Better testability about mysql

[trowski]

Processor should accept a connected Socket and use that for communication rather than creating it in connect(). Tests could then control the data on a mocked socket to test exactly the scenarios you suggested.

This package suffers a bit because it was originally written using watchers and all the horrible flow control logic that goes with that. I believe it's now at a point where I could do some further refactoring to improve testability as you mentioned, bringing it up to the standards of the other Amp packages.

[bwoebi]

Also: originally the Connection and Processor was a single monolith, until I realized there were some circular references.

I'm totally in favor of completely splitting the parsing part out and testing the processing of parsed data instead of having to test the protocol directly against the live mysql instance. (which is fine for integration, but can only test a small subset of the actual functionality)