[PATCH 1/4] Enforce ordering of bundles in REST test

[Stephen Finucane]

On Fri, 2017-09-08 at 07:20 +1000, Daniel Axtens wrote:
> Stephen Finucane <stephen at that.guru> writes:
> 
> > On Mon, 2017-09-04 at 01:14 +1000, Daniel Axtens wrote:
> > > This is required to make the tests pass on a recent version of
> > > postgres.
> > > 
> > > Signed-off-by: Daniel Axtens <dja at axtens.net>
> > > ---
> > >  patchwork/tests/test_rest_api.py | 2 +-
> > >  1 file changed, 1 insertion(+), 1 deletion(-)
> > > 
> > > diff --git a/patchwork/tests/test_rest_api.py
> > > b/patchwork/tests/test_rest_api.py
> > > index abffd17fddec..d4a84bd8c5ad 100644
> > > --- a/patchwork/tests/test_rest_api.py
> > > +++ b/patchwork/tests/test_rest_api.py
> > > @@ -738,7 +738,7 @@ class TestBundleAPI(APITestCase):
> > >          # authenticated user
> > >          # should see the public and private bundle
> > >          self.client.force_authenticate(user=user)
> > > -        resp = self.client.get(self.api_url())
> > > +        resp = self.client.get(self.api_url() + '?order=id')
> > 
> > Maybe I'm misunderstanding things here, but doesn't this suggest users
> > would always need to add the 'order' filter just to get things working
> > sanely on a PostgreSQL-backed instance? If so that doesn't sound very
> > useful, and it would suggest that the test is doing its job.
> 
> I didn't think bundles were intrinsically ordered - unlike
> patches/series which have a pretty obvious date-ordering, I think
> bundles are a bit more arbitrary. Is there an obvious sane ordering we
> should be enforcing? (IMO you could make a decent argument for either
> sorting by name or sorting by (creation|modification) date.)

That's a very good point. I looked at 'patchwork.view.bundle.bundle_list', and
noted that there is no default bundling applied, however, I do think we want to
avoid the same request returning different responses simply because of the
database backend in use. Given that there's no precedent here, I guess this
means we're free to choose? Name or ID (which is implicitly data-based) would
be equal in my mind.

> > It doesn't seem like there's a way to apply a default filter using the
> > 'OrderingFilter' [1]. However, could we simply add a sort to the queryset
> > returned by 'patchwork.api.bundle.BundleFilter.get_queryset'?
> 
> If we went down this route, I think we would probably want to order them
> using a class Meta ordering statement in the model (see e.g. the
> ordering of projects by linkname in models.py around line 101.) I assume
> the REST stuff honours that ordering.

The only disadvantage to taking this route is that it will introduce a
migration, and I'm not sure if a migration is something we should be in the
habit of backporting. Perhaps we modify 'get_queryset' for 'stable/2.0' (with
tests, of course) and take the Meta class approach for 'master'? Open to other
ideas here, of course.

Stephen