Fix unique together validator doesn't respect condition's fields

[kalekseev]

fixes #9358

[kalekseev]

not sure how reliable this method of fields extraction

[sevdog]

I duobt that this will work when the condition uses a cross-reference between two fields like: Q(first_field__gte=F('second_field')), however I dunno if an unique constraint like that makes sense.

I would like to see also a more complex condition like Q(fielda__isnull=True) | Q(fieldb=False) to be addressed by this validator.

[tomchristie]

@kalekseev - Could you talk me through this?

My understanding is that it's resolving an issue in #7438, is that correct?

[kalekseev]

@kalekseev - Could you talk me through this?

My understanding is that it's resolving an issue in #7438, is that correct?

UPDATE: Yes, this pr is supposed to fix the initial implementation of UniqueConstraint support but at the moment it doesn't work. The problem is that DRF doesn't check if fields in data match the unique constraint condition or not, so for example constraint is:

            models.UniqueConstraint(
                name="unique_constraint_model_together_uniq2",
                fields=('race_name', 'position'),
                condition=models.Q(fancy_conditions__gte=10),
            )

we have record in db that should be unique

        UniqueConstraintModel.objects.create(
            race_name='condition',
            position=1,
            global_id=10,
            fancy_conditions=10
        )

The data below should be valid because it has fancy_conditions=9 that doesn't match constraint condition models.Q(fancy_conditions__gte=10):

       serializer = UniqueConstraintSerializer(data={
            'race_name': 'condition',
            'position': 1,
            'global_id': 11,
            'fancy_conditions': 9,
        })

but this should fail

      serializer = UniqueConstraintSerializer(data={
            'race_name': 'condition',
            'position': 1,
            'global_id': 11,
            'fancy_conditions': 11,
        })