Bugzilla – Full Text Bug Listing
| Summary: | [pmi] Commit Activity should use Git author instead of committer (or in addition) | ||||||
|---|---|---|---|---|---|---|---|
| Product: | Community | Reporter: | Robin Stocker <robin> | ||||
| Component: | Dashboard | Assignee: | Portal Bugzilla Dummy Inbox <portal-inbox> | ||||
| Status: | RESOLVED FIXED | QA Contact: | |||||
| Severity: | normal | ||||||
| Priority: | P3 | CC: | daniel_megert, matthias.sohn, wayne.beaton | ||||
| Version: | unspecified | ||||||
| Target Milestone: | 2014-Q1 | ||||||
| Hardware: | All | ||||||
| OS: | All | ||||||
| Whiteboard: | |||||||
| Bug Depends on: | 370151 | ||||||
| Bug Blocks: | |||||||
| Attachments: |
|
||||||
|
Description
Robin Stocker
thanks for filing this For most changes the author did the heavy lifting and if the committer is different he rebased or did some minor additional tweaks or moved the commit to another branch. I got inspired last night and wrote a routine to capture the author data. It's running now. The new routine scans the Git repos and--for each commit--extracts the author, as well as any "Also-by" [1] entries. Once it's done running (and I have some data to play with), I'll rework the charts to use the new data. Initially, I'm going to pull the author names as they've been provided to Git. This may result in some authors having multiple entries (e.g. some commits moved over from CVS may have committer ids instead of names, or something). If there is a need, I think that we can be a bit more clever and match Git authors up to committer records to consolidate. But I think I've just started mumbling... The new implementation is a bit more resilient than the old; parallel and incremental scans are possible, meaning that we may be able to run the scan more than once a week. [1]http://wiki.eclipse.org/Development_Resources/Contributing_via_Git#The_Commit_Record The "Individual Commit Activity" chart is now based on the author of Git commits. As mentioned in comment #2, it includes "Also-by" entries, so any single commit may actually be claimed by one or more individuals. As before, the chart is based on the last three months of commits; it may be interesting to include a project lifetime chart as well (maybe somebody can open a new bug if that's considered valuable). Note that the charts currently display the name of the author as it is provided in the commit record. This means that some authors commits are split (e.g. "Tim Fox" and "purplefox" on the Vert.x project are the same person). Note also that it is based on absolute commit activity and will include merge commits that do not include any intellectual property. I am already capturing the data that I need to count only those commits that actually have file changes. I'm not sure how big an impact this will have on the charts, but it is probably worth investigating. This is experimental for now as it is based on some new code for gathering commit information. thanks, this looks more reasonable (checked for jgit and egit), we should do the same for the other chart "Organization Commit Activity" Thanks! +1 for also changing Organization Commit Activity. (In reply to Matthias Sohn from comment #4) > thanks, this looks more reasonable (checked for jgit and egit), we should do > the same for the other chart "Organization Commit Activity" I agree. That's going to take a little more work to map committers to organizations (it's relatively simple to just extract the author information from the commits). I've laid some of the groundwork already; I'll keep poking at it. Created attachment 241235 [details]
Slight appearance change
Organization commit activity now takes the authors into account.
You'll notice a small change in how it appears. Commits associated with "[Contribution]" come from a non-committer; those associated with "[Unaffiliated]" come from a committer who is not affiliated with a member company. Only member companies are shown (this part isn't new).
We're done here. All charts in the PMI use the author field. |