumer
+    -- consumes it.  Read IO requests to the IO device are performed in chunks
+    -- of 32KiB, this is referred to as @defaultChunkSize@ in the
+    -- documentation. One IO request may or may not read the full chunk. If the
+    -- whole stream is not consumed, it is possible that we may read slightly
+    -- more from the IO device than what the consumer needed.  Unless specified
+    -- otherwise in the API, writes are collected into chunks of
+    -- @defaultChunkSize@ before they are written to the IO device.
+
+    -- Streaming APIs work for all kind of devices, seekable or non-seekable;
+    -- including disks, files, memory devices, terminals, pipes, sockets and
+    -- fifos. While random access APIs work only for files or devices that have
+    -- random access or seek capability for example disks, memory devices.
+    -- Devices like terminals, pipes, sockets and fifos do not have random
+    -- access capability.
+
+    -- ** Read File to Stream
+    , read
+    -- , readUtf8
+    -- , readLines
+    -- , readFrames
+    , readByChunksUpto
+
+    -- -- * Array Read
+    -- , readArrayUpto
+    -- , readArrayOf
+
+    , readArrays
+    , readArraysOfUpto
+    -- , readArraysOf
+
+    -- ** Write File from Stream
+    , write
+    -- , writeUtf8
+    -- , writeUtf8Lines
+    -- , writeFrames
+    , writeByChunksOf
+
+    -- -- * Array Write
+    -- , writeArray
+    , writeArrays
+    , writeArraysOfUpto
+    , writev
+    , writevArraysOfUpto
+
+    -- -- * Random Access (Seek)
+    -- -- | Unlike the streaming APIs listed above, these APIs apply to devices or
+    -- files that have random access or seek capability.  This type of devices
+    -- include disks, files, memory devices and exclude terminals, pipes,
+    -- sockets and fifos.
+    --
+    -- , readIndex
+    -- , readSlice
+    -- , readSliceRev
+    -- , readAt -- read from a given position to th end of file
+    -- , readSliceArrayUpto
+    -- , readSliceArrayOf
+
+    -- , writeIndex
+    -- , writeSlice
+    -- , writeSliceRev
+    -- , writeAt -- start writing at the given position
+    -- , writeSliceArray
+    )
+where
+
+import Control.Monad.IO.Class (MonadIO(..))
+import Data.Word (Word8)
+import Foreign.ForeignPtr (withForeignPtr)
+-- import Foreign.ForeignPtr.Unsafe (unsafeForeignPtrToPtr)
+import Foreign.Ptr (plusPtr, castPtr)
+import Foreign.Storable (Storable(..))
+import GHC.ForeignPtr (mallocPlainForeignPtrBytes)
+-- import System.IO (Handle, hGetBufSome, hPutBuf)
+import System.IO (IOMode)
+import Prelude hiding (read)
+
+import qualified GHC.IO.FD as FD
+import qualified GHC.IO.Device as RawIO hiding (write)
+import qualified Streamly.FileSystem.FDIO as RawIO
+
+import Streamly.Mem.Array.Types (Array(..))
+import Streamly.Streams.Serial (SerialT)
+import Streamly.Streams.StreamD (toStreamD)
+import Streamly.Streams.StreamD.Type (fromStreamD)
+import Streamly.Streams.StreamK.Type (IsStream, mkStream)
+import Streamly.Mem.Array.Types (byteLength, defaultChunkSize, groupIOVecsOf)
+-- import Streamly.Fold (Fold)
+-- import Streamly.String (encodeUtf8, decodeUtf8, foldLines)
+
+import qualified Streamly.Mem.Array as A
+import qualified Streamly.Prelude as S
+import qualified Streamly.Streams.StreamD.Type as D
+
+-------------------------------------------------------------------------------
+-- References
+-------------------------------------------------------------------------------
+--
+-- The following references may be useful to build an understanding about the
+-- file API design:
+--
+-- http://www.linux-mag.com/id/308/ for blocking/non-blocking IO on linux.
+-- https://lwn.net/Articles/612483/ Non-blocking buffered file read operations
+-- https://en.wikipedia.org/wiki/C_file_input/output for C APIs.
+-- https://docs.oracle.com/javase/tutorial/essential/io/file.html for Java API.
+-- https://www.w3.org/TR/FileAPI/ for http file API.
+
+-------------------------------------------------------------------------------
+-- Handles
+-------------------------------------------------------------------------------
+
+-- XXX attach a finalizer
+-- | A 'Handle' is returned by 'openFile' and is subsequently used to perform
+-- read and write operations on a file.
+--
+newtype Handle = Handle FD.FD
+
+-- | File handle for standard input
+stdin :: Handle
+stdin = Handle FD.stdin
+
+-- | File handle for standard output
+stdout :: Handle
+stdout = Handle FD.stdout
+
+-- | File handle for standard error
+stderr :: Handle
+stderr = Handle FD.stderr
+
+-- XXX we can support all the flags that the "open" system call supports.
+-- Instead of using RTS locking mechanism can we use system provided locking
+-- instead?
+--
+-- | Open a file that is not a directory and return a file handle.
+-- 'openFile' enforces a multiple-reader single-writer locking on files. That
+-- is, there may either be many handles on the same file which manage input, or
+-- just one handle on the file which manages output. If any open handle is
+-- managing a file for output, no new handle can be allocated for that file. If
+-- any open handle is managing a file for input, new handles can only be
+-- allocated if they do not manage output. Whether two files are the same is
+-- implementation-dependent, but they should normally be the same if they have
+-- the same absolute path name and neither has been renamed, for example.
+--
+openFile :: FilePath -> IOMode -> IO Handle
+openFile path mode = fmap (Handle . fst) $ FD.openFile path mode True
+
+-------------------------------------------------------------------------------
+-- Array IO (Input)
+-------------------------------------------------------------------------------
+
+-- | Read a 'ByteArray' from a file handle. If no data is available on the
+-- handle it blocks until some data becomes available. If data is available
+-- then it immediately returns that data without blocking. It reads a maximum
+-- of up to the size requested.
+{-# INLINABLE readArrayUpto #-}
+readArrayUpto :: Int -> Handle -> IO (Array Word8)
+readArrayUpto size (Handle fd) = do
+    ptr <- mallocPlainForeignPtrBytes size
+    -- ptr <- mallocPlainForeignPtrAlignedBytes size (alignment (undefined :: Word8))
+    withForeignPtr ptr $ \p -> do
+        -- n <- hGetBufSome h p size
+        n <- RawIO.read fd p size
+        let v = Array
+                { aStart = ptr
+                , aEnd   = p `plusPtr` n
+                , aBound = p `plusPtr` size
+                }
+        -- XXX shrink only if the diff is significant
+        -- A.shrinkToFit v
+        return v
+
+-------------------------------------------------------------------------------
+-- Array IO (output)
+-------------------------------------------------------------------------------
+
+-- | Write an 'Array' to a file handle.
+--
+-- @since 0.7.0
+{-# INLINABLE writeArray #-}
+writeArray :: Storable a => Handle -> Array a -> IO ()
+writeArray _ arr | A.length arr == 0 = return ()
+writeArray (Handle fd) arr = withForeignPtr (aStart arr) $ \p -> do
+    RawIO.writeAll fd (castPtr p) aLen
+    {-
+    -- Experiment to compare "writev" based IO with "write" based IO.
+    iov <- A.newArray 1
+    let iov' = iov {aEnd = aBound iov}
+    A.writeIndex iov' 0 (RawIO.IOVec (castPtr p) (fromIntegral aLen))
+    RawIO.writevAll fd (unsafeForeignPtrToPtr (aStart iov')) 1
+    -}
+    where
+    aLen = byteLength arr
+
+-- | Write an array of 'IOVec' to a file handle.
+--
+-- @since 0.7.0
+{-# INLINABLE writeIOVec #-}
+writeIOVec :: Handle -> Array RawIO.IOVec -> IO ()
+writeIOVec _ iov | A.length iov == 0 = return ()
+writeIOVec (Handle fd) iov =
+    withForeignPtr (aStart iov) $ \p ->
+        RawIO.writevAll fd p (A.length iov)
+
+-------------------------------------------------------------------------------
+-- Stream of Arrays IO
+-------------------------------------------------------------------------------
+
+-- | @readArraysOfUpto size h@ reads a stream of arrays from file handle @h@.
+-- The maximum size of a single array is specified by @size@. The actual size
+-- read may be less than or equal to @size@.
+{-# INLINABLE _readArraysOfUpto #-}
+_readArraysOfUpto :: (IsStream t, MonadIO m)
+    => Int -> Handle -> t m (Array Word8)
+_readArraysOfUpto size h = go
+  where
+    -- XXX use cons/nil instead
+    go = mkStream $ \_ yld _ stp -> do
+        arr <- liftIO $ readArrayUpto size h
+        if A.length arr == 0
+        then stp
+        else yld arr go
+
+{-# INLINE_NORMAL readArraysOfUpto #-}
+readArraysOfUpto :: (IsStream t, MonadIO m)
+    => Int -> Handle -> t m (Array Word8)
+readArraysOfUpto size h = D.fromStreamD (D.Stream step ())
+  where
+    {-# INLINE_LATE step #-}
+    step _ _ = do
+        arr <- liftIO $ readArrayUpto size h
+        return $
+            case A.length arr of
+                0 -> D.Stop
+                _ -> D.Yield arr ()
+
+-- XXX read 'Array a' instead of Word8
+--
+-- | @readArrays h@ reads a stream of arrays from file handle @h@.
+-- The maximum size of a single array is limited to @defaultChunkSize@.
+-- 'readArrays' ignores the prevailing 'TextEncoding' and 'NewlineMode'
+-- on the 'Handle'.
+--
+-- > readArrays = readArraysOfUpto defaultChunkSize
+--
+-- @since 0.7.0
+{-# INLINE readArrays #-}
