Only supports reading, not writing, since TOML is lossy. For example, a simple a.b.c = 1
line could be written in a number of ways:
a.b.c = 1
a.b = { c = 1 }
a = { b.c = 1 }
[a]
b.c = 1
b = { c = 1 }
[a.b]
c = 1
Since reading/writing isn’t an idempotent operation, this library won’t even pretend to provide DecodeTOML
/EncodeTOML
typeclasses that imply that they’re inverses of each other.
Hopefully some other toml-writer
library may come along to make it easy to specify how to format your data in TOML (e.g. a combinator for table
vs inlineTable
), or you could use tomland
.
This library defines DecodeTOML
with an opaque Decoder a
as opposed to a Value -> DecodeM a
function, like aeson
does. In my opinion, this makes the common case of decoding config files much more straightforward, especially around nested fields, which are much more common in TOML than JSON. e.g.
-- aeson-like
instance DecodeTOML MyConfig where
decodeTOML :: Value -> DecodeM MyConfig
decodeTOML = withObject "MyConfig" $ \o ->
MyConfig
<$> o .: "field1"
<*> (o .: "field2" >>= (.: "field3"))
-- with toml-parser
instance DecodeTOML MyConfig where
tomlDecoder :: Decoder MyConfig
tomlDecoder =
MyConfig
<$> getField "field1"
<*> getFields ["field2", "field3"]
It also makes it easy to define ad-hoc decoders:
instance DecodeTOML MyConfig where
tomlDecoder = ...
alternativeDecoder :: Decoder MyConfig
alternativeDecoder = ...
-- uses tomlDecoder
decode "a = 1"
-- uses explicit decoder
decodeWith alternativeDecoder "a = 1"
As a bonus, it also makes for a less point-free interface when defining a decoder based on another decoder, which is kinda cool:
-- aeson-like
instance DecodeTOML MyString where
decodeTOML = fmap toMyString . decodeTOML
-- with toml-parser
instance DecodeTOML MyString where
tomlDecoder = toMyString <$> tomlDecoder
Ultimately, Decoder
is just a newtype around Value -> DecodeM a
, so we could always go back to it. Originally, I wanted to do something like jordan
, where this interface is required due to the way it parses and deserializes at the same time, but this isn’t possible with TOML due to the way TOML needs to be normalized.