Question

I am building an application that tracks karma that users give to each other. I want to be able to build a leaderboard over a rolling 7-day period and a list of people who have given you the most karma over a rolling 7-day period. I have come up the following model:

class KarmaAction(models.Model):
    user = models.ForeignKey(User)
    sender = models.ForeignKey(User)
    karma = models.IntegerField()
    created_at = models.IntegerField()

First off, will the above model scale well as the number of users grows? I am thinking of deleting any rows that are over 7 days old. Secondly, I want to know how I can create a list of senders with the corresponding sum of the karma they have given the last few days. Maybe it would look something like this:

supporters = {'User 1': 14, 'User 2': 12, 'User 3': 7, 'User 4': 2}

Is there any way to do this with the Django ORM?

Answer 1

The model seems reasonable to me; I don’t really see another way to do it. You need all of the information in there to be able to fulfill your requirements, anyway. (Though created_at should probably be a DateTimeField?) If it ends up too slow, you’ll probably just have to add aggressive caching.

You can do something like (untested, but I think it should work)

TIME_RANGE = datetime.timedelta(days=7)
now = datetime.datetime.now()

KarmaAction.objects.filter(created_at__gte=now - TIME_RANGE) \
                   .filter(user=recepient_user) \
                   .values('sender') \
                   .annotate(total_sent=Sum('karma')) \
                   .order_by('total_sent')

which will use the values to group by unique senders and then sum their karma values.

I added the order_by to make sure that if there’s a default ordering, it gets cleared and doesn’t interfere with the grouping. It also makes it easy to just get the top n senders of karma, though, by just adding e.g. [:5] to the end — if you make it -total_sent, of course :).

I don’t think the filter causes this problem, but I’m not positive without testing.