+readArrays :: (IsStream t, MonadIO m) => Handle -> t m (Array Word8)
+readArrays = readArraysOfUpto defaultChunkSize
+
+-------------------------------------------------------------------------------
+-- Read File to Stream
+-------------------------------------------------------------------------------
+
+-- TODO for concurrent streams implement readahead IO. We can send multiple
+-- read requests at the same time. For serial case we can use async IO. We can
+-- also control the read throughput in mbps or IOPS.
+
+-- | @readByChunksUpto chunkSize handle@ reads a byte stream from a file handle,
+-- reads are performed in chunks of up to @chunkSize@.  The stream ends as soon
+-- as EOF is encountered.
+--
+{-# INLINE readByChunksUpto #-}
+readByChunksUpto :: (IsStream t, MonadIO m) => Int -> Handle -> t m Word8
+readByChunksUpto chunkSize h = A.flattenArrays $ readArraysOfUpto chunkSize h
+
+-- TODO
+-- read :: (IsStream t, MonadIO m, Storable a) => Handle -> t m a
+--
+-- > read = 'readByChunks' A.defaultChunkSize
+-- | Generate a stream of elements of the given type from a file 'Handle'. The
+-- stream ends when EOF is encountered.
+--
+-- @since 0.7.0
+{-# INLINE read #-}
+read :: (IsStream t, MonadIO m) => Handle -> t m Word8
+read = A.flattenArrays . readArrays
+
+-------------------------------------------------------------------------------
+-- Writing
+-------------------------------------------------------------------------------
+
+-- | Write a stream of arrays to a handle.
+--
+-- @since 0.7.0
+{-# INLINE writeArrays #-}
+writeArrays :: (MonadIO m, Storable a) => Handle -> SerialT m (Array a) -> m ()
+writeArrays h m = S.mapM_ (liftIO . writeArray h) m
+
+-- | Write a stream of 'IOVec' arrays to a handle.
+--
+-- @since 0.7.0
+{-# INLINE writev #-}
+writev :: MonadIO m => Handle -> SerialT m (Array RawIO.IOVec) -> m ()
+writev h m = S.mapM_ (liftIO . writeIOVec h) m
+
+-- | Write a stream of arrays to a handle after coalescing them in chunks of
+-- specified size. The chunk size is only a maximum and the actual writes could
+-- be smaller than that as we do not split the arrays to fit them to the
+-- specified size.
+--
+-- @since 0.7.0
+{-# INLINE writeArraysOfUpto #-}
+writeArraysOfUpto :: (MonadIO m, Storable a)
+    => Int -> Handle -> SerialT m (Array a) -> m ()
+writeArraysOfUpto n h xs = writeArrays h $ A.coalesceChunksOf n xs
+
+-- | Write a stream of arrays to a handle after grouping them in 'IOVec' arrays
+-- of up to a maximum total size. Writes are performed using gather IO via
+-- @writev@ system call. The maximum number of entries in each 'IOVec' group
+-- limited to 512.
+--
+-- @since 0.7.0
+{-# INLINE writevArraysOfUpto #-}
+writevArraysOfUpto :: MonadIO m
+    => Int -> Handle -> SerialT m (Array a) -> m ()
+writevArraysOfUpto n h xs =
+    writev h $ fromStreamD $ groupIOVecsOf n 512 (toStreamD xs)
+
+-- GHC buffer size dEFAULT_FD_BUFFER_SIZE=8192 bytes.
+--
+-- XXX test this
+-- Note that if you use a chunk size less than 8K (GHC's default buffer
+-- size) then you are advised to use 'NOBuffering' mode on the 'Handle' in case you
+-- do not want buffering to occur at GHC level as well. Same thing applies to
+-- writes as well.
+
+-- | Like 'write' but provides control over the write buffer. Output will
+-- be written to the IO device as soon as we collect the specified number of
+-- input elements.
+--
+-- @since 0.7.0
+{-# INLINE writeByChunksOf #-}
+writeByChunksOf :: MonadIO m => Int -> Handle -> SerialT m Word8 -> m ()
+writeByChunksOf n h m = writeArrays h $ A.arraysOf n m
+
+-- > write = 'writeByChunks' A.defaultChunkSize
+--
+-- | Write a byte stream to a file handle. Combines the bytes in chunks of size
+-- up to 'A.defaultChunkSize' before writing.  Note that the write behavior
+-- depends on the 'IOMode' and the current seek position of the handle.
+--
+-- @since 0.7.0
+{-# INLINE write #-}
+write :: MonadIO m => Handle -> SerialT m Word8 -> m ()
+write = writeByChunksOf defaultChunkSize
+
+{-
+{-# INLINE write #-}
+write :: (MonadIO m, Storable a) => Handle -> SerialT m a -> m ()
+write = toHandleWith A.defaultChunkSize
+-}
+
+-------------------------------------------------------------------------------
+-- IO with encoding/decoding Unicode characters
+-------------------------------------------------------------------------------
+
+{-
+-- |
+-- > readUtf8 = decodeUtf8 . read
+--
+-- Read a UTF8 encoded stream of unicode characters from a file handle.
+--
+-- @since 0.7.0
+{-# INLINE readUtf8 #-}
+readUtf8 :: (IsStream t, MonadIO m) => Handle -> t m Char
+readUtf8 = decodeUtf8 . read
+
+-- |
+-- > writeUtf8 h s = write h $ encodeUtf8 s
+--
+-- Encode a stream of unicode characters to UTF8 and write it to the given file
+-- handle. Default block buffering applies to the writes.
+--
+-- @since 0.7.0
+{-# INLINE writeUtf8 #-}
+writeUtf8 :: MonadIO m => Handle -> SerialT m Char -> m ()
+writeUtf8 h s = write h $ encodeUtf8 s
+
+-- | Write a stream of unicode characters after encoding to UTF-8 in chunks
+-- separated by a linefeed character @'\n'@. If the size of the buffer exceeds
+-- @defaultChunkSize@ and a linefeed is not yet found, the buffer is written
+-- anyway.  This is similar to writing to a 'Handle' with the 'LineBuffering'
+-- option.
+--
+-- @since 0.7.0
+{-# INLINE writeUtf8ByLines #-}
+writeUtf8ByLines :: (IsStream t, MonadIO m) => Handle -> t m Char -> m ()
+writeUtf8ByLines = undefined
+
+-- | Read UTF-8 lines from a file handle and apply the specified fold to each
+-- line. This is similar to reading a 'Handle' with the 'LineBuffering' option.
+--
+-- @since 0.7.0
+{-# INLINE readLines #-}
+readLines :: (IsStream t, MonadIO m) => Handle -> Fold m Char b -> t m b
+readLines h f = foldLines (readUtf8 h) f
+
+-------------------------------------------------------------------------------
+-- Framing on a sequence
+-------------------------------------------------------------------------------
+
+-- | Read a stream from a file handle and split it into frames delimited by
+-- the specified sequence of elements. The supplied fold is applied on each
+-- frame.
+--
+-- @since 0.7.0
+{-# INLINE readFrames #-}
+readFrames :: (IsStream t, MonadIO m, Storable a)
+    => Array a -> Handle -> Fold m a b -> t m b
+readFrames = undefined -- foldFrames . read
+
+-- | Write a stream to the given file handle buffering up to frames separated
+-- by the given sequence or up to a maximum of @defaultChunkSize@.
+--
+-- @since 0.7.0
+{-# INLINE writeByFrames #-}
+writeByFrames :: (IsStream t, MonadIO m, Storable a)
+    => Array a -> Handle -> t m a -> m ()
+writeByFrames = undefined
+
+-------------------------------------------------------------------------------
+-- Random Access IO (Seek)
+-------------------------------------------------------------------------------
+
+-- XXX handles could be shared, so we may not want to use the handle state at
+-- all for these APIs. we can use pread and pwrite instead. On windows we will
+-- need to use readFile/writeFile with an offset argument.
+
+-------------------------------------------------------------------------------
+
+-- | Read the element at the given index treating the file as an array.
+--
+-- @since 0.7.0
+{-# INLINE readIndex #-}
+readIndex :: Storable a => Handle -> Int -> Maybe a
+readIndex arr i = undefined
+
+-- NOTE: To represent a range to read we have chosen (start, size) instead of
+-- (start, end). This helps in removing the ambiguity of whether "end" is
+-- included in the range or not.
+--
+-- We could avoid specifying the range to be read and instead use "take size"
+-- on the stream, but it may end up reading more and then consume it partially.
+
+-- | @readSliceWith chunkSize handle pos len@ reads up to @len@ bytes
+-- from @handle@ starting at the offset @pos@ from the beginning of the file.
+--
+-- Reads are performed in chunks of size @chunkSize@.  For block devices, to
+-- avoid reading partial blocks @chunkSize@ must align with the block size of
+-- the underlying device. If the underlying block size is unknown, it is a good
+-- idea to keep it a multiple 4KiB. This API ensures that the start of each
+-- chunk is aligned with @chunkSize@ from second chunk onwards.
+--
+{-# INLINE readSliceWith #-}
+readSliceWith :: (IsStream t, MonadIO m, Storable a)
+    => Int -> Handle -> Int -> Int -> t m a
+readSliceWith chunkSize h pos len = undefined
+
+-- | @readSlice h i count@ streams a slice from the file handle @h@ starting
+-- at index @i@ and reading up to @count@ elements in the forward direction
+-- ending at the index @i + count - 1@.
+--
+-- @since 0.7.0
+{-# INLINE readSlice #-}
+readSlice :: (IsStream t, MonadIO m, Storable a)
+    => Handle -> Int -> Int -> t m a
+readSlice = readSliceWith A.defaultChunkSize
+
+-- | @readSliceRev h i count@ streams a slice from the file handle @h@ starting
+-- at index @i@ and reading up to @count@ elements in the reverse direction
+-- ending at the index @i - count + 1@.
+--
+-- @since 0.7.0
+{-# INLINE readSliceRev #-}
+readSliceRev :: (IsStream t, MonadIO m, Storable a)
+    => Handle -> Int -> Int -> t m a
+readSliceRev h i count = undefined
+
+-- | Write the given element at the given index in the file.
+--
+-- @since 0.7.0
+{-# INLINE writeIndex #-}
+writeIndex :: (MonadIO m, Storable a) => Handle -> Int -> a -> m ()
