There’s no denying the growing movement within the Python community to prefer Rust over C for writing performance-sensitive extension modules. Leading the way there’s polars, a fast multi-threaded DataFrame library I’ve discussed in an earlier post, ruff which runs about 100x faster than the predominant (and painfully sluggish) flake8 linter, and even pydantic is preparing a 2.0 release with a pydantic-core backend written entirely in Rust. To add to this already-impressive corpus of projects, I am proud to annouce the release of dantzig, a Rust-powered linear programming library for Python!

Back in August, I started contributing to polars, an exciting new dataframe library that is modern, memory-efficient, and lightning FAST (check out the benchmarks here). It is under active development so bugs are identified (and resolved) at an incredible pace. One morning, I noticed a user reporting an error with the exponentially-weighted variance computation – the polars result differed significantly from what pandas produced. Intrigued, I began looking for the source of the disparity.