One issue we have right now is that our "unit" tests really comprise
both unit and functionality (that is end-to-end) tests. We need to
separate those out, and its part of the road map. But we haven't gotten
there yet.
Alan.
Charlie Groves wrote:
I'm using a local, single-node hadoop cluster to run pig's tests as I
work on it. Is this the preferred way to run the tests? I ask
because with this setup it takes several minutes for many of the tests
to complete, which is different than the normal quick unit test cycle
I'm used to.
I have set mapred.map.max.attempts and mapred.reduce.max.attempts to
1, so if I do break something, it fails after one attempt rather than
several. Is there anything else I can tweak in the setup to speed
things up?
Thanks,
Charlie