+writeIndex h i a = undefined
+
+-- | @writeSlice h i count stream@ writes a stream to the file handle @h@
+-- starting at index @i@ and writing up to @count@ elements in the forward
+-- direction ending at the index @i + count - 1@.
+--
+-- @since 0.7.0
+{-# INLINE writeSlice #-}
+writeSlice :: (IsStream t, Monad m, Storable a)
+    => Handle -> Int -> Int -> t m a -> m ()
+writeSlice h i len s = undefined
+
+-- | @writeSliceRev h i count stream@ writes a stream to the file handle @h@
+-- starting at index @i@ and writing up to @count@ elements in the reverse
+-- direction ending at the index @i - count + 1@.
+--
+-- @since 0.7.0
+{-# INLINE writeSliceRev #-}
+writeSliceRev :: (IsStream t, Monad m, Storable a)
+    => Handle -> Int -> Int -> t m a -> m ()
+writeSliceRev arr i len s = undefined
+-}
diff --git a/src/Streamly/FileSystem/FDIO.hs b/src/Streamly/FileSystem/FDIO.hs
new file mode 100644
index 000000000..e68841573
--- /dev/null
+++ b/src/Streamly/FileSystem/FDIO.hs
@@ -0,0 +1,200 @@
+{-# LANGUAGE CPP #-}
+{-# LANGUAGE BangPatterns #-}
+{-# LANGUAGE MagicHash #-}
+{-# LANGUAGE RecordWildCards #-}
+{-# LANGUAGE UnboxedTuples #-}
+
+#include "inline.hs"
+
+-- |
+-- Module      : Streamly.FileSystem.FDIO
+-- Copyright   : (c) 2019 Composewell Technologies
+-- Copyright   : (c) 1994-2008 The University of Glasgow
+--
+-- License     : BSD3
+-- Maintainer  : harendra.kumar@gmail.com
+-- Stability   : experimental
+-- Portability : GHC
+--
+-- Low level IO routines interfacing the operating system.
+--
+
+module Streamly.FileSystem.FDIO
+    ( write
+    , writeAll
+    , IOVec (..)
+    , writev
+    , writevAll
+    )
+where
+
+import Control.Concurrent (threadWaitWrite)
+import Control.Monad (when)
+import Data.Int (Int64)
+import Data.Word (Word8)
+import Foreign.C.Error (throwErrnoIfMinus1RetryMayBlock)
+import Foreign.C.Types (CSize(..), CInt(..), CBool(..))
+import Foreign.Ptr (plusPtr, Ptr)
+import System.Posix.Internals (c_write, c_safe_write)
+
+import GHC.IO.FD (FD(..))
+
+import Streamly.FileSystem.IOVec (IOVec(..), c_writev, c_safe_writev)
+
+-------------------------------------------------------------------------------
+-- IO Routines
+-------------------------------------------------------------------------------
+
+-- See System.POSIX.Internals in GHC base package
+
+-------------------------------------------------------------------------------
+-- Write without blocking the underlying OS thread
+-------------------------------------------------------------------------------
+
+foreign import ccall unsafe "rtsSupportsBoundThreads" threaded :: Bool
+
+isNonBlocking :: FD -> Bool
+isNonBlocking fd = fdIsNonBlocking fd /= 0
+
+#if !defined(mingw32_HOST_OS)
+
+-- "poll"s the fd for data to become available or timeout
+-- See cbits/inputReady.c in base package
+foreign import ccall unsafe "fdReady"
+  unsafe_fdReady :: CInt -> CBool -> Int64 -> CBool -> IO CInt
+
+writeNonBlocking :: String -> FD -> Ptr Word8 -> Int -> CSize -> IO CInt
+writeNonBlocking loc !fd !buf !off !len
+    | isNonBlocking fd = unsafe_write -- unsafe is ok, it can't block
+    | otherwise   = do
+        let isWrite = 1
+            isSocket = 0
+            msecs = 0
+        r <- unsafe_fdReady (fdFD fd) isWrite msecs isSocket
+        when (r == 0) $ threadWaitWrite (fromIntegral (fdFD fd))
+        if threaded then safe_write else unsafe_write
+
+    where
+
+    do_write call = fromIntegral `fmap`
+                      throwErrnoIfMinus1RetryMayBlock loc call
+                        (threadWaitWrite (fromIntegral (fdFD fd)))
+    unsafe_write  = do_write (c_write (fdFD fd) (buf `plusPtr` off) len)
+    safe_write    = do_write (c_safe_write (fdFD fd) (buf `plusPtr` off) len)
+
+writevNonBlocking :: String -> FD -> Ptr IOVec -> Int -> IO CInt
+writevNonBlocking loc !fd !iov !cnt
+    | isNonBlocking fd = unsafe_write -- unsafe is ok, it can't block
+    | otherwise   = do
+        let isWrite = 1
+            isSocket = 0
+            msecs = 0
+        r <- unsafe_fdReady (fdFD fd) isWrite msecs isSocket
+        when (r == 0) $ threadWaitWrite (fromIntegral (fdFD fd))
+        if threaded then safe_write else unsafe_write
+
+    where
+
+    do_write call = fromIntegral `fmap`
+                      throwErrnoIfMinus1RetryMayBlock loc call
+                        (threadWaitWrite (fromIntegral (fdFD fd)))
+    unsafe_write  = do_write (c_writev (fdFD fd) iov (fromIntegral cnt))
+    safe_write    = do_write (c_safe_writev (fdFD fd) iov (fromIntegral cnt))
+
+#else
+
+foreign import WINDOWS_CCONV safe "recv"
+   c_safe_recv :: CInt -> Ptr Word8 -> CInt -> CInt{-flags-} -> IO CInt
+
+foreign import WINDOWS_CCONV safe "send"
+   c_safe_send :: CInt -> Ptr Word8 -> CInt -> CInt{-flags-} -> IO CInt
+
+blockingWriteRawBufferPtr :: String -> FD -> Ptr Word8-> Int -> CSize -> IO CInt
+blockingWriteRawBufferPtr loc !fd !buf !off !len
+  = throwErrnoIfMinus1Retry loc $ do
+        let start_ptr = buf `plusPtr` off
+            send_ret = c_safe_send  (fdFD fd) start_ptr (fromIntegral len) 0
+            write_ret = c_safe_write (fdFD fd) start_ptr (fromIntegral len)
+        r <- bool write_ret send_ret (fdIsSocket fd)
+        when (r == -1) c_maperrno
+        return r
+      -- We don't trust write() to give us the correct errno, and
+      -- instead do the errno conversion from GetLastError()
+      -- ourselves. The main reason is that we treat ERROR_NO_DATA
+      -- (pipe is closing) as EPIPE, whereas write() returns EINVAL
+      -- for this case. We need to detect EPIPE correctly, because it
+      -- shouldn't be reported as an error when it happens on stdout.
+      -- As for send()'s case, Winsock functions don't do errno
+      -- conversion in any case so we have to do it ourselves.
+      -- That means we're doing the errno conversion no matter if the
+      -- fd is from a socket or not.
+
+-- NOTE: "safe" versions of the read/write calls for use by the threaded RTS.
+-- These calls may block, but that's ok.
+
+asyncWriteRawBufferPtr :: String -> FD -> Ptr Word8 -> Int -> CSize -> IO CInt
+asyncWriteRawBufferPtr loc !fd !buf !off !len = do
+    (l, rc) <- asyncWrite (fromIntegral (fdFD fd)) (fdIsSocket_ fd)
+                  (fromIntegral len) (buf `plusPtr` off)
+    if l == (-1)
+      then let sock_errno = c_maperrno_func (fromIntegral rc)
+               non_sock_errno = Errno (fromIntegral rc)
+               errno = bool non_sock_errno sock_errno (fdIsSocket fd)
+           in  ioError (errnoToIOError loc errno Nothing Nothing)
+      else return (fromIntegral l)
+
+writeNonBlocking :: String -> FD -> Ptr Word8 -> Int -> CSize -> IO CInt
+writeNonBlocking loc !fd !buf !off !len
+    | threaded  = blockingWriteRawBufferPtr loc fd buf off len
+    | otherwise = asyncWriteRawBufferPtr    loc fd buf off len
+
+#endif
+
+-- | @write FD buffer offset length@ tries to write data on the given
+-- filesystem fd (cannot be a socket) up to sepcified length starting from the
+-- given offset in the buffer. The write will not block the OS thread, it may
+-- suspend the Haskell thread until write can proceed.  Returns the actual
+-- amount of data written.
+write :: FD -> Ptr Word8 -> Int -> CSize -> IO CInt
+write = writeNonBlocking "Streamly.FileSystem.FDIO"
+
+-- XXX sendAll for sockets has a similar code, we can deduplicate the two.
+-- XXX we need to check the errno to determine if the loop should continue. For
+-- example, write may return without writing all data if the process file-size
+-- limit has reached, in that case keep writing in a loop is fruitless.
+--
+-- | Keep writing in a loop until all data in the buffer has been written.
+writeAll :: FD -> Ptr Word8 -> Int -> IO ()
+writeAll fd ptr bytes = do
+    res <- write fd ptr 0 (fromIntegral bytes)
+    let res' = fromIntegral res
+    if res' < bytes
+    then writeAll fd (ptr `plusPtr` res') (bytes - res')
+    else return ()
+
+-------------------------------------------------------------------------------
+-- Vector IO
+-------------------------------------------------------------------------------
+
+-- | @write FD iovec count@ tries to write data on the given filesystem fd
+-- (cannot be a socket) from an iovec with specified number of entries.  The
+-- write will not block the OS thread, it may suspend the Haskell thread until
+-- write can proceed.  Returns the actual amount of data written.
+writev :: FD -> Ptr IOVec -> Int -> IO CInt
+writev = writevNonBlocking "Streamly.FileSystem.FDIO"
+
+-- | Keep writing an iovec in a loop until all the iovec entries are written.
+writevAll :: FD -> Ptr IOVec -> Int -> IO ()
+writevAll fd iovec count = do
+    _res <- writev fd iovec count
+    {-
+    let res' = fromIntegral res
+    totalBytes = countIOVecBytes
+    if res' < totalBytes
+     then do
+        let iovec' = createModifiedIOVec
+            count' = ...
+        writeAll fd iovec' count'
+     else return ()
+    -}
+    return ()
diff --git a/src/Streamly/FileSystem/File.hs b/src/Streamly/FileSystem/File.hs
index 9e983d0b6..3472e3088 100644
--- a/src/Streamly/FileSystem/File.hs
+++ b/src/Streamly/FileSystem/File.hs
@@ -63,7 +63,7 @@ module Streamly.FileSystem.File
     -- -- * Array Read
     -- , readArrayOf
 
