django.db.models.sql.query.BaseQuery.get_count() works wrong with GROUP BY

[Maxim Syabro]

Query produced by get_count is

SELECT COUNT(*) FROM table

which returned one row, f.e. (5) . Ok, it's works.
Lets add an GROUP BY statement:

SELECT COUNT(*) FROM table GROUP BY field

It's returned an list of row with count of grouped rows, f.e. (4, 23, 1, 1, 5)
In ge_count we see

data = obj.execute_sql(SINGLE)
if not data:
    return 0
number = data[0]

So it's return only first value of the list instead of list's width. With our example it's 5.
Solutions are:

1. Return cursor's row_count
2. Modify query to

   SELECT (*) FROM (SELECT COUNT(*) FROM table GROUP BY field) AS t
   

[Malcolm Tredinnick]

This is true, but it's not currently intended to work with group_by. This is noted in the comment where the count query is set up. Making this work is part of the broader "adding aggregation" support to Django scheduled for 1.1. You'll want to follow #3566 for progress on that if you're interested.

Closing as wontfix, since the limitation is intentional, and the API you're using is not public API (there's no public group_by API at the moment) and we're developing the public API in #3566.

[bcurtu]

I know this bug/feature is fixed/implemented in django 1.1, however, since I can not move from 1.0 I have patched this problem:

Index: /home/bcm/wsp/www/django/db/models/sql/query.py
===================================================================
--- /home/bcm/wsp/www/django/db/models/sql/query.py	(revision 2833)
+++ /home/bcm/wsp/www/django/db/models/sql/query.py	(working copy)
@@ -228,6 +228,7 @@
         obj.select_related = False
         obj.related_select_cols = []
         obj.related_select_fields = []
+        obj.group_by=[] 
         if len(obj.select) > 1:
             obj = self.clone(CountQuery, _query=obj, where=self.where_class(),
                     distinct=False)