TestingTheNose

At the risk of fragmenting the original Wiki, OliBye wants to discuss TestingTheNose on JSPWiki as a real test of its powers.

OliBye believes any Nose we make to detect code smells should be tested on real code. I also believe that the tests should be run in the temporal dimension, i.e. not just on the latest code. This is important because at any given time a snapshot/build may be low on smells. It's the dynamic of smells appearing and being removed over time by refactoring that we need to be concious of.

If we had some tools that would allow for MiningVersionControl, we could do some of the following.

• Given a nose, our VersionControlMining tools should be able to pull out all the code that nose would detect.
• • This would allow us to check by hand that the nose is detecting the smells we think it should.
  • • I'm sure this is cause much debate as to what constitutes a smell, so maybe the results to be publishable in Wiki for discussion.

• Write noses that find refactorings.
• • These real refactorings would serve as good content for the XPD "real code" maxim.

Way off in the future it might be possible to score the noses findings, and highlight area's of the codebase that they haven't detected. This could make is possible to train noses, using various AI techneques, to recognise the code just before the refactoring of a bad smell.

I think DuplicationIsWhereItAllStarts -- NeilThorne