-    , readArraysUpto
+    , readArraysOfUpto
     -- , readArraysOf
     , readArrays
 
@@ -183,26 +183,26 @@ appendArray file arr = SIO.withFile file AppendMode (\h -> FH.writeArray h arr)
 -- Stream of Arrays IO
 -------------------------------------------------------------------------------
 
--- | @readArraysUpto size file@ reads a stream of arrays from file @file@.
+-- | @readArraysOfUpto size file@ reads a stream of arrays from file @file@.
 -- The maximum size of a single array is specified by @size@. The actual size
 -- read may be less than or equal to @size@.
-{-# INLINABLE readArraysUpto #-}
-readArraysUpto :: (IsStream t, MonadCatch m, MonadIO m)
+{-# INLINABLE readArraysOfUpto #-}
+readArraysOfUpto :: (IsStream t, MonadCatch m, MonadIO m)
     => Int -> FilePath -> t m (Array Word8)
-readArraysUpto size file = withFile file ReadMode (FH.readArraysUpto size)
+readArraysOfUpto size file = withFile file ReadMode (FH.readArraysOfUpto size)
 
 -- XXX read 'Array a' instead of Word8
 --
 -- | @readArrays file@ reads a stream of arrays from file @file@.
 -- The maximum size of a single array is limited to @defaultChunkSize@.
 --
--- > readArrays = readArraysUpto defaultChunkSize
+-- > readArrays = readArraysOfUpto defaultChunkSize
 --
 -- @since 0.7.0
 {-# INLINE readArrays #-}
 readArrays :: (IsStream t, MonadCatch m, MonadIO m)
     => FilePath -> t m (Array Word8)
-readArrays = readArraysUpto A.defaultChunkSize
+readArrays = readArraysOfUpto A.defaultChunkSize
 
 -------------------------------------------------------------------------------
 -- Read File to Stream
@@ -219,7 +219,7 @@ readArrays = readArraysUpto A.defaultChunkSize
 --
 {-# INLINE readByChunksUpto #-}
 readByChunksUpto :: (IsStream t, MonadIO m) => Int -> Handle -> t m Word8
-readByChunksUpto chunkSize h = A.flattenArrays $ readArraysUpto chunkSize h
+readByChunksUpto chunkSize h = A.flattenArrays $ readArraysOfUpto chunkSize h
 -}
 
 -- TODO
diff --git a/src/Streamly/FileSystem/Handle.hs b/src/Streamly/FileSystem/Handle.hs
index 8adc29bd9..cf1d41292 100644
--- a/src/Streamly/FileSystem/Handle.hs
+++ b/src/Streamly/FileSystem/Handle.hs
@@ -53,13 +53,13 @@ module Streamly.FileSystem.Handle
     -- , readUtf8
     -- , readLines
     -- , readFrames
-    -- , readByChunks
+    , readByChunksUpto
 
     -- -- * Array Read
     -- , readArrayUpto
     -- , readArrayOf
 
-    , readArraysUpto
+    , readArraysOfUpto
     -- , readArraysOf
     , readArrays
 
@@ -68,7 +68,7 @@ module Streamly.FileSystem.Handle
     -- , writeUtf8
     -- , writeUtf8ByLines
     -- , writeByFrames
-    , writeByChunks
+    , writeByChunksOf
 
     -- -- * Array Write
     , writeArray
@@ -172,13 +172,13 @@ writeArray h Array{..} = withForeignPtr aStart $ \p -> hPutBuf h p aLen
 -- Stream of Arrays IO
 -------------------------------------------------------------------------------
 
--- | @readArraysUpto size h@ reads a stream of arrays from file handle @h@.
+-- | @readArraysOfUpto size h@ reads a stream of arrays from file handle @h@.
 -- The maximum size of a single array is specified by @size@. The actual size
 -- read may be less than or equal to @size@.
-{-# INLINABLE _readArraysUpto #-}
-_readArraysUpto :: (IsStream t, MonadIO m)
+{-# INLINABLE _readArraysOfUpto #-}
+_readArraysOfUpto :: (IsStream t, MonadIO m)
     => Int -> Handle -> t m (Array Word8)
-_readArraysUpto size h = go
+_readArraysOfUpto size h = go
   where
     -- XXX use cons/nil instead
     go = mkStream $ \_ yld _ stp -> do
@@ -187,9 +187,9 @@ _readArraysUpto size h = go
         then stp
         else yld arr go
 
-{-# INLINE_NORMAL readArraysUpto #-}
-readArraysUpto :: (IsStream t, MonadIO m) => Int -> Handle -> t m (Array Word8)
-readArraysUpto size h = D.fromStreamD (D.Stream step ())
+{-# INLINE_NORMAL readArraysOfUpto #-}
+readArraysOfUpto :: (IsStream t, MonadIO m) => Int -> Handle -> t m (Array Word8)
+readArraysOfUpto size h = D.fromStreamD (D.Stream step ())
   where
     {-# INLINE_LATE step #-}
     step _ _ = do
@@ -206,12 +206,12 @@ readArraysUpto size h = D.fromStreamD (D.Stream step ())
 -- 'readArrays' ignores the prevailing 'TextEncoding' and 'NewlineMode'
 -- on the 'Handle'.
 --
--- > readArrays = readArraysUpto defaultChunkSize
+-- > readArrays = readArraysOfUpto defaultChunkSize
 --
 -- @since 0.7.0
 {-# INLINE readArrays #-}
 readArrays :: (IsStream t, MonadIO m) => Handle -> t m (Array Word8)
-readArrays = readArraysUpto A.defaultChunkSize
+readArrays = readArraysOfUpto A.defaultChunkSize
 
 -------------------------------------------------------------------------------
 -- Read File to Stream
@@ -221,15 +221,13 @@ readArrays = readArraysUpto A.defaultChunkSize
 -- read requests at the same time. For serial case we can use async IO. We can
 -- also control the read throughput in mbps or IOPS.
 
-{-
 -- | @readByChunksUpto chunkSize handle@ reads a byte stream from a file
 -- handle, reads are performed in chunks of up to @chunkSize@.  The stream ends
 -- as soon as EOF is encountered.
 --
 {-# INLINE readByChunksUpto #-}
 readByChunksUpto :: (IsStream t, MonadIO m) => Int -> Handle -> t m Word8
-readByChunksUpto chunkSize h = A.flattenArrays $ readArraysUpto chunkSize h
--}
+readByChunksUpto chunkSize h = A.flattenArrays $ readArraysOfUpto chunkSize h
 
 -- TODO
 -- read :: (IsStream t, MonadIO m, Storable a) => Handle -> t m a
@@ -267,11 +265,11 @@ writeArrays h m = S.mapM_ (liftIO . writeArray h) m
 -- input elements.
 --
 -- @since 0.7.0
-{-# INLINE writeByChunks #-}
-writeByChunks :: MonadIO m => Int -> Handle -> SerialT m Word8 -> m ()
-writeByChunks n h m = writeArrays h $ A.arraysOf n m
+{-# INLINE writeByChunksOf #-}
+writeByChunksOf :: MonadIO m => Int -> Handle -> SerialT m Word8 -> m ()
+writeByChunksOf n h m = writeArrays h $ A.arraysOf n m
 
--- > write = 'writeByChunks' A.defaultChunkSize
+-- > write = 'writeByChunksOf' A.defaultChunkSize
 --
 -- | Write a byte stream to a file handle. Combines the bytes in chunks of size
 -- up to 'A.defaultChunkSize' before writing.  Note that the write behavior
@@ -280,7 +278,7 @@ writeByChunks n h m = writeArrays h $ A.arraysOf n m
 -- @since 0.7.0
 {-# INLINE write #-}
 write :: MonadIO m => Handle -> SerialT m Word8 -> m ()
-write = writeByChunks A.defaultChunkSize
+write = writeByChunksOf A.defaultChunkSize
 
