I’d like to use this constraint on my Django model:

models.UniqueConstraint(
    fields=["edificio", "studio"],
    condition=models.Q(attiva=True),
    name="unique_attiva_per_edificio_studio",
    violation_error_message=_(
        "cannot have two active configurations for the same studio",
    ),
)

But since I’m on MariaDB (and I cannot change it) conditional constraints/indexes aren’t supported yet. So I’m extending the validate_constraints function in my EdificioConfigurazione model to check it.

def validate_constraints(self, exclude=None):
    super().validate_constraints(exclude)

    # TODO: this will be removed once MySQL/MariaDB supports conditional constraints
    manager = EdificioConfigurazione.objects
    multiple_attiva_per_edificio = (
        manager.values("edificio", "studio")
        .annotate(
            rcount=models.Count("id", filter=models.F("attiva")),
        )
        .filter(rcount__gt=1)
    )
    if multiple_attiva_per_edificio.exists():
        violation_error_message = _(
            "cannot have two active configurations for the same studio",
        )
        raise ValidationError(violation_error_message)

Assuming this is the right approach to do this (I’m open to suggestions), I cannot enforce this check since the instance to be added/modified (which is self) is not committed to the DB yet. How can I use the instance? My idea would be looking at the self.pk field and:

• if self.pk is None: I add the instance to the queryset before the checks
• else: the instance is already in the DB (I’m modifying it for example) so I need to update the queryset to use it instead of the one committed.

Once again, I don’t know if extending validate_constraints is the right approach but how would I implement this logic?