What's the motivation behind that change ? Why isn't the django validator enough ?

@xordoquy in our codebase we are using a different Django Paginator for some bigger tables (where thecount(*) takes too much time, and an estimated page-count is enough.

This would help us setting our own paginator without having to copy/paste thepaginate_queryset method

Out of curiosity, what do you use for estimated page counts?
Also, is modifyingget_paginated_response an alternative here, or is that not sufficient?

Follow up... I guessnum_pages performs a count() - might be good if we could avoid that?

we are using the postgres statistics for this table. it has currently around 3 Mio records, and the rough value is exact enough.

we already have modifiedget_paginated_response (similar to theHeader based example in the docs, which removes one call toPaginator.count.

Butpaginate_queryset callsPaginator.num_pages, which also callsPaginator.count.

(a count in these specific tables takes around 2-3 seconds)

Sounds sensible to me.
Would probably rename thedjango_paginator_class to simplypaginator_class

django_paginator_class seems right - differentiate from our pagination controls.
I'm not 100% convinced if we want to add public API around this or not, but on balance I guess probably okay as-is.

But paginate_queryset calls Paginator.num_pages, which also calls Paginator.count.

Indeed. Making this easier to eliminate might be a good plan

Milestoning to push this to the top for a final review.

Thanks@syphar!

thanks@tomchristie and@xordoquy :)