 {-
 {-# INLINE write #-}
diff --git a/src/Streamly/FileSystem/IOVec.hsc b/src/Streamly/FileSystem/IOVec.hsc
new file mode 100644
index 000000000..931e36c3d
--- /dev/null
+++ b/src/Streamly/FileSystem/IOVec.hsc
@@ -0,0 +1,66 @@
+{-# LANGUAGE CPP #-}
+{-# LANGUAGE BangPatterns #-}
+{-# LANGUAGE CApiFFI #-}
+{-# LANGUAGE ScopedTypeVariables #-}
+
+-- |
+-- Module      : Streamly.FileSystem.IOVec
+-- Copyright   : (c) 2019 Composewell Technologies
+--
+-- License     : BSD3
+-- Maintainer  : harendra.kumar@gmail.com
+-- Stability   : experimental
+-- Portability : GHC
+--
+-- Low level IO routines interfacing the operating system.
+--
+
+module Streamly.FileSystem.IOVec
+    ( IOVec(..)
+    , c_writev
+    , c_safe_writev
+    )
+where
+
+import Data.Word (Word8, Word64)
+import Foreign.C.Types (CInt(..))
+import Foreign.Ptr (Ptr)
+import Foreign.Storable (Storable(..))
+import System.Posix.Types (CSsize(..))
+
+-------------------------------------------------------------------------------
+-- IOVec
+-------------------------------------------------------------------------------
+
+#if !defined(mingw32_HOST_OS)
+
+#include <sys/uio.h>
+
+data IOVec = IOVec
+  { iovBase :: {-# UNPACK #-} !(Ptr Word8)
+  , iovLen  :: {-# UNPACK #-} !Word64
+  } deriving (Eq, Show)
+
+instance Storable IOVec where
+  sizeOf _ = #{size struct iovec}
+  alignment _ = #{alignment struct iovec}
+  peek ptr = do
+      base <- #{peek struct iovec, iov_base} ptr
+      len  :: #{type size_t} <- #{peek struct iovec, iov_len}  ptr
+      return $ IOVec base len
+  poke ptr vec = do
+      let base = iovBase vec
+          len  :: #{type size_t} = iovLen  vec
+      #{poke struct iovec, iov_base} ptr base
+      #{poke struct iovec, iov_len}  ptr len
+
+foreign import capi unsafe "sys/uio.h writev"
+   c_writev :: CInt -> Ptr IOVec -> CInt -> IO CSsize
+
+foreign import capi safe "sys/uio.h writev"
+   c_safe_writev :: CInt -> Ptr IOVec -> CInt -> IO CSsize
+
+#else
+c_writev = error "writev not implemented for windows"
+c_safe_writev = error "writev not implemented for windows"
+#endif
diff --git a/src/Streamly/Mem/Array.hs b/src/Streamly/Mem/Array.hs
index 35f938b79..bd2ce5df6 100644
--- a/src/Streamly/Mem/Array.hs
+++ b/src/Streamly/Mem/Array.hs
@@ -93,27 +93,32 @@ module Streamly.Mem.Array
     -- , newArray
     , writeN
     , write
-    , fromListN
-    , fromList
+    , A.fromListN
+    , A.fromList
 
     -- Folds
-    , toArrayN
+    , A.toArrayN
     -- , toArrays
     , toArray
 
-    -- Streams
+    -- Streams of arrays
     , arraysOf
 
     -- * Elimination
     -- 'GHC.Exts.toList' from "GHC.Exts" can be used to convert an array to a
     -- list.
 
-    , read
+    , A.read
     , readRev
-    , toList
+    , A.toList
+
+    -- Streams of arrays
     , flattenArrays
     -- , flattenArraysRev
-    , spliceArrays
+    , coalesceArrays
+    , coalesceChunksOf
+    , unlinesArraysBy
+    , splitArraysOn
 
     -- * Random Access
     , length
@@ -123,8 +128,10 @@ module Streamly.Mem.Array
     {-
     , readSlice
     , readSliceRev
+    -}
 
     , writeIndex
+    {-
     , writeSlice
     , writeSliceRev
     -}
@@ -140,13 +147,14 @@ where
 import Control.Monad (when)
 import Control.Monad.IO.Class (MonadIO(..))
 import Data.Functor.Identity (Identity)
+import Data.Word (Word8)
 import Foreign.ForeignPtr (withForeignPtr)
 import Foreign.ForeignPtr.Unsafe (unsafeForeignPtrToPtr)
 import Foreign.Ptr (minusPtr, plusPtr, castPtr)
 import Foreign.Storable (Storable(..))
 import Prelude hiding (length, null, last, map, (!!), read)
 
-import Streamly.Mem.Array.Types hiding (flattenArrays, newArray)
+import Streamly.Mem.Array.Types (Array(..), length)
 import Streamly.Streams.Serial (SerialT)
 import Streamly.Streams.StreamK.Type (IsStream)
 import Streamly.Fold.Types (Fold(..))
@@ -178,7 +186,7 @@ newArray len = undefined
 writeN :: (MonadIO m, Storable a) => Int -> SerialT m a -> m (Array a)
 writeN n m = do
     if n < 0 then error "writeN: negative write count specified" else return ()
-    fromStreamDN n $ D.toStreamD m
+    A.fromStreamDN n $ D.toStreamD m
 
 -------------------------------------------------------------------------------
 -- Elimination
@@ -189,7 +197,7 @@ writeN n m = do
 -- @since 0.7.0
 {-# INLINE_EARLY readRev #-}
 readRev :: (Monad m, IsStream t, Storable a) => Array a -> t m a
-readRev = D.fromStreamD . toStreamDRev
+readRev = D.fromStreamD . A.toStreamDRev
 -- XXX add fallback to StreamK rule
 -- {-# RULES "Streamly.Array.readRev fallback to StreamK" [1]
 --     forall a. S.toStreamK (readRev a) = K.revFromArray a #-}
@@ -199,7 +207,7 @@ _null :: Storable a => Array a -> Bool
 _null arr = length arr == 0
 
 {-# INLINE _last #-}
-_last :: forall a. Storable a => Array a -> Maybe a
+_last :: (MonadIO m, Storable a) => Array a -> m (Maybe a)
 _last arr = readIndex arr (length arr - 1)
 
 -------------------------------------------------------------------------------
@@ -306,11 +314,12 @@ foldbWith level f = undefined
 --
 -- @since 0.7.0
 {-# INLINE readIndex #-}
-readIndex :: Storable a => Array a -> Int -> Maybe a
+readIndex :: (MonadIO m, Storable a) => Array a -> Int -> m (Maybe a)
 readIndex arr i =
     if i < 0 || i > length arr - 1
-    then Nothing
-    else Just $ unsafeIndex arr i
+    then return Nothing
+    else liftIO $ withForeignPtr (aStart arr) $ \p ->
+            fmap Just $ peekElemOff p i
 
 {-
 -- | @readSlice arr i count@ streams a slice of the array @arr@ starting
@@ -332,14 +341,25 @@ readSlice arr i len = undefined
 readSliceRev :: (IsStream t, Monad m, Storable a)
     => Array a -> Int -> Int -> t m a
 readSliceRev arr i len = undefined
+-}
 
 -- | /O(1)/ Write the given element at the given index in the array.
 --
 -- @since 0.7.0
 {-# INLINE writeIndex #-}
 writeIndex :: (MonadIO m, Storable a) => Array a -> Int -> a -> m ()
-writeIndex arr i a = undefined
+writeIndex arr i a = do
+    let maxIndex = length arr - 1
+    if i < 0
+    then error "writeIndex: negative array index"
+    else if i > maxIndex
+         then error $ "writeIndex: specified array index " ++ show i
+                    ++ " is beyond the maximum index " ++ show maxIndex
+         else
+            liftIO $ withForeignPtr (aStart arr) $ \p ->
+                pokeElemOff p i a
 
+{-
 -- | @writeSlice arr i count stream@ writes a stream to the array @arr@
 -- starting at index @i@ and writing up to @count@ elements in the forward
 -- direction ending at the index @i + count - 1@.
@@ -381,7 +401,7 @@ toArraysInRange low high (Fold step initial extract) =
 {-# INLINE _toArraysOf #-}
 _toArraysOf :: (MonadIO m, Storable a)
     => Int -> Fold m a (SerialT Identity (Array a))
-_toArraysOf n = FL.lchunksOf n (toArrayN n) FL.toStream
+_toArraysOf n = FL.lchunksOf n (A.toArrayN n) FL.toStream
 
 -- XXX The realloc based implementation needs to make one extra copy if we use
 -- shrinkToFit.  On the other hand, the stream of arrays implementation may
@@ -401,7 +421,7 @@ bytesToCount x n =
     let elemSize = sizeOf x
     in n + elemSize - 1 `div` elemSize
 
-{-# INLINE toArrayMinChunk #-}
+{-# INLINE_NORMAL toArrayMinChunk #-}
 toArrayMinChunk :: forall m a. (MonadIO m, Storable a)
     => Int -> Fold m a (Array a)
 -- toArrayMinChunk n = FL.mapM spliceArrays $ toArraysOf n
@@ -416,11 +436,14 @@ toArrayMinChunk elemCount = Fold step initial extract
     initial = do
         when (elemCount < 0) $ error "toArrayMinChunk: elemCount is negative"
         liftIO $ A.newArray elemCount
-    step arr@(Array _ end bound) x | end == bound = do
-        arr1 <- liftIO $ reallocDouble 1 arr
+    step arr@(Array start end bound) x | end == bound = do
+        let p = unsafeForeignPtrToPtr start
+            oldSize = end `minusPtr` p
+            newSize = max (oldSize * 2) 1
+        arr1 <- liftIO $ A.realloc newSize arr
         insertElem arr1 x
     step arr x = insertElem arr x
-    extract = liftIO . shrinkToFit
+    extract = liftIO . A.shrinkToFit
 
 -- | Fold the whole input to a single array.
 --
@@ -429,7 +452,7 @@ toArrayMinChunk elemCount = Fold step initial extract
 -- @since 0.7.0
 {-# INLINE toArray #-}
 toArray :: forall m a. (MonadIO m, Storable a) => Fold m a (Array a)
-toArray = toArrayMinChunk (bytesToCount (undefined :: a) (mkChunkSize 1024))
+toArray = toArrayMinChunk (bytesToCount (undefined :: a) (A.mkChunkSize 1024))
 
 -- | Convert a stream of arrays into a stream of their elements.
 --
@@ -449,6 +472,39 @@ flattenArrays m = D.fromStreamD $ A.flattenArrays (D.toStreamD m)
 _flattenArraysRev :: (IsStream t, MonadIO m, Storable a) => t m (Array a) -> t m a
 _flattenArraysRev m = D.fromStreamD $ A.flattenArraysRev (D.toStreamD m)
 
