Issues installing python packages with CondaPkg

General Usage
1

Hi everyone!

I have a Julia package, which I have been developing for some time. Today I was making some changes to it, and at some point something got broken that I don’t know how to fix. The problem happens when I try to import my module, which uses PythonCall and installs a conda environment with CondaPkg (I also have a CondaPkg.toml file). Whenever I now try to import the module, I am having issues during the CondaPkg environment creation. Basically the entire conda environment is being created, but in the end I get an error for building the wheels of some python packages and the whole installation fails.

My first reaction was to revert to an old git commit where everything was working, but still the problem persisted. I have since tried to also completely reinstall all Julia packages, delete the entire contents of the .julia and .CondaPkg folders where the python environments are created. I don’t know what to search for at this point, and could really use some help.

Thanks so much for helping, I’ll be happy to provide any additional information that’s necessary :slight_smile:

2

It may a problem with a specific version of a Python package that breaks… Can you try lowering its version to an older one in CondaPkg.toml?
Just out of curiosity, how do you install the conda packages? I use this command to do it (in a Dockerfile): julia --project -e "using Pkg; Pkg.activate(\".\"); using CondaPkg; CondaPkg.update()"

3

In the end I managed to fix the problem by changing the dependencies to be installed using conda instead of pip in the CondaPkg.toml file. I suspect indeed there was a specific Python package version that was breaking things, but I couldn’t find it and after reading a recommendation to use conda when available this seemed to do the trick. On the upside it seems that the installation is also more robust now :slight_smile: .
Thanks so much for taking the time to answer and help out!