types for GenericForeignKey object_pk?

[Brian May]

Hello,

The documentation suggests that object_id associated with GenericForeignKey should be a PositiveIntegerField. To be precise, the documentation says: "For most models, this means a PositiveIntegerField." but doesn't give any more details.

However, this will not work for pk that are not integer. So for the case for django-guardian, where we want be more generic, we use a TextField.

I notice that object_id in django.contrib.admin.models LogEntry is a TextField - although it doesn't use a GenericForeignKey.

The problem with this is when making table joins on Postgresql - postgresql gets upset that the object_id is a string, and a different type on the destination table.

e.g. simplified, we have a db model:

class UserObjectPermissionBase(models.Model):
    content_type = models.ForeignKey(ContentType)
    object_pk = models.CharField(_('object ID'), max_length=255)
    content_object = GenericForeignKey(fk_field='object_pk')

We lookup certain objects from this table. For simplicity, lets assume they *all* point to Group (django.contrib.auth) - i.e. content_type points to the Group content type - as this is where the test case fails. Also we use values, as we only are interested in the object_pk field.

user_obj_perms_queryset = UserObjectPermissionBase.objects.filter(....).values('object_pk')

Then we want to lookup all groups that the results of this queryset point to:

queryset.filter(Q(pk__in=user_obj_perms_queryset))

On sqlite and Mysql this works fine. On Postgresql it fails with a Postgresql error:

======================================================================
ERROR: test_empty_perms_sequence (guardian.testapp.tests.test_shortcuts.GetObjectsForUser)
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/home/brian/tree/debian/upstream/django-guardian/guardian/testapp/tests/test_shortcuts.py", line 474, in test_empty_perms_sequence
    set(objects),
  File "/usr/lib/python2.7/dist-packages/django/db/models/query.py", line 162, in __iter__
    self._fetch_all()
  File "/usr/lib/python2.7/dist-packages/django/db/models/query.py", line 965, in _fetch_all
    self._result_cache = list(self.iterator())
  File "/usr/lib/python2.7/dist-packages/django/db/models/query.py", line 238, in iterator
    results = compiler.execute_sql()
  File "/usr/lib/python2.7/dist-packages/django/db/models/sql/compiler.py", line 840, in execute_sql
    cursor.execute(sql, params)
  File "/usr/lib/python2.7/dist-packages/django/db/backends/utils.py", line 64, in execute
    return self.cursor.execute(sql, params)
  File "/usr/lib/python2.7/dist-packages/django/db/utils.py", line 97, in __exit__
    six.reraise(dj_exc_type, dj_exc_value, traceback)
  File "/usr/lib/python2.7/dist-packages/django/db/backends/utils.py", line 64, in execute
    return self.cursor.execute(sql, params)
ProgrammingError: operator does not exist: integer = character varying
LINE 1: ..."name" FROM "auth_group" WHERE ("auth_group"."id" IN (SELECT...
                                                             ^
HINT:  No operator matches the given name and argument type(s). You might need to add explicit type casts.

The full SQL of the failing query in this case is:

SELECT "auth_group"."id", "auth_group"."name" FROM "auth_group" WHERE ("auth_group"."id" IN (SELECT U0."object_pk" FROM "guardian_userobjectpermission" U0 INNER JOIN "auth_permission" U2 ON ( U0."permission_id" = U2."id" ) WHERE (U0."user_id" = 1 AND U2."content_type_id" = 2)) OR "auth_group"."id" IN (SELECT U0."object_pk" FROM "guardian_groupobjectpermission" U0 INNER JOIN "auth_group" U1 ON ( U0."group_id" = U1."id" ) INNER JOIN "testapp_customuser_groups" U2 ON ( U1."id" = U2."group_id" ) INNER JOIN "auth_permission" U4 ON ( U0."permission_id" = U4."id" ) WHERE (U2."customuser_id" = 1 AND U4."content_type_id" = 2)))

Where I believe the problem is that "auth_group"."id" is integer and U0."object_pk" is a string, and they can't be compared.

... WHERE ... "auth_group"."id" IN (SELECT U0."object_pk" FROM ...) ...

So is this a problem with our code? Maybe GenericForeignKey must have an integer value, which excludes it from non-integer pks? Or is this a problem with Django?

Would be interested in feedback.

Thanks