Skip to content
← All work

Open source

Polars

Giving back to a tool I use every day.

Role
Open-source contributor
Hardest problem
Understanding a large codebase and making changes without breaking existing behaviour.
Result
Merged fixes, features, and documentation that now ship to everyone who installs Polars.

Why Polars

Polars is a DataFrame library built in Rust on top of Apache Arrow, with a lazy query engine that plans and optimises work before running it. It’s the tool I reach for first, whether I’m modelling loan performance at work or digging into a dataset of my own in the evening.

Use a library that much and you start to notice its rough edges. Eventually I wanted to help smooth them out.

  1. Scan

    Read dataset lazily, Parquet for example

  2. Optimise

    Push filters and column choices down

  3. Execute

    In parallel, on Arrow memory

  4. DataFrame

    Only what was asked for

Polars’ lazy engine plans and optimises a whole query before running it, reducing memory usage and improving performance.

What I contribute

My contributions tend to come from my own use, wherever a real workload ran into something that could be better:

  • Fixing bugs found in day-to-day analysis, with small, focused reproductions.
  • Improving the behaviour and performance of expressions I lean on heavily.
  • Clarifying documentation and examples so the next person has an easier time.

I’ve also spent time in the community forums talking directly to lead developers, whose time and advice I’m grateful for.

What I took from it

Working in a large, well-run Rust codebase has taught me a lot: how good code review works, how much testing is enough, and how a project decides what belongs in its core.

Open source raises the bar. A change you make will run on machines you’ll never see, for people you’ll never meet, so it has to be correct and easy for the next maintainer to follow.