Statistics about all packages in General

[tfiers]

from

• 88,\

  the question:

• [ ] "what is 80 percentile package name length?"
  • [ ] filtering out ~zero-user packages..
  • Maybe dependents ≥ 1 \ (I.e. implement

    • 87

  • Could query gh for nstars Ooh, or nclones: that's actual install numbers.

This becomes (again) no longer just pkg graph visualization. Could factor out to another pkg, sure.

You'd need caching, probably. (Timestamped folder somewhere) Some PkgGraph specific stable dir mayb.

This whole thing could be written in JavaScr.. ooh, could keep Julia, but save this in a gh-pages branch like way. Scheduled to run regularly on actions. (Maybe some smart caching exponential backoff like thang to not keep querying gh urls of dead pkgs)

[tfiers]

Estimate of file size (is git / gh as db ok or not): 8000 pkgs

Ah, we don't need to store svg's, as wasm dot exists.

And dep storage can be compressed: only direct deps. Hah what's diff over General then: nothing: you don't need your own db it's already there

So, a React app that, ig, uses the gh file api. (Is that efficient? There's lotsa toml files to query) Maybe a cron Action that converts the reg (dirs of TOML files) to an SQLite db? Small enough to send to browser or no?

BUTT: the killer feature is timeimports integration, and that really is Julia only