Django: Count Aggregates with Conditions

In Django, One-to-Many relations are modeled by the ForeignKey. Under some circumstances, we may wish to collect some information (e.g. count, min, max, avg, etc) from related objects. For example, in a blog app, we may wish to build a table which lists the title and the number of comments in each posts. This feature can be easily implemented with QuerySet.annotate() and Aggregation. In this post, I will go through a simple example that covers the basic usages of Django Aggregation.

Let's start with the Post and Comment model. In example/, two models are defined. The Comment model has a ForeignKey field, thus many Comment objects may be related to a Post object:

from django.db import models

class Post(models.Model):
    title = models.CharField(max_length=255)

class Comment(models.Model):
    post = models.ForeignKey('Post', on_delete=models.CASCADE)
    comment = models.CharField(max_length=255)
    hidden = models.BooleanField(default=False)
    num_thumbs_up = models.IntegerField(default=0)

Initialize the database with:

python makemigrations
python migrate

Start the Django shell with:

python shell

Then, build some test data with:

from example.models import Post, Comment

x = Post.objects.create(title='x')
y = Post.objects.create(title='y')

x1 = Comment.objects.create(post=x, comment='x1', num_thumbs_up=5)
x2 = Comment.objects.create(post=x, comment='x2', num_thumbs_up=10)

y1 = Comment.objects.create(post=y, comment='y1', num_thumbs_up=1)
y2 = Comment.objects.create(post=y, comment='y2', num_thumbs_up=3)
y3 = Comment.objects.create(post=y, comment='y3', num_thumbs_up=5)
y4 = Comment.objects.create(post=y, comment='y4', num_thumbs_up=7,

Here comes our first example. To list posts with the number of comments of each posts, we have to annotate our query with QuerySet.annotate() function. The code below annotate the query with an extra field named num_comments. The value of num_comments is derived from models.Count('comment'):

from django.db import models

ps = Post.objects.annotate(num_comments=models.Count('comment')).all()
for p in ps:
    print(p.title, p.num_comments)

The output shows that there are two comments in the first post, which has x as its title, and four comments in the second post, which has y as its title:

x 2
y 4

We may add some conditions while counting the related objects. For example, we may wish to count the good comments with 6 or more thumbs up. This can be implemented by adding a filter argument to the Count() function. The filter argument should be a Q() object, which specifies extra SQL expressions.

In this example, comment__num_thumbs_up__gte (separated by double underscores) means that the num_thumbs_up field of the related comment object must be greater than or equal to (gte) 6:

ps = Post.objects \
    .annotate(num_comments=models.Count('comment')) \
        'comment', filter=models.Q(comment__num_thumbs_up__gte=6))) \
for p in ps:
    print(p.title, p.num_comments, p.num_good_comments)

The output prints both num_comments and num_good_comments:

x 2 1
y 4 1

Q() objects accept more than one arguments. If two (or more) arguments are passed, only the objects that pass all tests are counted. For example, if we would like to hide the hidden comments from the output, we can add comment__hidden=False to the Q() objects:

ps = Post.objects \
        'comment', filter=models.Q(comment__hidden=False))) \
        'comment', filter=models.Q(
            comment__hidden=False, comment__num_thumbs_up__gte=6))) \
for p in ps:
    print(p.title, p.num_comments, p.num_good_comments)

The output no longer counts the hidden comment in post y:

x 2 1
y 3 0

Last, we can list the minimum (Min()), average (Avg()), and maximum (Max()) value of the num_thumbs_up field with:

ps = Post.objects \
    .annotate(min_thumbs_up=models.Min('comment__num_thumbs_up')) \
    .annotate(avg_thumbs_up=models.Avg('comment__num_thumbs_up')) \
    .annotate(max_thumbs_up=models.Max('comment__num_thumbs_up')) \
for p in ps:
    print(p.title, p.min_thumbs_up, p.avg_thumbs_up, p.max_thumbs_up)

The output shows that all comments in the first post have at least 5 thumbs up, 7.5 thumbs up on average, and at most 10 thumbs up and so forth:

x 5 7.5 10
y 1 4.0 7