Improve installation speed in Google Colab notebooks using uv pip compared to conda

[janpeter19]

Today the recommended installation of PyFMI is using conda-forge in a conda virtual environment.

When you use Google Colab notebooks it takes comparatively long time to first install miniconda and then PyFMI using conda-.forge channel, compared installing FMPy using uv pip.

Here is perhaps other ways to improve startup than switching from conda to uv pip that could be an alternative. I am interested to hear that too.

You can see for yourself at my GitHub page https://github.com/janpeter19. Each example has its own repository. And all examples can be run with either PyFMI or FMPy. Note the difference in time it takes to startup using PyFMI compared to FMPy.

For educational purpose it is useful to use Google Colab in this way. No installation on site. And students can run from the own computer at home as well using any Web-browser.

[PeterMeisrimelModelon]

Hi,

this is a known issue, see e.g., #210.

We will need to re-do the build system for PyFMI for Python 3.12 and hope we can resolve the issues that currently prevent wheel upload to PyPi to have up-to-date wheel available via pip.

/Peter

[janpeter19]

Thank you for a quick response.

I noted in August the change to default Python 3.12 for Google Colab,
without any announcement in the channel I expected https://medium.com/google-colab

Do you have any rough time schedule for this work of yours to update installation for Python 3.12?

(My experience using conda and conda-forge with Python 3.12 in Colab is very good.
But use of uv pip just do not work and the reasons for it you described above).