django-conditional-aggregates 0.3.2

Django aggregate functions which operate conditionally (i.e. generate SQL `CASE` statements)

(Django 1.4 and 1.5 needed an ugly hack but our tests pass, if you find any edge cases please post an issue)

Note: from Django 1.8 on this module is not needed as support is built-in: https://docs.djangoproject.com/en/1.8/ref/models/conditional-expressions/#case

Sometimes you need some conditional logic to decide which related rows to ‘aggregate’ in your aggregation function.

In SQL you can do this with a CASE clause, for example:

SELECT
    stats_stat.campaign_id,
    SUM(
        CASE WHEN (
            stats_stat.stat_type = a
            AND stats_stat.event_type = v
        )
        THEN stats_stat.count
        ELSE 0
        END
    ) AS impressions
FROM stats_stat
GROUP BY stats_stat.campaign_id

Note this is different to doing Django’s normal .filter(...).aggregate(Sum(...)) …what we’re doing is effectively inside the Sum(...) part of the ORM.

I believe these ‘conditional aggregates’ are most (perhaps only) useful when doing a GROUP BY type of query - they allow you to control exactly how the values in the group get aggregated, for example to only sum up rows matching certain criteria.

Usage:

pip install django-conditional-aggregates

from django.db.models import Q
from aggregates.conditional import ConditionalSum

# recreate the SQL example from above in pure Django ORM:
report = (
    Stat.objects
        .values('campaign_id')  # values + annotate => GROUP BY
        .annotate(
            impressions=ConditionalSum(
                'count',
                when=Q(stat_type='a', event_type='v')
            ),
        )
)

Note that standard Django Q objects are used to formulate the CASE WHEN(...) clause. Just like in the rest of the ORM, you can combine them with () | & ~ operators to make a complex query.

ConditionalSum and ConditionalCount aggregate functions are provided. There is also a base class if you need to make your own. The implementation of ConditionalSum is very simple and looks like this:

from aggregates.conditional import ConditionalAggregate, SQLConditionalAggregate

class ConditionalSum(ConditionalAggregate):
    name = 'ConditionalSum'

    class SQLClass(SQLConditionalAggregate):
        sql_function = 'SUM'
        default = 0