SerializerMethodField method does not get called

[B4rtware]

I am trying to understand how I would implement an additional field (token) inside the serializer class from rest_framework. This code snippet works in native django rest framework but unfortunately not in conjunction with graphene_django.

Using this serializer:

from rest_framework import serializers
from graphql_jwt.utils import jwt_payload, jwt_encode
from django.contrib.auth.models import User


class UserSerializer(serializers.ModelSerializer):
    # sending back token for initial login
    token = serializers.SerializerMethodField()

    def get_token(self, obj):
        # generate new token
        return jwt_encode(jwt_payload(obj))
        
    class Meta:
        model = User
        fields = ["password", "username", "email", "token"]

with this graphql mutation:

mutation {
  createUser(input: {username: "john_doe", password: "f", email: "john_doe@f.com"}) {
    errors {
      messages
    },
    username,
    token
  }
}

it returns the response:

{
  "data": {
    "createUser": {
      "errors": null,
      "username": "john_doe",
      "token": "john_doe"
    }
  }
}

Here you can see that the token will be filled with the used username instead of a generated token.

I really want to use those Serializer classes because of the validation function. I have already read about some limitations e.g "extra_kwargs" but did not found something specific about SerializerMethodField class. I have already thought about a workaround by initializing the Serializer class inside of a class UserMutation(graphene.Mutation) and calling its method is_valid but I hope for an explanation what I am doing wrong to use the documented procedure.

---

I would expect:

{
  "data": {
    "createUser": {
      "errors": null,
      "username": "john_doe",
      "token": "eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VybmFstZSI6InlkbGFkbWluIiwiZXhwIjoxNTY3NDMwMdzk3LCJvcmlnSWFs0IjoxNTY3NDMwMDk3fQ.zt-XQN5k8zWrAMpNWN-xt_vRe6topM7kzFVSm6CT0Vs"
    }
  }
}

---

• CPython 3.7.4 64bit
• Windows 10

---

• django 2.2.4
• graphene-django 2.5.0
• djangorestframework 3.10.2

[B4rtware]

After further investigation, I found the origin of the underlying problem. Digging to the method
graphene_django>rest_framework>mutation.py>SerializerMutation>perform_mutate and applying this modification

    from rest_framework import serializers                             # modified
    ...
    @classmethod
    def perform_mutate(cls, serializer, info):
        obj = serializer.save()

        kwargs = {}
        for f, field in serializer.fields.items():
            if not field.write_only:
                if isinstance(field, serializers.SerializerMethodField): # modified
                    kwargs[f] = field.to_representation(obj)             # modified
                else:                                                    # modified
                    kwargs[f] = field.get_attribute(obj)

        return cls(errors=None, **kwargs)

it worked.

---

The issue is at field.get_attribute(ob). Inspecting the SerializerMethodField inside of the rest_framework you can see that there is no get_attribute function instead it will use the get_attribute from the base class Field which I think (not really traced it completely) just returns the representation of the obj. In my case this is the user object and it will return the users username because __str__ is overridden to do so.

---

So I was interested in how the django_rest_framework handles the get_attribute issue. And I have found out that there is a method called to_representation ON the Serializer class which indeed returns all field names and also the correct values for each field:

OrderedDict([('password', 'f'), ('username', 'john_doe'), ('email', 'john_doe@f.com'), ('token', 'eyJ0eXAiOiJKV1QiLCJhbGciOiJIUzI1NiJ9.eyJ1c2VybmFtZSI6ImdkZGRkZHNzZHNkc3Nkc3Nzc2Rhc3Nkc2RkZGRkcyIsImV4cCI6MTU2NzUxNjc2Miwib3JpZ0lhdCI6MTU2NzUxNjQ2Mn0.fdiAgROOyqlV99S5R2mbV0DImp1YqONNVix6i67sCrs')])

And I am curious why the serializer.to_representation was not used and instead the loop over each field. Maybe I will get it after I am using this modification.

Using the mentioned method above, the function could be reduced to:

    @classmethod
    def perform_mutate(cls, serializer, info):
        obj = serializer.save()
        kwargs = serializer.to_representation(obj)
        return cls(errors=None, **kwargs)

Which I will now use as a workaround.
Any feedback is appreciated.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.