+-- XXX use an Array instead as separator? Or use a separate unlinesArraysBySeq
+-- API for that?
+--
+-- | Flatten a stream of arrays appending the given element after each
+-- array.
+--
+-- @since 0.7.0
+{-# INLINE unlinesArraysBy #-}
+unlinesArraysBy :: (MonadIO m, IsStream t, Storable a)
+    => a -> t m (Array a) -> t m a
+unlinesArraysBy x = D.fromStreamD . A.unlines x . D.toStreamD
+
+-- | Split a stream of arrays on a given separator byte, dropping the separator
+-- and coalescing all the arrays between two separators into a single array.
+--
+-- @since 0.7.0
+{-# INLINE splitArraysOn #-}
+splitArraysOn
+    :: (IsStream t, MonadIO m)
+    => Word8
+    -> t m (Array Word8)
+    -> t m (Array Word8)
+splitArraysOn byte s = D.fromStreamD $ A.splitOn byte $ D.toStreamD s
+
+-- | Coalesce adajcent arrays in incoming stream to form bigger arrays of a
+-- maximum specified size.
+--
+-- @since 0.7.0
+{-# INLINE coalesceChunksOf #-}
+coalesceChunksOf :: (MonadIO m, Storable a)
+    => Int -> SerialT m (Array a) -> SerialT m (Array a)
+coalesceChunksOf n xs = D.fromStreamD $ A.coalesceChunksOf n (D.toStreamD xs)
+
 -- |
 -- > arraysOf n = FL.groupsOf n (FL.toArrayN n)
 --
@@ -459,18 +515,17 @@ _flattenArraysRev m = D.fromStreamD $ A.flattenArraysRev (D.toStreamD m)
 arraysOf :: (IsStream t, MonadIO m, Storable a)
     => Int -> t m a -> t m (Array a)
 arraysOf n str =
-    D.fromStreamD $ fromStreamDArraysOf n (D.toStreamD str)
+    D.fromStreamD $ A.fromStreamDArraysOf n (D.toStreamD str)
 
 -- XXX Both of these implementations of splicing seem to perform equally well.
 -- We need to perform benchmarks over a range of sizes though.
 
-{-# INLINE _spliceArraysRealloced #-}
-_spliceArraysRealloced :: (MonadIO m, Storable a)
-    => SerialT m (Array a) -> m (Array a)
-_spliceArraysRealloced s = do
-    buffered <- P.foldr S.cons S.nil s
-    len <- S.sum (S.map length buffered)
-
+-- CAUTION! length must more than equal to lengths of all the arrays in the
+-- stream.
+{-# INLINE spliceArraysLenUnsafe #-}
+spliceArraysLenUnsafe :: (MonadIO m, Storable a)
+    => Int -> SerialT m (Array a) -> m (Array a)
+spliceArraysLenUnsafe len buffered = do
     arr <- liftIO $ A.newArray len
     end <- S.foldlM' writeArr (aEnd arr) buffered
     return $ arr {aEnd = end}
@@ -479,42 +534,36 @@ _spliceArraysRealloced s = do
 
     writeArr dst Array{..} =
         liftIO $ withForeignPtr aStart $ \src -> do
-                        let len = aEnd `minusPtr` src
-                        memcpy (castPtr dst) (castPtr src) len
-                        return $ dst `plusPtr` len
+                        let count = aEnd `minusPtr` src
+                        A.memcpy (castPtr dst) (castPtr src) count
+                        return $ dst `plusPtr` count
 
-{-# INLINE spliceArraysBuffered #-}
-spliceArraysBuffered :: forall m a. (MonadIO m, Storable a)
+{-# INLINE _spliceArraysBuffered #-}
+_spliceArraysBuffered :: (MonadIO m, Storable a)
     => SerialT m (Array a) -> m (Array a)
-spliceArraysBuffered s = do
+_spliceArraysBuffered s = do
+    buffered <- P.foldr S.cons S.nil s
+    len <- S.sum (S.map length buffered)
+    spliceArraysLenUnsafe len s
+
+{-# INLINE spliceArraysRealloced #-}
+spliceArraysRealloced :: forall m a. (MonadIO m, Storable a)
+    => SerialT m (Array a) -> m (Array a)
+spliceArraysRealloced s = do
     idst <- liftIO $ A.newArray (bytesToCount (undefined :: a)
-                                (mkChunkSizeKB 4))
-    arr <- S.foldlM' appendArr idst s
-    liftIO $ shrinkToFit arr
+                                (A.mkChunkSizeKB 4))
 
-    where
-
-    appendArr dst@(Array _ end bound) src  = liftIO $ do
-        let srcLen = aEnd src `minusPtr` unsafeForeignPtrToPtr (aStart src)
-        dst1 <-
-            if end `plusPtr` srcLen >= bound
-            then reallocDouble srcLen dst
-            else return dst
-
-        withForeignPtr (aStart dst1) $ \_ -> do
-            withForeignPtr (aStart src) $ \psrc -> do
-                let pdst = aEnd dst1
-                memcpy (castPtr pdst) (castPtr psrc) srcLen
-                return $ dst1 { aEnd = pdst `plusPtr` srcLen }
+    arr <- S.foldlM' A.spliceWithDoubling idst s
+    liftIO $ A.shrinkToFit arr
 
 -- | Given a stream of arrays, splice them all together to generate a single
 -- array. The stream must be /finite/.
 --
 -- @since 0.7.0
-{-# INLINE spliceArrays #-}
-spliceArrays :: (MonadIO m, Storable a) => SerialT m (Array a) -> m (Array a)
-spliceArrays = spliceArraysBuffered
--- spliceArrays = _spliceArraysRealloced
+{-# INLINE coalesceArrays #-}
+coalesceArrays :: (MonadIO m, Storable a) => SerialT m (Array a) -> m (Array a)
+coalesceArrays = spliceArraysRealloced
+-- spliceArrays = _spliceArraysBuffered
 
 -- | Create an 'Array' from a stream. This is useful when we want to create a
 -- single array from a stream of unknown size. 'writeN' is at least twice
@@ -548,18 +597,18 @@ write = FL.foldl' toArray
 {-# INLINE transformWith #-}
 transformWith :: (MonadIO m, Storable a, Storable b)
     => (SerialT m a -> SerialT m b) -> Array a -> m (Array b)
-transformWith f arr = FL.foldl' (toArrayMinChunk (length arr)) $ f (read arr)
+transformWith f arr = FL.foldl' (toArrayMinChunk (length arr)) $ f (A.read arr)
 
 -- | Fold an array using a 'Fold'.
 --
 -- @since 0.7.0
 {-# INLINE foldArray #-}
 foldArray :: (MonadIO m, Storable a) => Fold m a b -> Array a -> m b
-foldArray f arr = FL.foldl' f (read arr)
+foldArray f arr = FL.foldl' f (A.read arr)
 
 -- | Fold an array using a stream fold operation.
 --
 -- @since 0.7.0
 {-# INLINE foldWith #-}
 foldWith :: (MonadIO m, Storable a) => (SerialT m a -> m b) -> Array a -> m b
-foldWith f arr = f (read arr)
+foldWith f arr = f (A.read arr)
diff --git a/src/Streamly/Mem/Array/Types.hs b/src/Streamly/Mem/Array/Types.hs
index f1c9efae0..d1618280e 100644
--- a/src/Streamly/Mem/Array/Types.hs
+++ b/src/Streamly/Mem/Array/Types.hs
@@ -23,29 +23,34 @@ module Streamly.Mem.Array.Types
       Array (..)
 
     -- * Construction
-    , unsafeInlineIO
     , withNewArray
     , newArray
     , unsafeSnoc
     , snoc
-    , shrinkToFit
-    , memcpy
-    , memcmp
+    , spliceWithDoubling
 
     , fromList
     , fromListN
     , fromStreamDN
     -- , fromStreamD
+
+    -- * Streams of arrays
     , fromStreamDArraysOf
     , flattenArrays
     , flattenArraysRev
+    , coalesceChunksOf
+    , groupIOVecsOf
+    , splitOn
 
     -- * Elimination
     , unsafeIndexIO
     , unsafeIndex
     , length
+    , byteLength
+    , byteCapacity
     , foldl'
     , foldr
+    , splitAt
 
     , toStreamD
     , toStreamDRev
@@ -59,7 +64,11 @@ module Streamly.Mem.Array.Types
     , defaultChunkSize
     , mkChunkSize
     , mkChunkSizeKB
-    , reallocDouble
+    , unsafeInlineIO
+    , realloc
+    , shrinkToFit
+    , memcpy
+    , memcmp
 
     , unlines
     )
@@ -74,11 +83,11 @@ import Data.Word (Word8)
 import Foreign.C.String (CString)
 import Foreign.C.Types (CSize(..), CInt(..))
 import Foreign.ForeignPtr
-       (ForeignPtr, withForeignPtr, touchForeignPtr)
+       (ForeignPtr, withForeignPtr, touchForeignPtr, plusForeignPtr)
 import Foreign.ForeignPtr.Unsafe (unsafeForeignPtrToPtr)
-import Foreign.Ptr (plusPtr, minusPtr, castPtr)
+import Foreign.Ptr (plusPtr, minusPtr, castPtr, nullPtr)
 import Foreign.Storable (Storable(..))
-import Prelude hiding (length, foldr, read, unlines, unwords)
+import Prelude hiding (length, foldr, read, unlines, splitAt)
 import Text.Read (readPrec, readListPrec, readListPrecDefault)
 
 import GHC.Base (Addr#, realWorld#)
@@ -89,6 +98,7 @@ import GHC.Ptr (Ptr(..))
 
 import Streamly.Fold.Types (Fold(..))
 import Streamly.SVar (adaptState)
+import Streamly.FileSystem.FDIO (IOVec(..))
 
