Loic Bistuer has been working on a patch to allow using custom querysets
with prefetch_related(). There was a lot of trying out different APIs. For
example the kwargs based API discussed earlier on this mailing list was
tried, but it turns out that due to prefetch ordering issues it wasn't
actually workable.
The final API is somewhat close to what the original R() object API in
ticket #17001 was. Custom prefetches are done with Prefetch objects. Simple
prefetches are possible with either Prefetch or the traditional lookup
approach:
Article.objects.prefetch_related('authors') ==
Article.objects.prefetch_related(Prefetch('authors'))
Custom queryset can be provided. For example this will fetch to
authors.all() with different ordering:
Article.objects.prefetch_related(Prefetch('authors',
queryset=Author.objects.order_by('name')))
It is also possible to fetch to a list:
Article.objects.prefetch_related(Prefetch('authors',
queryset=Author.objects.order_by('name'),
to_attr='authors_ordered_by_name'))
Now each fetched article will have authors_ordered_by_name list.
Notably when prefetching to a list performance will be up to 2x better
compared to prefetching into a queryset (thus solving #20577). OTOH you
can't do further queryset operations on the list. But if you want that,
then your prefetch will be wasted in any case.
It is also possible to fetch the same relation multiple times to different
attrs:
Article.objects.prefetch_related(
Prefetch('authors', queryset=Author.objects.order_by('name'),
to_attr='authors_ordered_by_name'),
Prefetch('authors', queryset=Author.objects.order_by('age'),
to_attr='authors_ordered_by_age')
)
Finally, nesting prefetches with custom querysets is possible:
authors_starting_with_a_with_friends =
Author.objects.filter(name__istartswith='a').prefetch_related('friends')
Article.objects.prefetch_related(
Prefetch('authors', queryset=authors_starting_with_a_with_friends,
to_attr='authors_a'))
Above the 'friends' prefetch could be a Prefetch with custom queryset.
While reviewing this I wondered about multidb support. Should there be
automatic .using() call for the provided queryset if it doesn't have
explicitly set db? Currently the queryset is used as-is, which can mean it
queries different database than the outer queryset. I don't know multidb
that well so opinions welcome.
Apart of the multidb issue I consider the patch as ready for checkin.
If you are interested in this feature reviewing is the best way to move
this forward. While code and docs reviews are welcome, just trying this out
with your use-cases is an excellent way to review the provided
functionality.
This issue's ticket is https://code.djangoproject.com/ticket/17001, and
pull request can be found from https://github.com/django/django/pull/1826.
- Anssi
--
You received this message because you are subscribed to the Google Groups
"Django developers" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
To post to this group, send email to [email protected].
Visit this group at http://groups.google.com/group/django-developers.
To view this discussion on the web visit
https://groups.google.com/d/msgid/django-developers/94009676-b398-4c11-82fe-f550b06c235a%40googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
