[storm] Do aggregate queries in storm imply an ORDER BY?

[James Henstridge]

On Sat, Jun 20, 2009 at 3:08 AM, Jason Baker<jbaker at zeomega.com> wrote:
> There's one test in tests/store/base.py that I'm running in to trouble with:
>
>    def test_find_group_by_table(self):
>        result = self.store.find(
>            (Sum(FooValue.value2), Foo), Foo.id == FooValue.foo_id)
>        result.group_by(Foo)
>        foo1 = self.store.get(Foo, 10)
>        foo2 = self.store.get(Foo, 20)
>        self.assertEquals(list(result), [(5, foo1), (16, foo2)])
>
> This test is failing because list(result) looks like this:  [(16,
> foo2), (5, foo1)].  This is the same result, but in a different order.
>  I know at least at the SQL level, a query such as the above wouldn't
> be guaranteed to be in any order.  But is this the case with storm?

A while back, I was working on a randomisation branch for Storm.  This
added a flag that would automatically append "random()" (or whatever
the DB uses) to the end of ORDER BY clauses.  The intended use was to
prevent tests from depending on undefined ordering.

I haven't yet finished the branch, since it generated invalid SQL in a
few cases but I did try turning it on for the Storm test suite itself
and ran into a few problems like the one you've found.  I am inclined
to consider such tests broken rather than say that back ends that fail
on those tests are broken.

I guess it'd be worth trying out my branch again, and at least merge
fixes for the failures it discovers.

James.