#9982 closed Bug (fixed)
Inconsistent behavior on model save depending on whether OneToOneField is a primary key
Reported by: | Owned by: | ||
---|---|---|---|
Component: | Database layer (models, ORM) | Version: | dev |
Severity: | Normal | Keywords: | OneToOneField primary_key IntegrityError |
Cc: | Can Sarıgöl | Triage Stage: | Accepted |
Has patch: | no | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
Hi,
When using a OneToOneField to make one model extend another, I noticed a behavioral inconsistency depending on whether the related model's OneToOneField was also the primary key or not.
The inconsistency is to do with when save() can be called on model instances.
Consider the following example models:
class BaseModel(models.Model): pass class ExtendedModel(models.Model): link = models.OneToOneField(BaseModel)
It is not possible to first instantiate these models and later on save them - an IntegrityError would be raised when saving the ExtendedModel. E.g.,
o1 = BaseModel() o2 = ExtendedModel(link=o1) o1.save() o2.save()
when o2 is saved it will raise an IntegrityError because it doesn't have a value for the primary key of the model it is related to. I think o1's primary key is copied when o2 is initialized, and is None at this point because o1 isn't saved yet.
However, if the definition of ExtendedModel is changed so that the OneToOneField is also the primary key, it now becomes possible to save the model instances in this order. To test this I changed the model definition as follows:
class ExtendedModel(models.Model): link = models.OneToOneField(BaseModel, primary_key=True)
When using this model definition it seems that o1's primary key is copied when o2 is saved, as opposed to when o2 is initialized. This inconsistency caused a lot of head scratching here.
Thanks!
Sean
Change History (11)
comment:1 by , 16 years ago
Owner: | changed from | to
---|---|
Status: | new → assigned |
comment:3 by , 16 years ago
milestone: | → 1.1 |
---|---|
Triage Stage: | Unreviewed → Accepted |
comment:4 by , 16 years ago
milestone: | 1.1 → 1.2 |
---|
Pushing to 1.2: this is an inconsistancy, not an outright bug.
comment:5 by , 15 years ago
milestone: | 1.2 |
---|---|
Owner: | removed |
Status: | assigned → new |
Moving off the 1.2 milestone. As noted in the first comment, this should be raising an error in the case that appears to work, in the best case. So it's not a case of something not working that should.
comment:6 by , 14 years ago
Severity: | → Normal |
---|---|
Type: | → Bug |
comment:10 by , 6 years ago
Resolution: | → fixed |
---|---|
Status: | new → closed |
Version: | 1.0 → master |
Fixed in 519016e5f25d7c0a040015724f9920581551cab0.
The second case shouldn't work either. I'll have to check whether it's really doing something sensible there (as opposed to just not raising an error). It may be because of something in place for model inheritance that it's working; I seem to recall we added an admin-related workaround that was a bit of a hack. In which case, it'll just need documentation.
Django cannot read your mind and does not automatically save things. So you do need to save something before using it as a reference. The first case is consistent and you should always save before using something as a related instance.
Leaving open until I understand and document (at a minimum) what's happening.