Reliable ways to quantify package usage (specifically core astropy)

[pllim]

From Astropy Coordination Meeting 2021. When the discussions happened, it was about astropy core library, though the solution might be applicable to any number of Python packages in this Project. From the meeting notes:

• Still difficult to measure “usage” of astropy (citations is biased, visits to astropy.org vs docs is biased, etc.)
  • Might be useful to publish some plots from the web analytics for the project to use. This is something people can understand pretty well, esp. Trending.
  • @adrn: Agreed! I wanted to make a dashboard with all of this info, updated automatically, eventually at astropy/astropy-sustainability (initial attempt here)
• @bsipocz: Number of issues or PRs per subpackage is a coarse indicator as well
  • @adrn: I wonder if it would be useful for usage to compare the number of issues raised by people who have submitted PRs to astropy vs. not (per subpackage)
  • @pllim: Like this?

---

Currently, everyone seems to have their own favorite scripts to run for usage stats; e.g.,:

• https://github.com/astropy/astropy-procedures/tree/main/visualizations_demographics (Had not been updated for 3 years, last updated by @keflavich.)
• https://github.com/adrn/astropy-sustainability (As mentioned by @adrn above.)
• https://github.com/bsipocz/astropy-graphs/tree/extend_plots (Original scripts from @astrofrog modified by @bsipocz.)
• https://github.com/pllim/astropy-graphs/tree/lim-edit-2021 (Yet another modified version by @pllim.)
• Scripts by @taldcroft (private communications).

[pllim]

There is also https://github.com/scientific-python/devstats

And @jeffjennings is reviving a dashboard.

[jeffjennings]

I'll have a PR for my version of repo and citation stats in the next day or two.

[jeffjennings]

The PR I mentioned in my last message is here - https://github.com/astropy/astropy/pull/16822