Hey, I'm a TA at some CS courses, can this be used to find similar codebases? Maybe it useful to find similar programmers, find possible code copies and/or identify the author of a snippet. Any ideas?
Students code winds up looking like their sources because they copy the control structures because there's only so many sane ways to do those problems without introducing fluff.
At least this is what my friend's experience using plagiarism checking tools was.
Indeed, that's been my experience too; if you assign something "Hello world" level, you're going to get a bunch of 100% correct solutions that look almost the same, and then the other... odd ones which are almost certainly incorrect in some way.
It's only when you assign something far more nontrivial, and with many correct ways to do it, that you will get more diverse solutions; and even then, the chances of the correct solutions being very similar are high.
I think it's promising, but to evaluate it against existing solutions and make it production ready is probably still a good chunk of work.
For simple code similarity there are Moss, JPlag and Sherlock (linked in the paper )