How to group by AND aggregate with Django

Guy Bowden picture Guy Bowden · Nov 15, 2012 · Viewed 68k times · Source

I have a fairly simple query I'd like to make via the ORM, but can't figure that out..

I have three models:

Location (a place), Attribute (an attribute a place might have), and Rating (a M2M 'through' model that also contains a score field)

I want to pick some important attributes and be able to rank my locations by those attributes - i.e. higher total score over all selected attributes = better.

I can use the following SQL to get what I want:

select location_id, sum(score) 
    from locations_rating 
    where attribute_id in (1,2,3) 
    group by location_id order by sum desc;

which returns

 location_id | sum 
-------------+-----
          21 |  12
           3 |  11

The closest I can get with the ORM is:

Rating.objects.filter(
    attribute__in=attributes).annotate(
    acount=Count('location')).aggregate(Sum('score'))

Which returns

{'score__sum': 23}

i.e. the sum of all, not grouped by location.

Any way around this? I could execute the SQL manually, but would rather go via the ORM to keep things consistent.

Thanks

Answer

Aamir Adnan picture Aamir Adnan · Nov 15, 2012

Try this:

Rating.objects.filter(attribute__in=attributes) \
    .values('location') \
    .annotate(score = Sum('score')) \
    .order_by('-score')