Update the CompositeKeyTest to use a populated schema.

[xrash]

I was facing a bug while paginating a Doctrine entity that had a composite primary key and getting this exception:

Single id is not allowed on composite primary key in entity [...]

After poking in the code, I actually found that the exception was being thrown by Doctrine, and there was already a test for that. To my amusement, the test was passing. I tried to do in my code exactly what the test was doing, but no dice.

So I poked a little bit more and discovered that the CompositeKeyTest was running with an empty schema, and that's why it succeded and my code didn't. It took me to just populate the testing schema with some fixtures for the test to fail just like my code.

After a while, I was able to make it work using some crazy hint which I am not sure if it works generally, but worked for me.

Can you guys help to check if those changes make sense at all?

Thanks!

[pilot]

so if run test with $query->setHint(Query::HINT_CUSTOM_OUTPUT_WALKER, false); and populate fixtures we should get Single id is not allowed on composite primary key in entity [...] ?

[xrash]

I don't know if you "should" get that in the sense of expected behavior, but that's what is happening.

[lunetics]

stumbled across the same problem, what did you do to make it work @xrash ?

[xrash]

@lunetics Try using these options:

$query->setHint('knp_paginator.count', $count);
$query->setHint(UsesPaginator::HINT_FETCH_JOIN_COLLECTION, false);
$options = array('wrap-queries' => true);
$result = $paginator->paginate($query, $page, $limit, $options);

[lunetics]

hey @xrash thanks. that finally works with wrap-queries.

I Think the problem lies, that even if you supply the "knp_paginator.count", the doctrine pagintor still has a 0 count and "tries" to build a count query (which fails because of composite primary key).

But i'm not so deep into that, so i think maybe someone should look over that.

[lunetics]

Just saw, even when supplying a count, the doctrine count query is initiated!

[xrash]

@lunetics, that's a nice reasoning.

It looks like the behavior when dealing with composite keys is actually documented here, so it's more like a "lack of feature" instead of a bug. However, I think that the solution proposed in the documentation does not work, so we should also update the docs and/or implement this feature.

I will try to have a look on that soon enough, but still this PR makes sense alone by itself (in the current state of things) and the docs are in another repo. Also, this PR is kinda like dead =p.