There are three things I hate about Python. The first is that I want to know the data types and it has been difficult for me to navigate between dictionaries, and their equivalents in yaml/json and how pandas handles them vs numpy vs pyspark. The second thing is that I hade how quickly libraries and dependencies get out of synch and I can’t just move binaries around to be executed elsewhere without headaches, like I can in Rust. The third thing I hate about Python is that it’s interpreted and plays fast and loose with things I’d rather be sure of. The final thing is that I’m so used to hating Python (having chosen Ruby back in 2011) that I hate Python.
Since the beginning of this month, I have decided to take Data Science seriously because I also hate K8s and I don’t want to be just a DevOps monkey. So I’m hunkering down and taking a 50 hour course. Not the MIT 600.2X but something approaching that. I finally found somebody on Udemy whose English I can parse transparently. It makes a huge difference. It’s Jose from Pierian. So now I am in that realm where I start to realize how stupid I am for not being stupid. Which is to say I’m hearing finally, in way that gets through my thick skull, what I needed to know to make sense of stuff I have mostly ignored. IE the basics for people who would take Python deadly seriously.
I should also say that the other thing that kept me going deep into Python had a lot to do with what I had to go through vis a vis library mess to get Jupyter notebooks to work properly on Mac. vu made that work in just a few moments. Now venv are truly disposable in a way they weren’t with poetry.
So, yeah. Right now I’m discovering exactly how stupid I am in Python despite the fact that I’ve been maintaining a lot of code that I didn’t write, and getting angry about.
The thing that did it was a video by ArjanCodes that introduced me to uv. UV replaces poetry, immediately. Now this handles objection #2 and attenuates objection #4. But also since I’m turning away from big data at scale and towards fast, distributed OLAP and simplicity, I have reasons to disambiguate all the stuff that seemed beneath consideration and was of no interest to my old customer base. I should come out of the other side of this feeling pretty good about Python, without going the Mojo route.
So the company is Astral. I’m happy about it.
Also, since I’ve gone DuckDB, I’m going to start publishing models and code I’ve been working on from way back in the day. Here’s a teaser. These are the customers and projects I’ve been a part of going back 25 years. Can you believe it?
And some of those are high level directories. There’s also another section of about 20 large projects that I have given codenames because of their proprietary nature. It will take a lot of work to convert that out to something that could go public, but in the above cases, most of these are more than seven years old.