Happy December, all.

This week:

• I wrap up my exploration of Posit's Databot LLM
• Talk about the rv and uv package managers

Databot and NHANES Part 3

I wrap up my testing of Databot (the Posit LLM) by evaluating the quality of the code that it generated. For the most part, it does well, but does skip a step in ordering a factor.

It goes to show that you still need a basic mental model of how exploratory data analysis works before you can do serious work with Databot.

As Joe Cheng said in his keynote: "Databot is not a Flotation Device" - it will not do the difficult work for you.

uv and rv: Package Managers of the Future

Julia Silge, in her talk "How I got unstuck with Python" (a good watch) highlighted the importance of package management in reducing frustrations with Python. One of those frustrations is Dependency Hell. If you are not familiar with Dependency Hell - it is that some packages are dependent on different versions of Python, and versions of other libraries and dependencies.

Enter uv: a Rust-based package manager that aims to solve dependency hell for Python users. It uses virtual environments to manage difficult package installations, limiting installs to a project folder. That way, you can have multiple projects that use multiple versions of packages and R and they won't conflict or give you errors.

There is now rv, which is the equivalent for R. In my experience with rv so far, they mirrored most of the structure of uv, but adapted for R. It shows a lot of promise.

If there's more interest in how these package managers work, let me know in the comments and I'll do a little writeup.

Take Care of Yourself

These have been incredibly trying times for all of us, and I feel like we are all running a marathon to stay in place.

Please take time for yourself this season. You're all doing the best you can, so be gentle with yourself.

Best, Ted