altair-upset: The Evolution of UpSet plots in Altair
Altair is the only plotting library that I’ve felt like loved me back. Maybe ggplot2 would love me back as well, but it’s locked away in the tower of R.
I went on a bit of a lark when I stumbled upon the original upset-altair-notebook from HMS-DBMI. It was just a Jupyter Notebook, and only worked with Altair 4, which met none of my requirements.
The original notebook was great, but I needed something I could quickly pip install and use across projects. Plus, I needed to use Altair 5 for marimo, I wanted to future-proof this tool for the community.
First step was packaging. I’m a big fan of not reinventing the wheel, so I kept the core visualization logic but wrapped it in a proper Python package structure. This meant:
Here’s where it gets interesting - I created a snapshot of the Altair 4 functionality before diving into the Altair 5 boss battle. Think of it like saving your game before a major fight - if something goes wrong, you can always roll back to a working state.
Why? Because I’ve learned from enough bioinformatics murder mysteries that breaking changes in dependencies can be a nightmare. I tried to just port the function to Altair 5 and wound up with some really weird functionality that I couldn’t make sense of.
