filtering over an Sum on a BooleanField uses BooleanField's get_db_prep_lookup on the value

[Pavel Anossov]

With a model like this:

class Test(models.Model):
    active = models.BooleanField()
    group = models.IntegerField()

this queryset:

Test.objects.values('group').annotate(s=Sum('active')).filter(s__gt=123456)

As well as this one, and one with any other value

Test.objects.values('group').annotate(s=Sum('active')).filter(s__gt='wharrgabl')

Generates this SQL and params

SELECT "test_test"."group", SUM("test_test"."active") AS "s" FROM "test_test" GROUP BY "test_test"."group" HAVING SUM("test_test"."active") > ? LIMIT 21

(True,)

That is, any value is converted to bool by BooleanField's get_db_prep_lookup called from ​here.

[Aymeric Augustin]

In general, aggregation only works when the result of the aggregate has the same type as the original value.

Arithmetic on booleans doesn't make sense, even though Python attemps to perform it:

>>> bool(True - True)
False
>>> bool(True / False)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
ZeroDivisionError: integer division or modulo by zero

Specifically, your code won't work on databases where true and false are stored as 't' and 'f'.

[Pavel Anossov]

This is just somewhat inconsistent — the docs say only numeric fields are allowed in Avg, but it is not enforced, and Avg works on BooleanFields because the result is always a FloatField. A mention in the Sum docs or an exception instead of undefined behaviour would be nice.

[Aymeric Augustin]

It's an accident of duck-typing that Avg works on boolean fields, and as said in my first comment, it certainly doesn't work on all databases.