 import qualified Streamly.Streams.StreamD.Type as D
 import qualified Streamly.Streams.StreamK as K
@@ -165,6 +175,9 @@ foreign import ccall unsafe "string.h memcpy" c_memcpy
 foreign import ccall unsafe "string.h strlen" c_strlen
     :: CString -> IO CSize
 
+foreign import ccall unsafe "string.h memchr" c_memchr
+    :: Ptr Word8 -> Word8 -> CSize -> IO (Ptr Word8)
+
 -- XXX we are converting Int to CSize
 memcpy :: Ptr Word8 -> Ptr Word8 -> Int -> IO ()
 memcpy dst src len = c_memcpy dst src (fromIntegral len) >> return ()
@@ -222,14 +235,14 @@ withNewArray count f = do
 unsafeSnoc :: forall a. Storable a => Array a -> a -> IO (Array a)
 unsafeSnoc arr@Array{..} x = do
     when (aEnd == aBound) $
-        error "BUG: unsafeAppend: writing beyond array bounds"
+        error "BUG: unsafeSnoc: writing beyond array bounds"
     poke aEnd x
     touchForeignPtr aStart
     return $ arr {aEnd = aEnd `plusPtr` (sizeOf (undefined :: a))}
 
 {-# INLINE snoc #-}
-snoc :: forall a. Storable a => Array a -> a -> Array a
-snoc arr@Array {..} x = unsafePerformIO $
+snoc :: forall a. Storable a => Array a -> a -> IO (Array a)
+snoc arr@Array {..} x = do
     if (aEnd == aBound)
     then do
         let oldStart = unsafeForeignPtrToPtr aStart
@@ -250,37 +263,14 @@ snoc arr@Array {..} x = unsafePerformIO $
         touchForeignPtr aStart
         return $ arr {aEnd = aEnd `plusPtr` (sizeOf (undefined :: a))}
 
-
--- | Remove the free space from an Array.
-shrinkToFit :: forall a. Storable a => Array a -> IO (Array a)
-shrinkToFit arr@Array{..} = do
-    assert (aEnd <= aBound) (return ())
-    if aEnd /= aBound
-    then do
-        let oldStart = unsafeForeignPtrToPtr aStart
-        let size = aEnd `minusPtr` oldStart
-        newPtr <- mallocPlainForeignPtrAlignedBytes
-                    size (alignment (undefined :: a))
-        withForeignPtr newPtr $ \pNew -> do
-            memcpy (castPtr pNew) (castPtr oldStart) size
-            touchForeignPtr aStart
-            let end = pNew `plusPtr` size
-            return $ Array
-                { aStart = newPtr
-                , aEnd   = end
-                , aBound = end
-                }
-    else return arr
-
--- | Expand the free space in the array doubling the size with a minimum
--- exapnsion of the specified amount of bytes.
-{-# NOINLINE reallocDouble #-}
-reallocDouble :: forall a. Storable a => Int -> Array a -> IO (Array a)
-reallocDouble minIncrease Array{..} = do
+-- | Reallocate the array to the specified size in bytes. If the size is less
+-- than the original array the array gets truncated.
+{-# NOINLINE realloc #-}
+realloc :: forall a. Storable a => Int -> Array a -> IO (Array a)
+realloc newSize Array{..} = do
     assert (aEnd <= aBound) (return ())
     let oldStart = unsafeForeignPtrToPtr aStart
     let size = aEnd `minusPtr` oldStart
-        newSize = max (size * 2) (size + minIncrease)
     newPtr <- mallocPlainForeignPtrAlignedBytes
                 newSize (alignment (undefined :: a))
     withForeignPtr newPtr $ \pNew -> do
@@ -292,6 +282,17 @@ reallocDouble minIncrease Array{..} = do
             , aBound = pNew `plusPtr` newSize
             }
 
+-- | Remove the free space from an Array.
+shrinkToFit :: forall a. Storable a => Array a -> IO (Array a)
+shrinkToFit arr@Array{..} = do
+    assert (aEnd <= aBound) (return ())
+    if aEnd /= aBound
+    then do
+        let oldStart = unsafeForeignPtrToPtr aStart
+        let size = aEnd `minusPtr` oldStart
+        realloc size arr
+    else return arr
+
 -- XXX when converting an array of Word8 from a literal string we can simply
 -- refer to the literal string. Is it possible to write rules such that
 -- fromList Word8 can be rewritten so that GHC does not first convert the
@@ -340,15 +341,30 @@ unsafeIndexIO Array {..} i =
 unsafeIndex :: forall a. Storable a => Array a -> Int -> a
 unsafeIndex arr i = let !r = unsafeInlineIO $ unsafeIndexIO arr i in r
 
--- | /O(1)/ Get the length of the array.
+-- | /O(1)/ Get the byte length of the array.
+--
+-- @since 0.7.0
+{-# INLINE byteLength #-}
+byteLength :: Array a -> Int
+byteLength Array{..} =
+    let p = unsafeForeignPtrToPtr aStart
+        len = aEnd `minusPtr` p
+    in assert (len >= 0) len
+
+-- | /O(1)/ Get the length of the array i.e. the number of elements in the
+-- array.
 --
 -- @since 0.7.0
 {-# INLINE length #-}
 length :: forall a. Storable a => Array a -> Int
-length Array{..} =
+length arr = byteLength arr `div` sizeOf (undefined :: a)
+
+{-# INLINE byteCapacity #-}
+byteCapacity :: Array a -> Int
+byteCapacity Array{..} =
     let p = unsafeForeignPtrToPtr aStart
-        aLen = aEnd `minusPtr` p
-    in assert (aLen >= 0) (aLen `div` sizeOf (undefined :: a))
+        len = aBound `minusPtr` p
+    in assert (len >= 0) len
 
 {-# INLINE_NORMAL toStreamD #-}
 toStreamD :: forall m a. (Monad m, Storable a) => Array a -> D.Stream m a
@@ -729,8 +745,9 @@ defaultChunkSize :: Int
 defaultChunkSize = mkChunkSizeKB 32
 
 {-# INLINE_NORMAL unlines #-}
-unlines :: MonadIO m => D.Stream m (Array Char) -> D.Stream m Char
-unlines (D.Stream step state) = D.Stream step' (OuterLoop state)
+unlines :: forall m a. (MonadIO m, Storable a)
+    => a -> D.Stream m (Array a) -> D.Stream m a
+unlines sep (D.Stream step state) = D.Stream step' (OuterLoop state)
     where
     {-# INLINE_LATE step' #-}
     step' gst (OuterLoop st) = do
@@ -743,7 +760,7 @@ unlines (D.Stream step state) = D.Stream step' (OuterLoop state)
             D.Stop -> D.Stop
 
     step' _ (InnerLoop st _ p end) | p == end =
-        return $ D.Yield '\n' $ OuterLoop st
+        return $ D.Yield sep $ OuterLoop st
 
