SerializerMethodField - dict with known keys

[jmaslanka]

Hey, let's assume we have this field in a serializer:

field = SerializerMethodField()

def get_field(self, obj):
    return {
        "type": "some_type",
        "data": OtherSerializer(obj.objects, many=True).data,
        ...
    }

Is there a way to represent it as:

{
    "type": "some_type",
    "data": {
        <serializer data>
    },
    ....
}

I've tried:

• using typing.TypedDict as type hint to get_field - results in "Unknown type hint"
• @extend_schema_field accepts only either serializer or OpenApiTypes. I don't see a way to edit additionalProperties for OpenApiTypes.OBJECT.

Seems like a common problem to me. Any ideas?

[tfranzel]

hi @jmaslanka, indeed that still incomplete. better type hint resolution is still a todo. it currently only support Optional, basic types, and OpenApiTypes.

[jmaslanka]

Hey @tfranzel , thanks for the response. I ended up creating simple serializer, like:

class FieldData(serializers.Serializer):
    type = CharField()
    data = OtherSerializer(many=True)

and linked it using extend_schema_field.

[tfranzel]

@jmaslanka good stuff! luckily there multiple ways to achieve this. better type hint parsing is definitely on my list.

[tfranzel]

fyi: you could have also done this:

from drf_spectacular.utils import extend_schema_field, inline_serializer

@extend_schema_field(inline_serializer(
    name='FieldData',
    fields={
        "type": CharField(),
        "data": OtherSerializer(many=True),
    }
))
def get_field(self, obj):
     pass

also we improved type hint functionality including basic TypedDict. however mixing type defintions with Serializers and Serializer fields may look ok, but are probably wrong according to mypy. so my above example is probably the "cleanest" solution. TypedDict with arbitrarily mixed basic types should work now.