Create PR activity report

[nchammas]

It would be nice to have some sort of report that shows basic PR-related metrics about the health of the project.

I imagine y'all have some ideas about this already. I'd suggest the following metrics for starters, which could perhaps be calculated over a configurable period of time:

• Number of PRs opened, merged, and closed without merging
• Number of stale, hanging, and abandoned PRs (as described in Create dashboard for viewing stale, hanging, and abandoned PRs #1)
• Average time to first reviewer comment
• Average time to PR close (either merged or not)

[JoshRosen]

I like these suggestions. This shouldn't be too hard to implement; we could probably put these metrics at the top of the main page to provide an "at-a-glance" activity report, similar to the GitHub "pulse" screen.

I'm not sure how useful "Number of PRs reviewed per reviewer" will be, though, since some folks may leave super-detailed commentary on only a handful of PRs, while other folks might comment more broadly, and I don't want to "gamify" the review process.

[nchammas]

Yeah, I added that "Number of PRs reviewed per reviewer" metric in specifically as a gamification thing, but that could easily become counter-productive. I've taken it out.

[JoshRosen]

Unfortunately, the GitHub API won't distinguish between merged and closed PRs because we close them from squashed commits that weren't part of the PR. Because we've been closing some old PRs by merging dummy commits, it might be tricky to auto-detect whether a PR was merged or not.

There are a few other approaches we could use, though:

• Check whether the PR author matches the close commit author.
• Check whether the close commit's message says that it closed the PR. The dev/merge_spark_pr.py script uses a consistent commit message convention.

[nchammas]

Some of the stats proposed here are now available at: http://issuestats.com/github/apache/spark

[nchammas]

Well, it looks like issuestats.com went the way of the dodo.

More to the point: Is the PR dashboard still actively used?

In any case, I'll close this since it's clearly not an essential feature.