Question

I am trying to count daily records for some model, but I would like the count was made only for records with some fk field = xy so I get list with days where there was a new record created but some may return 0.

class SomeModel(models.Model):
    place = models.ForeignKey(Place)
    note = models.TextField()
    time_added = models.DateTimeField()

Say There’s a Place with name=”NewYork”

data = SomeModel.objects.extra({'created': "date(time_added)"}).values('created').annotate(placed_in_ny_count=Count('id'))

This works, but shows all records.. all places.

Tried with filtering, but it does not return days, where there was no record with place.name="NewYork". That’s not what I need.

Answer 1

It looks as though you want to know, for each day on which any object was added, how many of the objects created on that day have a place whose name is New York. (Let me know if I’ve misunderstood.) In SQL that needs an outer join:

SELECT m.id, date(m.time_added) AS created, count(p.id) AS count
  FROM myapp_somemodel AS m
  LEFT OUTER JOIN myapp_place AS p
       ON m.place_id = p.id
       AND p.name = 'New York'
  GROUP BY created

So you can always express this in Django using a raw SQL query:

for o in SomeModel.objects.raw('SELECT ...'):   # query as above
    print 'On {0}, {1} objects were added in New York'.format(o.created, o.count)

Notes:

1. I haven’t tried to work out if this is expressible in Django’s query language; it may be, but as the developers say, the database API is “a shortcut but not necessarily an end-all-be-all.”)

2. The m.id is superfluous in the SQL query, but Django requires that “the primary key … must always be included in a raw query”.

3. You probably don’t want to write the literal 'New York' into your query, so pass a parameter instead: raw('SELECT ... AND p.name = %s ...', [placename]).