Codec makes it simple to write composable bidirectional serializers with a consistent interface.

Just define your data type normally:

data RecordB = RecordB
  { recordBString :: String
  , recordBDouble :: Double
  } deriving (Eq, Ord, Show)

and then associate each field with a codec using the =. operator:

recordBObjCodec :: JSONCodec RecordB
recordBObjCodec = asObject "RecordB" $
  RecordB
    <$> recordBString =. field "string"
    <*> recordBDouble =. field "double"

That's it! If you want, you can now define ToJSON and FromJSON instances, or just use it directly:

instance ToJSON RecordB where
  toJSON = toJSONCodec recordBObjCodec
  toEncoding = toEncodingCodec recordBObjCodec
instance FromJSON RecordB where
  parseJSON = parseJSONCodec recordBObjCodec

Support can be added for almost any serialization library, but aeson and binary support are included.

JSON example:

data RecordA = RecordA
  { recordAInt :: Int
  , recordANestedObj :: RecordB
  , recordANestedArr :: RecordB
  , recordANestedObjs :: [ RecordB ]
  } deriving (Eq, Ord, Show)
data RecordB = RecordB
  { recordBString :: String
  , recordBDouble :: Double
  } deriving (Eq, Ord, Show)
recordACodec :: JSONCodec RecordA
recordACodec = asObject "RecordA" $
  RecordA
    <$> recordAInt =. field "int"
    <*> recordANestedObj =. field' "nestedObj" recordBObjCodec
    <*> recordANestedArr =. field' "nestedArr" recordBArrCodec
    <*> recordANestedObjs =. field' "nestedObjs" (arrayOf' id id recordBObjCodec)
recordBObjCodec :: JSONCodec RecordB
recordBObjCodec = asObject "RecordB" $
  RecordB
    <$> recordBString =. field "string"
    <*> recordBDouble =. field "double"
-- serialize to array elements
recordBArrCodec :: JSONCodec RecordB
recordBArrCodec = asArray "RecordB" $
  RecordB
    <$> recordBString =. element
    <*> recordBDouble =. element

Binary example:

data RecordA = RecordA
  { recordAInt64 :: Int64
  , recordAWord8 :: Word8
  , recordANestedB :: RecordB
  } deriving (Eq, Ord, Show)
data RecordB = RecordB
  { recordBWord16 :: Word16
  , recordBByteString64 :: BS.ByteString
  } deriving (Eq, Ord, Show)
recordACodec :: BinaryCodec RecordA
recordACodec =
  RecordA
    <$> recordAInt64 =. int64le
    <*> recordAWord8 =. word8
    <*> recordANestedB =. recordBCodec
recordBCodec :: BinaryCodec RecordB
recordBCodec =
  RecordB
    <$> recordBWord16 =. word16host
    <*> recordBByteString64 =. byteString 64

A Codec is just a combination of a deserializer r a, and a serializer c -> w a.

data CodecFor r w c a = Codec
  { codecIn :: r a
  , codecOut :: c -> w a
  }
type Codec r w a = CodecFor r w a a

With binary for example, r is Get and w is PutM. The reason we have an extra parameter c is so that we can associate a Codec with a particular field using the =. operator:

(=.) :: (c' -> c) -> CodecFor r w c a -> CodecFor r w c' a

Codec is an instance of Functor, Applicative, Monad and Profunctor. You can serialize in any order you like, regardless of field order in the data type:

recordBCodecFlipped :: BinaryCodec RecordB
recordBCodecFlipped = do
  bs64 <- recordBByteString64 =. byteString 64
  RecordB
    <$> recordBWord16 =. word16host
    <*> pure bs64

Contributors

=. operator and Profunctor approach thanks to Xia Li-yao