Switch binary package

[singpolyma]

This package currently uses binary-strict, an older package that simply cloned binary because the error handling in binary was abysmal at the time. The binary package has improved considerably, and I would recommend moving to it.

The cereal package is also another (more actively maintained) option, but I believe things will be moving back to binary in general, now that the package is so much better.

[blanu]

I use binary-strict for the Data.Binary.Strict.BitGet and BitPut monads for parsing bit-aligned data. The binary and cereal packages seems to only support byte-aligned data. Parsing of bit-aligned data is required for Huffman encoding.

[blanu]

Strangely, there is a Data.Binary.BitPut, but no Data.Binary.BitGet?

[singpolyma]

Looking at the code, it seems the whole package dependency is only used for implementing bitpack and bitunpack, which are more succinctly (and without any dependency) written as:

bitpack :: [Bool] -> B.ByteString
bitpack = B.pack . map packByte . takeWhile (not . null) . unfoldr (Just . splitAt 8)
    where
    packByte = foldl' (\i b -> (i `shiftL` 1) .|. (fromIntegral $ fromEnum b)) 0

bitunpack :: S.ByteString -> Either String [Bool]
bitunpack = Right . concatMap (\byte -> map (testBit byte) [7,6..0]) . S.unpack

I can prepare a pull request for these versions (and dropping the binary-strict) dependency, if this meets with your approval.

You will also note that bitunpack never fails, so the type signature could be changed to reflect that.

[singpolyma]

The bitpack dependency also happens to be used in Dust.Model.Packet, but that can be easily replaced with calls to cereal (which is already a dependency anyway). I would include such a change in any pull request.

[blanu]

Sounds good!

[blanu]

Done!