带有源字段的UniqueValidator

Question

I have model UserProfile:

class UserProfile(models.Model):
   user = models.OneToOneField(User)

and serializer:

email = serializers.CharField(source='user.email', required=False,
                                  validators=[UniqueValidator(queryset=User.objects.all())])

Serialization works fine, but deserialization doesn't — it tries to file 'user.email' field in User model, and, of course, fails.

If I change User to UserProfile in the queryset, it fails with another error:

invalid literal for int() with base 10: 'admin@localhost'

Is it possible to set different sources for serialization and deserialization?

Answer 1

By default, the UniqueValidator expects that the source (or field name, if no source is given) can be used in the queryset that is provided to filter out the existing object . So by using a source that spans a relation (and has a dot), it will try to use that name when filtering the queryset which fails, as you've noticed.

You can fix this by subclassing UniqueValidator to override the filter_queryset method to filter it differently.

class CustomUniqueValidator(UniqueValidator):

    def filter_queryset(self, value, queryset):
        """
        Filter the queryset to all instances matching the given attribute.
        """
        filter_kwargs = {"email": value}
        return queryset.filter(**filter_kwargs)

This hard codes email as the filter, one possible option for not hard coding it would be to split self.field_name and get the last part of the dotted source.