Best practice for installing custom Python packages in LinuxCNC?

 
Hi all,

I’ve developed a custom Python package — and not just a loose .py file, but a real multi-module package, complete with pyproject.toml, unit tests, internal structure, and even a __str__ for logging prettiness. Yes, I know, it’s total overkill for something that just drives the lubrication pump… but hey, it’s fun

The code I wrote can be found here if anyone is interested: github.com/johanvergeer/maho-mh400e-conf...ree/main/lubrication

Now I want to use this package from a Python script that’s launched via loadusr in my machine config.

Some context:

• The script uses linuxcnc, hal, and qtvcp modules.
• LinuxCNC runs on the global system Python interpreter as far as I can see, which means my package has to be available outside of any virtual environment.
• Installing with pip install . into the global Python environment technically works, but it’s strongly discouraged on modern Debian-based systems due to PEP 668. I’ve already seen Poetry try to uninstall half of /usr/lib/python3 
  
• Virtualenvs don’t help much either, since they can’t see LinuxCNC’s compiled modules.
• Adding my repo folder to sys.path is possible, but feels like duct tape.

So… what’s the best practice here?

• Is it acceptable to install your own packages into something like /usr/lib/python3/dist-packages/?
• Is there an officially recommended way to include custom Python packages that need to be imported across multiple files or used in different HAL components?

Or am I simply in “here be dragons” territory and should stick to a bunch of flat .py files with global state? 

Any advice or patterns you use would be greatly appreciated!

Cheers,
Johan
 

Python programming is generally something of a band-aid programmed free-for-all.

Java has conventions and structure.

Python is just "whatever floats your goat".

Fully agreed — if LinuxCNC ever gets a proper Java API, I’ll gladly throw my Python into the chip tray

Until then, I’ll just keep pretending my Python package has structure and discipline… while knowing deep down it’s all held together by logging statements and good intentions

Backwards?

In what sense?

Java is far, far more feature-rich than Python.

Python does not even have switch statements!


I think the main reason I like Java is the editor (Intellij IDEA CE) is free and good.

PyCharm just is not that great.

And I believe any life without true access modifiers is not a good life.

Not all python executable builders actually work.