On Tue, Aug 14, 2018 at 7:47 AM Jeff King <peff@xxxxxxxx> wrote: > > One problem there is that the prefixes are ambiguous (e.g., Jacob Keller > shares with me, and I think at least one other over the years). You > could look at the author of the tip commit, but that's not always right > (and in fact, counting just merged topics misses bug-fixes that get > applied directly on top of other people's topics). And of course there's > the notion that "topic" might be a documentation typo fix, or it might > be the entire range-diff program. One could take all topics and see if you have at least one commit in there. But that would mostly measure how much of an allrounder you are in the code base (e.g. bug or style fixes such as Ramsay's "please squash this" would be in many topics if not squashed). There are other players who are very deep into one area of the code, and probably have fewer series. > I think "surviving lines" is another interesting metric, though it also > has flaws (if I s/sha1/oid/ on your line, it becomes my line; even > though my change is useful and should be counted, it's probably not as > important as whatever the code was doing in the first place). I wonder if we could measure the entropy added to the code base instead. A patch that does s/sha1/oid/ (or introduction of a repository argument) might compress very well, whereas new code might not compress very well. ;-)
