Right now it seems that the plugin walks all files to get modified filenames but this is slow for large projects. As an alternative (I don't know if this would fully work) it would be faster to just check the diff using whatever version control system was in use. It's pretty easy to detect the VC based on hidden dot files.

Here is an example of finding all modified files using git:

git diff --name-status | grep .py | awk '{ print $2 }'
git diff --cached --name-status | grep .py | awk '$1 != "D" { print $2 }'

This would ignore deleted files (although, hmm, maybe that should trigger tests)

"When X changes, always run test suite Y. X might not be Python; it might be a BDD script."

If I run nosetests with an individual test file, bazinga seems to produce empty .nosehashes files, and will not do anything to prevent rerunning an unchanged test.

If I provide multiple files, the same thing happens.

When providing whole directories, I seem to get some entries into the .nosehashes files, but far fewer than I would expect, and far fewer than snakefood finds when running it individually.

So far, I couldn't get it to work, maybe a debug option would be nice, which allows to print information about the changed dependencies which caused a test to run.

... when determining whether a test should be run or not.

For instance, in gitdb:

cd gitdb
nosetests-2.6 --with-bazinga test/

Runs all the tests as I would expect it.

The next time, it runs no test as I would expect it.

Now I change the file test/tes_util.py, and the tests will be rerun, as expected.

Changing a dependency of that file, util.py, will not trigger the test to be rerun though. It doesn't appear to pick up changes in files the test depends on.

See my .nosebazinga file for debugging: https://gist.github.com/1066688
I modified the file to hide portions of the paths, the actual path was absolute, not relative.