#12163 closed (fixed)
Regression: Pickling error saving session with unsaved model instances (*_Deferred_)
| Reported by: | Ryan Fugger | Owned by: | nobody |
|---|---|---|---|
| Component: | Database layer (models, ORM) | Version: | dev |
| Severity: | Keywords: | pickle, session, model, instance, deferred | |
| Cc: | arv@… | Triage Stage: | Unreviewed |
| Has patch: | no | Needs documentation: | no |
| Needs tests: | no | Patch needs improvement: | no |
| Easy pickings: | no | UI/UX: | no |
Description
When attempting to store freshly-created unsaved model instances in a session, I get the following error:
PicklingError: Can't pickle <class 'order.models.OrderItem_Deferred_'>: attribute lookup order.models.OrderItem_Deferred_ failed
The object is an instance of OrderItem. Apparently django instantiates it as !OrderItem_Deferred_, and then pickle can't deal with it?
Using svn revision 11723. Worked fine in 11616.
Attachments (2)
Change History (11)
comment:1 by , 15 years ago
| milestone: | → 1.2 |
|---|
comment:2 by , 15 years ago
I'll try to work up a test case when I have time. Strangely, when I reverted to r11616, the __reduce__ function was broken (model used before assignment), even though it was definitely working fine three days ago. So maybe I'm doing something wrong, although I don't think my surrounding code has changed since then.
In the meantime, replacing model = self.__class__ with model = self._meta.proxy_for_model near the start of db.models.base.Model.__reduce__ seems to fix the issue for me. I'm not sure if this is a good general solution or not, because I don't really know how deferred fields work.
by , 15 years ago
| Attachment: | base.py.diff added |
|---|
Set model to _meta.proxy_for_model by default for deferred
comment:3 by , 15 years ago
(Sorry, probably better not to replace model = self.__class__, but just to set model = self._meta.proxy_for_model as a default when self._deferred is True.
comment:4 by , 15 years ago
| Cc: | added |
|---|
comment:5 by , 15 years ago
I'd just like to add that it's revision 11691 that causes this, using 11690 should fix things for now.
comment:6 by , 15 years ago
Thanks Andrew. r11690 does work. I hadn't cleared out my session last time I tried a pre-11691 revision, so pickled *_Deferred_ objects were still hanging around and causing errors.
comment:7 by , 15 years ago
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
There have been some recent changes to the infrastructure around pickling of deferred fields - evidently you have found a case that wasn't covered by our test suite.
Can you please provide a complete test case - i.e., a minimal set of models, and the exact sequence of commands that fails.