     step' _ (InnerLoop st startf p end) = do
         x <- liftIO $ do
@@ -751,4 +768,272 @@ unlines (D.Stream step state) = D.Stream step' (OuterLoop state)
                     touchForeignPtr startf
                     return r
         return $ D.Yield x (InnerLoop st startf
-                            (p `plusPtr` (sizeOf (undefined :: Char))) end)
+                            (p `plusPtr` (sizeOf (undefined :: a))) end)
+
+{-# INLINE spliceTwo #-}
+spliceTwo :: (MonadIO m, Storable a) => Array a -> Array a -> m (Array a)
+spliceTwo arr1 arr2 = do
+    let src1 = unsafeForeignPtrToPtr (aStart arr1)
+        src2 = unsafeForeignPtrToPtr (aStart arr2)
+        len1 = aEnd arr1 `minusPtr` src1
+        len2 = aEnd arr2 `minusPtr` src2
+
+    arr <- liftIO $ newArray (len1 + len2)
+    let dst = unsafeForeignPtrToPtr (aStart arr)
+
+    liftIO $ do
+        memcpy (castPtr dst) (castPtr src1) len1
+        touchForeignPtr (aStart arr1)
+        memcpy (castPtr (dst `plusPtr` len1)) (castPtr src2) len2
+        touchForeignPtr (aStart arr2)
+    return arr { aEnd = dst `plusPtr` (len1 + len2) }
+
+-- Splice a new array into a pre-reserved array. The user must ensure that
+-- there is enough space in the array.
+{-# INLINE spliceWith #-}
+spliceWith :: (MonadIO m) => Array a -> Array a -> m (Array a)
+spliceWith dst@(Array _ end bound) src  = liftIO $ do
+    let srcLen = byteLength src
+    if end `plusPtr` srcLen > bound
+    then error "Bug: spliceIntoUnsafe: Not enough space in the target array"
+    else
+        withForeignPtr (aStart dst) $ \_ -> do
+            withForeignPtr (aStart src) $ \psrc -> do
+                let pdst = aEnd dst
+                memcpy (castPtr pdst) (castPtr psrc) srcLen
+                return $ dst { aEnd = pdst `plusPtr` srcLen }
+
+-- Splice a new array into a preallocated array, doubling the space if there is
+-- no space in the target array.
+{-# INLINE spliceWithDoubling #-}
+spliceWithDoubling :: (MonadIO m, Storable a)
+    => Array a -> Array a -> m (Array a)
+spliceWithDoubling dst@(Array start end bound) src  = do
+    assert (end <= bound) (return ())
+    let srcLen = aEnd src `minusPtr` unsafeForeignPtrToPtr (aStart src)
+
+    dst1 <-
+        if end `plusPtr` srcLen >= bound
+        then do
+            let oldStart = unsafeForeignPtrToPtr start
+                oldSize = end `minusPtr` oldStart
+                newSize = max (oldSize * 2) (oldSize + srcLen)
+            liftIO $ realloc newSize dst
+        else return dst
+    spliceWith dst1 src
+
+data SpliceState s arr
+    = SpliceInitial s
+    | SpliceBuffering s arr
+    | SpliceYielding arr (SpliceState s arr)
+    | SpliceFinish
+
+-- XXX can use general grouping combinators to achieve this?
+-- | Coalesce adajcent arrays in incoming stream to form bigger arrays of a
+-- maximum specified size.
+--
+-- @since 0.7.0
+{-# INLINE_NORMAL coalesceChunksOf #-}
+coalesceChunksOf :: (MonadIO m, Storable a)
+    => Int -> D.Stream m (Array a) -> D.Stream m (Array a)
+coalesceChunksOf n (D.Stream step state) = D.Stream step' (SpliceInitial state)
+
+    where
+
+    {-# INLINE_LATE step' #-}
+    step' gst (SpliceInitial st) = do
+        r <- step gst st
+        case r of
+            D.Yield arr s -> return $
+                let len = byteLength arr
+                 in if len >= n
+                    then D.Skip (SpliceYielding arr (SpliceInitial s))
+                    else D.Skip (SpliceBuffering s arr)
+            D.Skip s -> return $ D.Skip (SpliceInitial s)
+            D.Stop -> return $ D.Stop
+
+    step' gst (SpliceBuffering st buf) = do
+        r <- step gst st
+        case r of
+            D.Yield arr s -> do
+                let len = byteLength buf + byteLength arr
+                if len > n
+                then return $ D.Skip (SpliceYielding buf (SpliceBuffering s arr))
+                else do
+                    buf' <- if byteCapacity buf < n
+                            then liftIO $ realloc n buf
+                            else return buf
+                    buf'' <- spliceWith buf' arr
+                    return $ D.Skip (SpliceBuffering s buf'')
+            D.Skip s -> return $ D.Skip (SpliceBuffering s buf)
+            D.Stop -> return $ D.Skip (SpliceYielding buf SpliceFinish)
+
+    step' _ SpliceFinish = return D.Stop
+
+    step' _ (SpliceYielding arr next) = return $ D.Yield arr next
+
+data GatherState s arr
+    = GatherInitial s
+    | GatherBuffering s arr Int
+    | GatherYielding arr (GatherState s arr)
+    | GatherFinish
+
+-- | @groupIOVecsOf maxBytes maxEntries@ groups arrays in the incoming stream
+-- to create a stream of 'IOVec' arrays with a maximum of @maxBytes@ bytes in
+-- each array and a maximum of @maxEntries@ entries in each array.
+--
+-- @since 0.7.0
+{-# INLINE_NORMAL groupIOVecsOf #-}
+groupIOVecsOf :: MonadIO m
+    => Int -> Int -> D.Stream m (Array a) -> D.Stream m (Array IOVec)
+groupIOVecsOf n maxIOVLen (D.Stream step state) =
+    D.Stream step' (GatherInitial state)
+
+    where
+
+    {-# INLINE_LATE step' #-}
+    step' gst (GatherInitial st) = do
+        r <- step (adaptState gst) st
+        case r of
+            D.Yield arr s -> do
+                let p = unsafeForeignPtrToPtr (aStart arr)
+                    len = byteLength arr
+                iov <- liftIO $ newArray maxIOVLen
+                iov' <- liftIO $ unsafeSnoc iov (IOVec (castPtr p)
+                                                (fromIntegral len))
+                if len >= n
+                then return $ D.Skip (GatherYielding iov' (GatherInitial s))
+                else return $ D.Skip (GatherBuffering s iov' len)
+            D.Skip s -> return $ D.Skip (GatherInitial s)
+            D.Stop -> return $ D.Stop
+
+    step' gst (GatherBuffering st iov len) = do
+        r <- step (adaptState gst) st
+        case r of
+            D.Yield arr s -> do
+                let p = unsafeForeignPtrToPtr (aStart arr)
+                    alen = byteLength arr
+                    len' = len + alen
+                if len' > n || length iov >= maxIOVLen
+                then do
+                    iov' <- liftIO $ newArray maxIOVLen
+                    iov'' <- liftIO $ unsafeSnoc iov' (IOVec (castPtr p)
+                                                      (fromIntegral alen))
+                    return $ D.Skip (GatherYielding iov
+                                        (GatherBuffering s iov'' alen))
+                else do
+                    iov' <- liftIO $ unsafeSnoc iov (IOVec (castPtr p)
+                                                    (fromIntegral alen))
+                    return $ D.Skip (GatherBuffering s iov' len')
+            D.Skip s -> return $ D.Skip (GatherBuffering s iov len)
+            D.Stop -> return $ D.Skip (GatherYielding iov GatherFinish)
+
+    step' _ GatherFinish = return D.Stop
+
+    step' _ (GatherYielding iov next) = return $ D.Yield iov next
+
+-- | Create two slices of an array without copying the original array. The
+-- specified index @i@ is the first index of the second slice.
+--
+-- @since 0.7.0
+splitAt :: forall a. Storable a => Int -> Array a -> (Array a, Array a)
+splitAt i arr@Array{..} =
+    let maxIndex = length arr - 1
+    in  if i < 0
+        then error "sliceAt: negative array index"
+        else if i > maxIndex
+             then error $ "sliceAt: specified array index " ++ show i
+                        ++ " is beyond the maximum index " ++ show maxIndex
+             else let off = i * sizeOf (undefined :: a)
+                      p = unsafeForeignPtrToPtr aStart `plusPtr` off
+                in ( Array
+                  { aStart = aStart
+                  , aEnd = p
+                  , aBound = p
+                  }
+                , Array
+                  { aStart = aStart `plusForeignPtr` off
+                  , aEnd = aEnd
+                  , aBound = aBound
+                  }
+                )
+
+-- Drops the separator byte
+{-# INLINE breakOn #-}
+breakOn :: MonadIO m
+    => Word8 -> Array Word8 -> m (Array Word8, Maybe (Array Word8))
+breakOn sep arr@Array{..} = liftIO $ do
+    let p = unsafeForeignPtrToPtr aStart
+    loc <- c_memchr p sep (fromIntegral $ aEnd `minusPtr` p)
+    return $
+        if loc == nullPtr
+        then (arr, Nothing)
+        else
+            ( Array
+                { aStart = aStart
+                , aEnd = loc
+                , aBound = loc
+                }
+            , Just $ Array
+                    { aStart = aStart `plusForeignPtr` (loc `minusPtr` p + 1)
+                    , aEnd = aEnd
+                    , aBound = aBound
+                    }
+            )
+
+data SplitState s arr
+    = Initial s
+    | Buffering s arr
+    | Splitting s arr
+    | Yielding arr (SplitState s arr)
+    | Finishing
+
+-- | Split a stream of arrays on a given separator byte, dropping the separator
+-- and coalescing all the arrays between two separators into a single array.
+--
+-- @since 0.7.0
+{-# INLINE_NORMAL splitOn #-}
+splitOn
+    :: MonadIO m
+    => Word8
+    -> D.Stream m (Array Word8)
+    -> D.Stream m (Array Word8)
+splitOn byte (D.Stream step state) = D.Stream step' (Initial state)
+
+    where
+
+    {-# INLINE_LATE step' #-}
+    step' gst (Initial st) = do
+        r <- step gst st
+        case r of
+            D.Yield arr s -> do
+                (arr1, marr2) <- breakOn byte arr
+                return $ case marr2 of
+                    Nothing   -> D.Skip (Buffering s arr1)
+                    Just arr2 -> D.Skip (Yielding arr1 (Splitting s arr2))
+            D.Skip s -> return $ D.Skip (Initial s)
+            D.Stop -> return $ D.Stop
+
+    step' gst (Buffering st buf) = do
+        r <- step gst st
+        case r of
+            D.Yield arr s -> do
+                (arr1, marr2) <- breakOn byte arr
+                buf' <- spliceTwo buf arr1
+                return $ case marr2 of
+                    Nothing -> D.Skip (Buffering s buf')
+                    Just x -> D.Skip (Yielding buf' (Splitting s x))
+            D.Skip s -> return $ D.Skip (Buffering s buf)
+            D.Stop -> return $
+                if byteLength buf == 0
+                then D.Stop
+                else D.Skip (Yielding buf Finishing)
+
+    step' _ (Splitting st buf) = do
+        (arr1, marr2) <- breakOn byte buf
+        return $ case marr2 of
+                Nothing -> D.Skip $ Buffering st arr1
+                Just arr2 -> D.Skip $ Yielding arr1 (Splitting st arr2)
+
+    step' _ (Yielding arr next) = return $ D.Yield arr next
+    step' _ Finishing = return $ D.Stop
diff --git a/src/Streamly/String.hs b/src/Streamly/String.hs
index 92a072c22..5ca9d2c66 100644
--- a/src/Streamly/String.hs
+++ b/src/Streamly/String.hs
@@ -187,7 +187,7 @@ words = FL.wordsBy isSpace toArray
 
 {-# INLINE unlines #-}
 unlines :: (MonadIO m, IsStream t) => t m (Array Char) -> t m Char
-unlines = D.fromStreamD . A.unlines . D.toStreamD
+unlines = D.fromStreamD . A.unlines '\n' . D.toStreamD
 
 {-# INLINE unwords #-}
 unwords :: (MonadAsync m, IsStream t) => t m (Array Char) -> t m Char
diff --git a/streamly.cabal b/streamly.cabal
index 7e4081f67..490bb818e 100644
--- a/streamly.cabal
+++ b/streamly.cabal
@@ -207,6 +207,9 @@ library
                      , Streamly.Sink
                      , Streamly.Pipe.Types
 
+                     , Streamly.FileSystem.IOVec
+                     , Streamly.FileSystem.FDIO
+
     exposed-modules:   Streamly.Prelude
                      , Streamly.Time
                      , Streamly
@@ -217,8 +220,9 @@ library
 
                     -- IO devices
                      , Streamly.Mem.Array
-                     , Streamly.FileSystem.File
+                     , Streamly.FileSystem.FD
                      , Streamly.FileSystem.Handle
+                     , Streamly.FileSystem.File
                      , Streamly.Network.Socket
                      , Streamly.Network.Server
                      , Streamly.Network.Client