Maintain uniqueness on django-localized-fields - django

Question

I'm trying to avoid having duplicate localized items stored in a Django-rest-framework app, django-localalized-fields package with a PostgreSQL database I can't find any way to make this work.
(https://pypi.org/project/django-localized-fields/)
I've tried writing custom duplicate detection logic in the Serializer, which works for create, but for update the localized fields become null (they are required fields, so I receive a not null constraint error). It seems to be django-localized-fields utility which is causing this problem.
The serializer runs correctly (create/update) when I'm not overriding create/update in the serializer by defining them separately.
I've also tried adding unique options to the database in the model, which does not work - duplicates are still created. Using the standard unique methods, or the method in the django-localized-fields documentation (uniqueness=['en', 'ro']).
I've also tried the UniqueTogetherValidator in Django, which also doesn't seem to support HStore/localizedfields.
I'd appreciate some help in tracking down either how to fix the update in the serializer or place a unique constraint in the database. Since django-localized-fields uses hstore in PostgreSQL it must be a common enough problem for applications using hstore to maintain uniqueness.
For those who aren't familiar, Hstore stores items as key/value pairs within a database. Here's an example of how django-localized-fields stores language data within the database:
"en"=>"english word!", "es"=>"", "fr"=>"", "frqc"=>"", "fr-ca"=>""

Answer 1

django-localized-fields constraint unique values only per the same language. If you want to achieve that values in a row don't collide with values in another row, you have to validate them on Django and database level.
Validation in Django
In Django you can create custom function validate_hstore_uniqueness, which is called everytime is model validated.
def validate_hstore_uniqueness(obj, field_name):
value_dict = getattr(obj, field_name)
cls = obj.__class__
values = list(value_dict.values())
# find all duplicite existing objects
duplicite_objs = cls.objects.filter(**{field_name+'__values__overlap':values})
if obj.pk:
duplicite_objs = duplicite_objs.exclude(pk=obj.pk)
if len(duplicite_objs):
# extract duplicite values
existing_values = []
for obj2 in duplicite_objs:
existing_values.extend(getattr(obj2, field_name).values())
duplicate_values = list(set(values) & set(existing_values))
# raise error for field
raise ValidationError({
field_name: ValidationError(
_('Values %(values)s already exist.'),
code='unique',
params={'values': duplicate_values}
),
})
class Test(models.Model):
slug = LocalizedField(blank=True, null=True, required=False)
def validate_unique(self, exclude=None):
super().validate_unique(exclude)
validate_hstore_uniqueness(self, 'slug')
Constraint in DB
For DB constraint you have to use constraint trigger.
def slug_uniqueness_constraint(apps, schema_editor):
print('Recreating trigger quotes.slug_uniqueness_constraint')
# define trigger
trigger_sql = """
-- slug_hstore_unique
CREATE OR REPLACE FUNCTION slug_uniqueness_constraint() RETURNS TRIGGER
AS $$
DECLARE
duplicite_count INT;
BEGIN
EXECUTE format('SELECT count(*) FROM %I.%I ' ||
'WHERE id != $1 and avals("slug") && avals($2)', TG_TABLE_SCHEMA, TG_TABLE_NAME)
INTO duplicite_count
USING NEW.id, NEW.slug;
IF duplicite_count > 0 THEN
RAISE EXCEPTION 'Duplicate slug value %', avals(NEW.slug);
END IF;
RETURN NEW;
END;
$$ LANGUAGE plpgsql;
DROP TRIGGER IF EXISTS slug_uniqueness_constraint on quotes_author;
CREATE CONSTRAINT TRIGGER slug_uniqueness_constraint
AFTER INSERT OR UPDATE OF slug ON quotes_author
FOR EACH ROW EXECUTE PROCEDURE slug_uniqueness_constraint();
"""
cursor = connection.cursor()
cursor.execute(trigger_sql)
And enable it in migrations:
class Migration(migrations.Migration):
dependencies = [
('quotes', '0031_auto_20200109_1432'),
]
operations = [
migrations.RunPython(slug_uniqueness_constraint)
]
Probably is a good idea to also create GIN db index for speeding up lookups:
CREATE INDEX ON test_table using GIN (avals("slug"));