caspermeijn/jord-rs

Jord is a Rust library that handles geographic math, the kind of calculations you need when working with positions on Earth. Think of it as a toolkit for answering questions like: how far apart are these two points? If I travel 1,000 meters at a bearing of 200 degrees, where do I end up? What's the closest point where two moving objects will meet? It converts between different ways of representing locations (like GPS coordinates versus 3D Earth-centered vectors), measures distances along the Earth's curved surface, and works with areas and paths drawn on the globe. The library supports two models of the Earth: a simple sphere and a more accurate ellipsoid (the WGS-84 shape that GPS systems use, which accounts for Earth being slightly squished at the poles). It uses a mathematical approach called "n-vector" representation, which avoids problems that traditional latitude/longitude math runs into near the poles and when crossing the dateline, places where longitude becomes ambiguous or signs flip. Who would use this? Anyone building navigation, mapping, or tracking software. A drone company might use it to calculate intercept courses. A logistics app could compute great-circle distances between warehouses. A maritime or aviation tool might rely on it for bearing calculations across oceans. Game developers building worlds on spherical planets could use it for path-finding and area containment checks. The project is notably rigorous about correctness. It includes solutions to all ten standard test problems from NavLab, a well-known reference site for geographic calculations, so you can verify the library produces accurate results against established benchmarks. The algorithms are grounded in published academic literature on spherical geometry and navigation, not just quick approximations. The code is written in Rust, which means it's designed for performance and safety, useful if you're running these calculations at scale or in systems where errors could have real-world consequences like navigation errors.