What programming/technical projects have you been working on?

I'm way behind everyone here, but I began my first semester of programming classes. First time back in school since 2016-ish when I was a PhD student studying number theory.

I did small bits of coding for school (Matlab, R, lots of Tex), but never used any of the common development languages. I aways wanted to learn, and this year I needed a means to access a student scholarship at the college where I teach to help pay for my kid's daycare.

So far I've compiled my first few baby programs in C++ in visual studio, solved dozens of problems in zybooks labs, and started to grasp the syntax a bit. It's all moving very slowly, and I'm not sure the zybooks approach is the best for really learning, but I'm trying to save and compile code outside of the zybooks environment for anything that's more than a few lines. Hopeful that we get to some critical thinking stuff soon.

I should be writing documentation for repeatTest but instead I’m improving it in minor ways. I published a release tweaking some API’s and dividing the package up into entrypoints because it seemed like the thing to do for a JSR package. Better to do the breaking API changes now, I suppose, since I don’t have any users yet.

Currently we're deploying k3s at work. Right now I'm setting up my dev cluster - 1 control-plane, 3 worker nodes, 2 storage nodes for longhorn.

I'm using kube-vip + traefik for lb and reverse proxy. I had longhorn deployed but this was using the default root disk on the storage nodes so I'm re-configuring them to use a dedicated disk on each system for storage.

A lot of the work I'm doing is proof of concept / eval to show my team so we can do planning for what production will be like. Namely how we need to change our deployment pipeline since we want to use argocd + rollouts (with analysis in the future)

I still have spare nodes available so I might do a Talos deployment using metallb and nginx since I've seen people have opinions on k3s for production use. Thankfully we're a small team with a small user base internally for our applications so I don't think there's a wrong answer to this as before, we used docker swarm and it was fine.

Yesterday I procrastinated on homework so badly that I made an entire pomodoro timer (trust me bro, it'll help with productivity)

Been trying to set up an LLM on my 10900K 3090 24GB VRAM. Keep running into issues getting PyTorch to compile against the CUDA libraries. Keeps pulling down the CPU version.

Using conda. pip was making me crazy.

I'm trying to simulate a stock market based on historical data so I can experiment with a trading algorithm.

I've been playing around with game dev engines. I don't really intend to output anything worth sharing, but I've had two friends express interest in making a game and have well-formed unique concepts so I thought perhaps I'll pop out a prototype. One friend has talked about their idea for a long time so I thought a prototype might motivate them to make it happen.

They're mainly narrative driven games, so at first I looked at renpy and narrat. One of the projects involved some RPG lite systems, so narrat seemed more appropriate. I didn't like the js approach though and the documentation was lacking. It seems directed at folks new to coding, but also accidentally assumes a basis as an experienced web dev.

So I went over to the newest darling, Godot. It's much more complex, feels a bit like using a factory robot arm to crack open a beer, but I'd prefer to learn something more generalizable. I'm quite pleased with the documentation so far and the demos available out of the box cover a wide range of topics. I've mostly just explored the systems and learning the relationship between scenes and nodes. I did get an extremely rudimentary dialogue system set up that pulls a script from a file, now I'm working on keying the lines to indicate "stage directions" like swapping sprites for different speakers to start. The goal would be to make it so my friend could essentially write the scenes similar to a play script as the dialogue input files.

Battling my way through a data quality project.

Have a script that hits an API to run some jobs caching fresh data processed by code under test to a table to compare with baseline data.

Some new stuff for me: asyncio, aiohttp, Datacompy to compare dataframes, and a touch of Bazel because I'm working in a big ass monorepo now.

I'm trying to learn haskell, I've gone through learn you a haskell, going to do some project euler, then maybe advent of code if I don't think of anything else to do in it. If anybody has recommendations on what sort of things they try and build when beginning to learn a new language I'd be interested, also for any other haskell (or functional in general I guess) learning resources

I created a web portal request system that extends my YTmusic time shifting application I shared here previously.

The program lets users search youtube music and request them PIC1 PIC2. The program then checks my library and any pending requests for duplicates and if not found it adds them to a work queue. When the tape dub server kicks off the worker it has these requests already on the work list and gathers the music albums for the requested tracks. I have this all linked to my plex library to build up a music collection again.

I have the code and dockerfile hosted on my git lab here if anyone wants to mess around with it. Will require the tapebub server running as well to operate.

First time using picocss and was verry happy with it. I've been looking for a basic light weight framework to replace bootstrap for small web programs. Just enough to get stuff on the screen looking 1/2 way decent on desktop and mobile wtihout bogging me down on learning the ins and outs of a frame work.

This week I added some timezone flags for my dateutil wrapper CLI program. It was a bit harder than I expected because my tests were failing when changing the system timezone. I tried using freezegun but it wasn't covering all the system timezone callpoints. In the end I decided to only write tests for the explicit timezone parts and skip testing the localtime parts for now. But I also discovered the pytest-regressions plugin which made it easy to write a bunch of data regression tests without having to copy and paste failing values with correct ones over and over:

@pytest.mark.parametrize("p", [["1970-01-01 00:00:01"], ["--from-unix", "1"]])
@pytest.mark.parametrize("fz", [["-fz", "America/New_York"], ["-fz", "America/Chicago"]])
@pytest.mark.parametrize("tz", [["-tz", "America/New_York"], ["-tz", "America/Chicago"]])
@pytest.mark.parametrize("s", [[], ["-d"], ["-t"]])
@pytest.mark.parametrize("f", [[], ["-TZ"]])
def test_lb_timestamps_tz_utc(data_regression, p, fz, tz, s, f, capsys):
    lb(["timestamps"] + p + fz + tz + s + f)
    captured = capsys.readouterr().out.strip()
    data_regression.check(captured)

The data_regression fixture will create a test output file that should be committed to the repo. It takes in a dict by default but luckily strings also work. You can use --force-regen as in pytest tests/text/test_timestamps.py --force-regen to replace the failing test data with the new output data, otherwise it will keep failing until you fix the code.

I work on a laboratory information management system (LIMS) for my job, which is just a case manager for patients & lab tests. The feature work has been pretty limited, I’ve mostly been doing bug fixes and performance improvements. To get a better understanding why it’s architected the way it is I’m implementing a basic version from scratch using the same framework, SpringBoot.

My version feels a little too basic. It’s just a crud app without the secret sauce that makes commercial LIMS profitable, like their plug and play interfacing with lab machines. I don’t know the first thing about the communication between those instruments and the LIMS at my job, it would be nice to know the scale of the work that connects them.

On the other hand I’ve been making another crud app in Go, not really anything more or less complex. Working on both at the same time highlights everything I dislike about Java. It feels juvenile to contribute to any lang vs lang discourse because I know every tool has its advantages, but I find Java and Spring to be cumbersome. These companies are pretty locked in with the wealth of Java knowledge their senior and staff engineers have, especially when scaling LIMS. I wonder what it would take to implement the secret sauce in Go and how the growing pains would be different.