The titles of admin list_filters aren't translated anymore

[Matthias Kestenholz]

FieldListFilter.title is not initialized with verbose_name of fields anymore but with field_path instead. Because of this the admin change_list shows 'Nach status' and 'Nach visitor_name' instead of 'Nach Status' and 'Nach Besucher'.

[Matthias Kestenholz]

The attached patch probably needs improvement, but I'll leave that to someone else.

I think the tests are verifying the wrong behavior here, too.

[dmclain]

This looks like a duplicate of #15221

[Julien Phalip]

#15221 is a related, but different issue. #15221 is for displaying the _RelatedFieldListFilter_'s title properly.

But thanks, I was just going to point to it :-)

This ticket here is a regression (whereas the issue in #15221 has always been there), hence marking as blocker. The attached patch checks for the existence of a verbose_name attribute, similar to the approach in the patch for #15221 (which I'm also going to update in a minute).

[Matthias Kestenholz]

I don't think the patch fixes the whole issue. We should be able to display a proper verbose_name even for related objects (to support something like list_filter = ('related__object__attribute',)

Or am I misunderstanding something?

[Julien Phalip]

@mk: Agreed, which is what #15221 is suggesting. However, the patch here addresses the regression only, which is a slightly different issue.

FieldListFilter, as a super class, should avoid trying to guess about its child classes' business. The best it can do is provide a sensible default (i.e. verbose_name if possible, or field_path if not). The patch in #15221 lets RelatedFieldListFilter (i.e. the child class) override this default behaviour.

Again, this ticket is about fixing a regression that was introduced by [16144]. #15221 is about fixing a behaviour that has always existed.

Hope that makes sense :-)

[Matthias Kestenholz]

Ah yes, indeed. I was mislead by my own stupid patch.

Thanks!

[Jannis Leidel]

In [16171]:

Fixed #15971 -- Fixed regression introduced in r16144 that prevented the use of the field's verbose_name attribute. Thanks to mk and Julien.