search

ansible-community / stats-collections

RShiny app to display statistics for the Ansible Collections
GNU General Public License v3.0
1 stars 2 forks source link

Abandoned collections #2

Open GregSutcliffe opened 4 years ago

GregSutcliffe commented 4 years ago

from @andersson007:

other ways to identify abandoned collections using any kinds of visualization?

Mostly this comes down to identifying "abandoned" vs "stable" - survival analysis of Issues is probably a good starting point, compared to the collections group as a whole.

Output here could be sensitive? Do we want to "name and shame" the collections we think are abandoned? Or is this data for @gundalow et al. to consume and reach out with?

GregSutcliffe commented 4 years ago

From IRC tonight, this is probably a combination of time-to-close (for issues/prs), time-to-first-comment (possibly just from maintainers), and time-to-release. If all those are trending up for some time, then we likely have a problem.

Andersson007 commented 4 years ago

@GregSutcliffe imo, would be also good to see number of downloads from galaxy for every monitored collection (counters on that dashboard), so we might focus on popular collections. Also, on collection-dedicated stat page (is it exist?:) would be interesting to see:

  1. Graphical representation of releases for a collection (if possible to get pairs version:date from galaxy).
  2. Graphical representation of downloads rate change, so we could track the collection's popularity. Maybe monthly diffs of downloads or something

just ideas that crossed my mind:)

GregSutcliffe commented 4 years ago

Thanks for the feedback @Andersson007! Some thoughts:

1) is already possible, I think - if you pick "Galaxy Releases" from the picker, you get a graphical view of the days between each release. I thought this was more interesting than just having a flat line with Date on the X-Axis. 2) We don't seem to be able to get downloads per version - only a total download count for the collection as a whole (the API call is here if you want to see). So we can't use that for the release graph.

We could track that value over time, yes. I generally avoid the actual value (there's a lot of potential systemic bias from things like CI etc), but the trend could be useful, I guess.