Opened 13 years ago
Closed 9 years ago
#18081 closed Bug (fixed)
Syncdb doesn't create database constraints for foreign keys referencing a proxy model.
Reported by: | Anssi Kääriäinen | Owned by: | nobody |
---|---|---|---|
Component: | Database layer (models, ORM) | Version: | 1.4 |
Severity: | Normal | Keywords: | |
Cc: | Simon Charette | Triage Stage: | Accepted |
Has patch: | no | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
Add this test to modeltests/proxy_models/tests.py:
class TransactionalProxyModelTests(TransactionTestCase): def test_proxy_fk(self): """ Test that the DB contains proper foreign keys for proxy model references. """ @transaction.commit_on_success def create_failing_pk(): t = TrackerUser.objects.create(status='bar') Improvement.objects.create(summary='foof', version='foof', reporter_id=1, associated_bug_id=1, assignee=t) self.assertRaises(IntegrityError, create_failing_pk)
On MySQL this does not fail when using InnoDB. It should fail, as both reporter and associated bug with ID=1 are missing.
The reason for this is that in django/db/backends/creation.py, sql_for_pending_references() is a check:
if not model._meta.managed or model._meta.proxy: skip creation
Now, that is incorrect: the model in question is the model we are _referring_, not the model from where the PK is from. So, under MySQL foreign keys to proxy models do not get enforced in the DB.
In addition the same bug is there for other databases, too. However, this one does not show under PostgreSQL as the foreign key is created inline. However, were the order of the models different in models.py also PostgreSQL would fail to create the foreign key.
The fix is luckily very simple: just remove the "or proxy" part. The managed part is correct: the referenced model can be a view for example, in which case referencing it would be a failure.
Attachments (1)
Change History (9)
comment:1 by , 12 years ago
Triage Stage: | Unreviewed → Accepted |
---|
comment:2 by , 12 years ago
Cc: | added |
---|
comment:3 by , 9 years ago
by , 9 years ago
Attachment: | 18081-test.diff added |
---|
comment:4 by , 9 years ago
I don't see why that line in the SchemaEditor
should be wrong. Proxy models don't have any database representation on their own (https://docs.djangoproject.com/en/1.8/topics/db/models/#proxy-models) and thus don't have any indices. Without digging into the issue too deep, this seems more like a bug in the assignment of related objects to me.
comment:5 by , 9 years ago
Cc: | added; removed |
---|---|
Summary: | Proxy model foreign keys not created properly by syncdb → Syncdb doesn't create database constraints for foreign keys referencing a proxy model. |
This is not an issue anymore since calls to _create_fk_sql are not conditional to not field.target_field.model._meta.proxy
(create_model, add_field, alter_field).
I think a more appropriate regression test could live in test_operations
and issue an AddField
of a foreign key pointing to ProxyPony
and make sure the constraint exists. Naturally the test should be skipUnlessDBFeature('supports_foreign_keys')
.
comment:8 by , 9 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
I am not sure if this is still an issue with the
SchemaEditor
backends. I updated Anssi's test (attached) and it passes at 28cb272a7279e6dfc4d5c53838ebf7343c3e66b5, however, theSchemaEditor._model_indexes_sql()
method still hasa condition to skip proxy models.