btw: Georust has released GeoJSON 1.0: the library has been stable for a long time, but @ikawe recently shipped a significant perf improvement (it was already extremely fast. It is now faster) and we felt that it was time to commit to a stable 1.0 release: https://crates.io/crates/geojson / https://docs.rs/geojson/1.0.0/geojson/

A fun thing I do when I see people being wildly aggressive on social media is to also block them on Github so I never have to worry about accidentally working with them. Yes, this is because it's mostly people who present themselves as software developers (or some adjacent bullshit job) on social media. Yes, this has been prompted by yet more wildly inappropriate comments directed at the Harfbuzz devs.

My most recent delightful jj insight: if I want to squash changes to a specific file (or set of files) made in a range of one or more commits into a different (usually earlier) target commit, it’s completely trivial (jj squash -f abc::xyz -t foo filename). The mere idea of attempting this in git would make me giggle in nervous terror.

Do I have any European followers who work with and / or run environmental sensor networks? (size not terribly important, but probably < 1000 stations / sensors). If this is you say hello: I'd love to talk about a PoC I'm working on.

Herbie (https://herbie.uwplse.org) was immensely useful to me when I was writing my first Rust crate in 2015, via the rust-herbie-lint plugin. But that plugin architecture was deprecated quickly, and it only worked with one specific Nightly compiler version. Nearly 11 years later, I've finally managed to resurrect it as a Dylint plugin: https://github.com/urschrei/herbie-lint.

I want to use Automerge for something (replacing Firebase real-time sync), but it looks like I’m a bit too early because there’s like one auth-enabled Dockerised sync server implementation available and I don’t like the look of it. Hm.

If you use Claude Code, you've been able to use it to search your local Zotero library and analyse the results using my pyzotero API client and a skill I wrote for it. Now, that skill can also query Semantic Scholar for gap analysis suggestions and related papers: https://github.com/urschrei/zotero_search_skill.

Modern computers are fast: using FlatGeoBuf as the on-disk format, georust’s new Voronoi module can generate and save a Voronoi diagram from all 2,693,630 valid UK postcode centroids in ~10 seconds without jumping through any code optimisation hoops. The actual Voronoi generation takes just over 5s. That's single-core performance on an M2 Pro, and includes round-tripping the coordinates from WGS84 -> OSGB36 -> WGS84. If you exclude the projection work, it’s well under 10s total.