Fix #117; Make sure to prepare begin batch statements.

[nemosupremo]

See discussion @ #117. I haven't tested this yet, so can someone ensure the tests pass?

[0x6e6562]

Does anybody know whether Travis supports multiple versions of Cassandra?

[nemosupremo]

I'm not sure about Travis, but this bug actually affected Cassandra 1.x and 2.x fyi (Cassandra 2 still supports the begin batch ... apply batch language).

[0x6e6562]

I appreciate that @tux21b, @mihasya and yourself @nemothekid have been driving this issue out, so I was just looking over it to see if I could speed things up so that this patch can land. The content of the patch on visual inspection looks fine to me, but I'd like to see how the test coverage for both C* 1.x and 2.x is working out. Also, I'm just coming from the side, so I'd need to pick up the context as well. Hence anything to make the maintainer's life easier is much appreciated - @mihasya did you have some tests that you wanted to run this patch against?

[tux21b]

LGTM. I think the patch looks good in its current state (we can add additional tests later, but there is already one and it is definitely an improvement). Many thanks!

[0x6e6562]

Thanks for taking this one @tux21b - that's saved me having to find my way into this one :-)

[nemosupremo]

Part of this PR is a test that @mihasya wrote (d3ed3a1). In any case I don't think the CQL API for begin batch is any different on Cassandra 1.x and Cassandra 2.x, I think calling this Cassandra 1 bug is incorrect. (If you look at the commit list and travis build, my first fix had a mistake which didn't actually prepare the statement, which caused the build to fail in Travis). So as it stands now begin batch is broken in gocql on all versions of Cassandra.

Now I believe you do have a cause for worry though. In our production cluster (where we also make heavy use of begin batch) we are still on Cassandra 1.2 and I believe quite a number of cassandra users haven't made the switch to 2.0, so its probably wise to make sure we don't break anything in 1.2. However according to the docs I can't find any mention that any feature that is supported in Cassandra 1.2 is implemented any differently in 2.0, so I'm not sure if blocking on Travis for multiple version testing is a good use of time.

[0x6e6562]

@nemothekid Thanks for taking the time to put this level of detail into the conversation - should we run into any related issues, at least we have some kind of documented history on the subject matter.

[mihasya]

👍 thanks for the patch @nemothekid