Add support for python enum.Enum

[maroshmka]

Do not raise WrongTypeError when value is of type Enum. E.g. it can be Enum of ints, so first try to convert the value of int to Enum type.

For further details see tests.

[konradhalas]

Hi @maroshmka, thank you for your PR :)

In this case you should use cast parameter:

class MyEnum(enum.Enum):
    a: int = 1
    b: int = 3

@dataclass
class X:
    e: MyEnum = MyEnum.a

x = X()

result = from_dict(X, {"e": 1}, Config(cast=["e"]))

assert result == x

[maroshmka]

thanks, i'll try to use it.

It's just not very scalable tho. When you have bunch of dataclasses in your project and you add a Enum to random one, you need to add cast to every usage of from_dict with that class :/

And what if I have enum inside inner dataclass? Can cast do it?

@dataclass
class X:
    e: MyEnum = MyEnum.a

@dataclass
class Y:
    x: X

y = Y()

# can I do something like this ???
result = from_dict(Y, {"x": {"e": 1}}, Config(cast=["x.e"]))

assert result == x

If not I may implement that :)

[konradhalas]

Agree, this is why I'm thinking about two possible solutions:

• add cast_all parameter,
• get rid of cast at all - there are more mature libs (like DRF or Marshmallow) which are focused on data validation/serialization. dacite should be used in conjunction with such lib and this lib should generate proper values (enums in your case) which user should pass to dacite to build data class.

[maroshmka]

That seems like a bigger issue. But I think theres a difference - mostly when you want some special deserialization, it is because of your own created type. Enum is not own type, its python (not primitive tho).

• I would add cast_all, to try to deserialize all the attributes as they were defined in dataclass 👍
• add serializer arg, which can be drf.Serializer type to use for deserialization -> imho this is not needed, at least now.

btw: cast=["x.e"] dot notation works (if someone would be asking), maybe it would be nice to add it to doc?

Thank you for discussion 🙂