django-group-by 0.1.1

Group by arbitrary model fields

This package provides a mixin for Django QuerySets that adds a method group_by that behaves mostly like the values method, but with one difference: its iterator does not return dictionaries, but a model-like object with model instances for related values.

Installation

Install from this repository:

pip install -e https://github.com/kako-nawao/django-group-by.git

Usage

Create a QuerySet subclass with the GroupByMixin to use in a model’s manager:

# models.py

from django.db import models
from django.db.models.query import QuerySet
from django_group_by import GroupByMixin

class BookQuerySet(QuerySet, GroupByMixin):
    pass

class Book(Model):
    objects = BookQuerySet.as_manager()

    title = models.CharField(max_length=100)
    author = models.ForeignKey('another_app.Author')
    publication_date = models.DateField()
...

Then use it just like values, and you’ll get a similar query set:

>>> some_rows = Book.objects.group_by('title', 'author', 'author__nationality).distinct()
>>> some_rows.count()
4

The difference is that every row is not a dictionary but an AggregatedGroup instance, with only the grouped fields:

>>> row = some_rows[0]
>>> row
<AggregatedGroup for Book>
>>> row.title
The Colour of Magic
>>> row.publication_date
*** AttributeError: 'AggregatedGroup' object has no attribute 'publication_date'

But of course the main advantage is that you also get related model instances, as far as you want:

>>> row.author
<Author: Terry Pratchett>
>>> row.author_nationality
<Nation: Great Britain>

Related Model ID Only

The previous example shows a difference in behaviour from Django’s values: we’re grouping by the foreign key author, but instead of getting only the ID we get the full instance. Why? Because it’s more useful, and I think that getting {‘author’: 5} as a result is just weird.

If you just want the ID you can specify it:

>>> some_rows = Book.objects.group_by('title', 'author_id', 'author__nationality_id).distinct()