In the serializer used to create a model I want to rename my model field to (field_name)_id so it's clearer for API consumers that this field is an ID field. The model also has a unique_together
constraint on some fields. However when validation runs in the serializer, it fails with a KeyError that the field does not exist:
...rest_framework/utils/serializer_helpers.py", line 148, in __getitem__
return self.fields[key]
KeyError: 'question'
Is there a simple way to get this to work? Minimal example code below.
Model
class MyModel(Model):
question = ForeignKey('uppley.Question', null=False, on_delete=PROTECT)
user = ForeignKey('catalystlab.User', null=False, on_delete=PROTECT)
class Meta:
unique_together = ('question', 'user',)
Serializer
class MyCreateSerializer(ModelSerializer):
question_id = PrimaryKeyRelatedField(
write_only=True,
source='question',
queryset=Question.objects.all(),
)
user_id = PrimaryKeyRelatedField(
write_only=True,
source='user',
queryset=User.objects.all(),
)
class Meta:
model = MyModel
fields = ('question_id', 'user_id',)
test.py - test for demonstration purposes
question = QuestionFactory()
user = UserFactory()
data = {
'question_id': question.id,
'user_id': user.id,
}
serializer = MyCreateSerializer(data=data, write_only=True)
is_valid = serializer.is_valid(raise_exception=True) #KeyError exception raised here.
Previously with DRF 3.10.3 this all worked fine, however with 3.11.0 this now throws a KeyError as mentioned above.
What I have tried
Removing the source
field on PrimaryKeyRelatedField
for user_id
and question_id
in the Serializer actually results in bypassing the unique_together validation in DRF and the KeyError is avoided. However the validated data is not mapped back to the original field names (user
and question
). In this case we have to manually change the keys back to their original names before we can create an instance of the Model from the validated data.
Is there a better way to do this?