any progress here?

Not really, I've dropped the ball a bit on this one. Will try to make some progress...

Oh well, that's all passing now... Hopefully that covers the main use case reported in#9378

There is also the other semi-related bug#9358 which is NOT in scope of this PR, I believe it will be fixed by#9360

Is a new version expected with this fix and others related like#9360 ? The support for UniqueConstraint is not complete and may cause issues…

thanks!

I'm not entirely sure under which circumstances, but this change turns aSerializerMethodField() whose name appears in the uniqueness constraint of the underlying model into aHiddenField(default=None).

I'll keep investigating.

Here is a (minimal???) reproducer:

deftest_tba():classTestModel(models.Model):field_1=models.IntegerField(null=True)field_2=models.IntegerField(null=True)classMeta:unique_together= (("field_1","field_2"),)classTestSerializer(serializers.ModelSerializer):field_1=serializers.SerializerMethodField()defget_field_1(self)->str:return"TEST"classMeta:model=TestModelfields= ["field_1","field_2"]fields=TestSerializer().fieldsassertisinstance(fields["field_1"],serializers.SerializerMethodField)

    assert isinstance(fields["field_1"], serializers.SerializerMethodField)E   AssertionError: assert FalseE    +  where False = isinstance(HiddenField(default=None), <class 'rest_framework.fields.SerializerMethodField'>)E    +    where <class 'rest_framework.fields.SerializerMethodField'> = serializers.SerializerMethodField

I'm not entirely sure under which circumstances, but this change turns a SerializerMethodField() whose name appears in the uniqueness constraint of the underlying model into a HiddenField(default=None).

I'll keep investigating.

Seeing the same thing in my codebase when upgrading to 3.16!