Automatic prefetching for Django
Never worry about n+1 performance problems again
This project aims to automatically perform the correct
select_relatedand
prefetch_relatedcalls for your django-rest-framework code. It does this by inspecting your serializers, seeing what fields they use, and what models they refer to, and automatically calculating what needs to be prefetched.
pip install django-auto-prefetching
This is a ViewSet mixin you can use, which will automatically prefetch the needed objects from the database. In most circumstances this will be all the database optimizations you'll ever need to do:
Simply add it after your ModelViewSet class.
from django_auto_prefetching import AutoPrefetchViewSetMixin from rest_framework.viewsets import ModelViewSetclass BaseModelViewSet(AutoPrefetchViewSetMixin, ModelViewSet): queryset = YourModel.objects.all() serializer_class = YourModelSerializer
It supports all types of relational fields, (many to many, one to many, one to one, etc.) out of the box.
The
AutoPrefetchViewSetMixincannot see what objects are being accessed in e.g. a
SerializerMethodField. If you use objects in there, you might need to do some additional prefetches. If you do this and override
get_queryset, you will have to call
prefetchmanually as the mixin code is never reached.
import django_auto_prefetching from rest_framework.viewsets import ModelViewSetclass BaseModelViewSet(django_auto_prefetching.AutoPrefetchViewSetMixin, ModelViewSet): serializer_class = YourModelSerializer
def get_queryset(self): # Simply do the extra select_related / prefetch_related here # and leave the mixin to do the rest of the work queryset = YourModel.objects.all() queryset = queryset.select_related('my_extra_field') return django_auto_prefetching.prefetch(queryset, self.serializer_class)
Currently the project is currently being tested against Python 3.6 and 3.7 and Django 2.2 Pull Requests to support other versions are welcome.
The project is currently being used without issues in a medium-sized Django project(20.000 lines of code)
Contributions are welcome! To get the tests running, do the following: - Clone the repository. - If you don't have it, install pipenv - Install the dependencies with
pipenv sync --dev- Activate the virtualenv created by pipenv by writing
pipenv shell- Run the tests with
./manage.py test
MIT