hacky pass at uuid sorting

[kurtjx]

An attempt at addressing #217

This seems to work but I'm not all-together in love with it. One major deal breaker issue, if you expect your column names to actually be DateTypes rather than time uuids, you will get time uuids back and be massively confused. I'm not sure how to address that without making the stubs fully aware of the column family definition.

So this is really not a pull request, but an ask for better ideas I guess. Thanks.

[thobbs]

Why not just do:

items = sorted(items, key=lambda uuid: uuid.time)

when the comparator is TimeUUID?

[kurtjx]

I didn't know you can do that with uuid objects - that is better. However, I don't think the mock cassa will even know what the comparator type is in the current implementation. We could check the type() on items i guess but still seems pretty hacky.

[thobbs]

I'm okay with it being somewhat hacky just for stubs purposes. How about checking to see if the first column is an instance of UUID and has a version attribute with value 1?

[kurtjx]

Sorry for the slow turn-around. I think this is what we want more or less. Embarrassingly I'm having a hard time getting the tests to run on my dev box (actually getting a local cassandra to run).

[thobbs]

Thanks! I think that looks pretty good.