Opened 6 years ago
Closed 6 years ago
#29893 closed Bug (wontfix)
Unexpected behavior of DetailView on a model defining __len__() returning False
Reported by: | Max | Owned by: | Max |
---|---|---|---|
Component: | Generic views | Version: | 2.1 |
Severity: | Normal | Keywords: | |
Cc: | Max | Triage Stage: | Unreviewed |
Has patch: | yes | Needs documentation: | no |
Needs tests: | no | Patch needs improvement: | no |
Easy pickings: | no | UI/UX: | no |
Description
Description
When using Django's DetailView on a model defining a __len__()
method, the model object is not always added to context. This is the case, when __len__()
of the model object returns zero. The behavior is caused by BaseDetailView.get()
calling SingleObjectMixin.get_context_data()
which checks like this, if self.object
was set before bySingleObjectMixin.get_object()
:
if self.object: # adds self.object to context
Because this check is happening implicitly and not against None
, pythons default implementation calling self.object.__len__()
(https://docs.python.org/3/library/stdtypes.html#truth-value-testing) causes this condition to evaluate to False, at least if __len__()
returns zero. This leads to not having the wanted object in context data.
Steps to reproduce:
Create an empty model defining a __len__
method returning zero and a DetailView like this:
class TestModel(models.Model): def __len__(self): return 0 class TestModelDetailView(DetailView): model = TestModel template_name = 'test.html' # some template
After creating a TestModel object, the described behavior can be reproduced when calling the Detailview (url pattern like this /test/<int:pk>/
).
Proposed solution
Simply checking if self.object is the default value None prevents this not easy to find bug.
SingleObjectMixin.get_queryset()
also has a similar check
if self.model: # some action
which does not seem to cause problems, because self.model
is a class and no object. It should be considered to also check against the default value for self.model here, which is None.
Change History (4)
comment:1 by , 6 years ago
Cc: | added |
---|---|
Owner: | changed from | to
Status: | new → assigned |
comment:2 by , 6 years ago
comment:3 by , 6 years ago
I agree that the case of having len defined on a model may be rare, but i would strongly propose doing the modification of this check. Although this may not be exhaustive (you mentioned cases like self.instance
), i think it is worth it, because the current situation leads to highly unexpected behavior, which can be definitely avoided in this particular case.
comment:4 by , 6 years ago
Resolution: | → wontfix |
---|---|
Status: | assigned → closed |
I agree with Simon that trying to adapt the Django ecosystem to this convention is likely an unreasonable burden. You can write to the DevelopersMailingList if you want to try to convince the community otherwise. It would certainly help your argument if you explained why you want to define __len__()
as such. Others may suggest a better solution for the problem you're trying to solve.
I guess adding an
is None
would do for this case but I wouldn't be surprised if a lot of other cases still break. For example, there's probably a few model forms or admin APIs that do something alongif self.instance
and would need to be adjusted as well.Given defining
__len__()
on a model instance is an edge case and your issue can be worked around by defining adef __bool__(self): return True
I'm tempted to close this issue as _wont fix_ but I'll let other contributors chime in. I feel like the burden of making sure all APIs dealing with optional model instances truthiness appropriately is not worth the small benefit here.