[patch] extra_lookup_kwargs and extra_context in infodicts should accept callbacks

[hugo]

the callback should be called with the request and the dict of the parsed urlpattern to allow dynamic creation of extra_lookup_kwargs based on url parts or request.GET stuff (for example to throw in some filtering).

[hugo]

#779 is a duplicate

[Tom Tobin <korpios@…>]

Patch against r1217 to implement optional callback for extra_lookup_kwargs

[Tom Tobin <korpios@…>]

As per discussion in #779, I have attached a patch to implement half of the requested behavior in this ticket (the extra_lookup_kwargs callbacks). If this looks good, I'll try to implement the other half for extra_context callbacks.

[Tom Tobin <korpios@…>]

I should note that, in the patch, object_list in list_detail has the only callback called with just one argument (request), rather than two, as there are no significant URL parts there.

[Tom Tobin <korpios@…>]

Patch against r1263 to implement ticket request (covers both extra_context and extra_lookup_kwargs)

[Tom Tobin <korpios@…>]

I've attached a patch against r1263 (trunk) to implement all of the requested behavior in this ticket. All related callback functions must accept two positional arguments: request and lookup_kwargs (which will be an empty dictionary in extra_lookup_kwargs of list_detail.object_list and the extra_context of create_update.create_object).

I've been debating whether to have those two callback functions only take request, but opted for the behavior in the patch as to allow for a single defined callback function to handle both cases (with and without a corresponding lookup_kwargs to pass in).

[Luke Plant]

You can now use the 'context_processors' argument to generic views to have dynamic 'extra_context'. As for lookup_kwargs, you can easily write a wrapper function that filters the QuerySet as required then passes it on to the generic view function, and I think this approach, which is so easy, is better than trying to cram everything to the generic views and further complicating their interface.

I think this bug and #779 can be closed now.

[Adrian Holovaty]

Closing for the reason outlined by lukeplant in the previous comment.