Django (1.2 beta) will reset the database(s) between every test that runs, meaning each test runs on an empty DB. However, the database(s) are not flushed. One of the effects of flushing the database is the auto_increment counters are reset.
Consider a test which pulls data out of the database by primary key:
class ChangeLogTest(django.test.TestCase):
def test_one(self):
do_something_which_creates_two_log_entries()
log = LogEntry.objects.get(id=1)
assert_log_entry_correct(log)
log = LogEntry.objects.get(id=2)
assert_log_entry_correct(log)
This will pass because only two log entries were ever created. However, if another test is added to ChangeLogTest and it happens to run before test_one, the primary keys of the log entries are no longer 1 and 2, they might be 2 and 3. Now test_one fails.
This is actually a two part question:
- Is it possible to force
./manage.py testto flush the database between each test case? - Since Django doesn’t flush the DB between each test by default, maybe there is a good reason. Does anyone know?
Have a look on the implementation of the command of django.core.management.commands.flush.py.
You can call the flush command from inside your test call (maybe in TestCase.setUp):
Yes there is: Speed up. Flushing and reloading many data from json takes a while…
Maybe you should have a look on TransactionTestCase