Opened 10 years ago
Closed 10 years ago
#23001 closed Bug (fixed)
Annotation breaks with deferred fields and select_related
Reported by: | jarshwah | Owned by: | Josh Smeaton |
---|---|---|---|
Component: | Database layer (models, ORM) | Version: | dev |
Severity: | Normal | Keywords: | |
Cc: | Triage Stage: | Ready for checkin | |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
I've discovered a bug that I think was introduced by: https://github.com/django/django/commit/0b6f05ede648ce62a5c91c7c38a0a362711f0656
The problem manifests as annotations being assigned an incorrect value when there are deferred fields and select_related is in use.
Passing Test:
def test_annotate_defer(self): qs = Book.objects.annotate( page_sum=Sum("pages")).defer('name').filter(pk=1) rows = [ (1, "159059725", 447, "The Definitive Guide to Django: Web Development Done Right") ] self.assertQuerysetEqual( qs.order_by('pk'), rows, lambda r: (r.id, r.isbn, r.page_sum, r.name) )
Failing Test:
def test_annotate_defer_select_related(self): qs = Book.objects.select_related('contact').annotate( page_sum=Sum("pages")).defer('name').filter(pk=1) rows = [ (1, "159059725", 447, "Adrian Holovaty", "The Definitive Guide to Django: Web Development Done Right") ] self.assertQuerysetEqual( qs.order_by('pk'), rows, lambda r: (r.id, r.isbn, r.page_sum, r.contact.name, r.name) )
The problem is in the iterator method of django.db.models.query, which results in the aggregate_start variable being off by len(deferred):
if load_fields and not fill_cache: # Some fields have been deferred, so we have to initialize # via keyword arguments. skip = set() init_list = [] for field in fields: if field.name not in load_fields: skip.add(field.attname) else: init_list.append(field.attname) model_cls = deferred_class_factory(self.model, skip) else: model_cls = self.model init_list = [f.attname for f in fields]
fill_cache is set to true when select_related is in use, which skips to the else block. The else block doesn't take deferred fields into account like the if block does. I'll attempt to patch this by including similar logic in the else block (skipping deferred fields), but I'm not yet sure what else that might affect.
Change History (5)
comment:1 by , 10 years ago
comment:3 by , 10 years ago
Triage Stage: | Unreviewed → Ready for checkin |
---|---|
Type: | Uncategorized → Bug |
comment:4 by , 10 years ago
Owner: | changed from | to
---|---|
Status: | new → assigned |
comment:5 by , 10 years ago
Resolution: | → fixed |
---|---|
Status: | assigned → closed |
Both tests pass on the 1.7 branch.