Enable absolute imports for modules using Queue.

[jcrobak]

When running on Linux with code on a case-insensitive file system,
imports of the Queue module fail because python resolves the
wrong file (It is trying to use a relative import of queue.py in
the kafka directory). This change forces absolute imports via PEP328.

[jcrobak]

ping?

[jcrobak]

Hi @mumrah - are you willing to consider this PR? Or do I need to come up with a different solution?

[mumrah]

@jcrobak I'm fine with this change, but I don't have a local linux machine to test it on. Have you run the integration tests with this change?

[mumrah]

Also, sorry for the long delay

[jcrobak]

Thanks @mumrah. I have run the unit tests on mac and linux. It seems as though they were also run as part of the travis ci build, right? https://travis-ci.org/mumrah/kafka-python/builds/12852012

[mumrah]

Yea, TravisCI built this PR fine, but the problem you describe wasn't being shown previously - so can't really say if it worked just from Travis.

To clarify, the unit tests are currently disabled and the integration tests that actually start up ZooKeeper and Kafka are the ones I'm interested in. Are these the tests you ran?

Thanks

[jcrobak]

Yea, TravisCI built this PR fine, but the problem you describe wasn't being shown previously - so can't really say if it worked just from Travis.

Gotcha. Yes, I've verified that this fixes the issue I've seen.

To clarify, the unit tests are currently disabled and the integration tests that actually start up ZooKeeper and Kafka are the ones I'm interested in. Are these the tests you ran?

Yeah, that seems to be which runs when I execute tox. The tests do fail in python 2.6.8, but they pass in python 2.7.5. Output looks like this: https://gist.github.com/jcrobak/7454409

[mumrah]

That looks like the mostly broken unit tests. Take at look at how Travis runs the integration tests and make sure things pass

[jcrobak]

Sorry for the delay. I can't get the integration tests to run via python -m test.test_integration even though they run via tox. I've tried several times both on master and on my branch. There seems to be a race condition that causes test_switch_leader (connection reset by peer), test_switch_leader_async (stack overflow), and TestFailover.tearDownClass ('NoneType' object has no attribute 'stop') to error out.

But the fact that local tox and travis builds are both passing seems to be a pretty strong indication that my code didn't introduce an import-related regression.

[kevinastone]

👍

Ran into this issue.