(adding spectrometer-dev so we can find this discussion later)
On Tue, Jul 26, 2016 at 4:20 AM, Stephen Kitt <[email protected]> wrote: > On Mon, 25 Jul 2016 11:01:11 -0400 > Colin Dixon <[email protected]> wrote: > > Some of you might have noticed that Spectrometer has been given some > > new life: > > http://spectrometer.opendaylight.org/ > > > > Today, Thanh, Vasu, TomN, and a few others will do a demo of what we > > have and how it works. > > This is all quite nifty, especially given that we can have (modulo > development effort of course) specific, ODL-relevant views of the data. > > It occurs to me that the following may be useful (and I couldn't find > the info in Spectrometer): > * a breakdown of commit sizes (so we can see which projects merge huge > patches) > This should be possible to add. We're still in the early stages where there's lots of features we want to add but only have limited resources so can't add everything at once :) > * distinction of author v. committer (to measure review activity — how > many committers merge patches in a project?); this may be complicated > by the fact that projects have different merge strategies (rebasing > projects have the reviewer as the patch committer, merging projects > have the reviewer as the merge author) > Yes this is something we should be able to add in the future. My current idea is that: author = contributor (the person in the author field inside of a git commit object). I'm trying to use the word "contributor" in relevant places in the UI. committer is a committer on the project, we can probably query ldap or some metadata file in a project repo (maybe PROJECT_INFO.yaml) and identify certain contributors who have committer powers. Additional analyzing Gerrit information is totally in the roadmap. > * a synthetic, global view of lines/author/project (to quickly identify > projects with a high bus factor) — this is already easy to find per > project, but a global view would perhaps be quite handy > Sounds like a great idea. > * measuring "blame" as well as insertions/deletions — i.e., who wrote > the code which is currently in the project (obviously this is rather > hand-wavy, since e.g. reformatting code will change credit without > any code work being done) > I'm not sure how to get this detail yet and I suspect it will probably be very performance intensive as it will probably have to analyze every line in every file for this data, we recently discovered that even showing the insertions/deletions is one of the causes of spectrometer's performance issues. But I can definitely see value in maybe having something that shows you for a commit object, the lines of code being changed who wrote the previous versions of them. > * some different measure of deletions in the LOC charts (e.g. Red Hat > has -114,615 LOC in the project now) > I agree and think the whole LOC measurement code needs to be re-evaluated. I think aggregating all of the lines added and deleted is probably not the right way to display this data and I haven't found it useful personally either. Does Spectrometer pull in Git data only, or does it have all the Gerrit > data? When I've done commit analyses in the past, I've found it useful > to correlate Gerrit users with email addresses in commits — this > sometimes simplifies the task of merging contributors who use multiple > email addresses (or whose email address changes). > Currently it gets the project list from Gerrit, and everything else comes from Git. We've been focusing on Git as we thought the most important data to get shown that would have the most impact is the pie charts displaying commit information from various projects. We definitely have Gerrit in the roadmap and other sources such as: - mailing lists - jira - sonar - jenkins - bugzilla Basically whatever we can poll data from we'll try to get it and provide an API for people to query but these details will be in the future as we're currently focusing on the Git side of things before we move on. Regards, Thanh
_______________________________________________ Discuss mailing list [email protected] https://lists.opendaylight.org/mailman/listinfo/discuss
