Upper bound version constraint on Numpy?

[nollety]

I was wondering why does radis sets an upper bound version constraint on Numpy in its setup file?

🐛 Describe the issue

When adding radis to a Python environment where a version of Numpy larger than 1.22.3 is required (either from the current project, or from other dependencies), the package manager is going to run into a resolution error as radis requires numpy<=1.22.3.

🎯 Expected behavior

Usually, it is not recommended to have upper bound version constraint for dependencies, but I don't know there are specific reasons here to avoid Numpy versions larger than 1.22.3?

💡 Possible solutions

Unless radis is incompatible with numpy>1.22.3, we could simply update the constraint set by this line in the setup.py file (similarly in the development conda environment file) to:

"numpy>=1.22.3"

(perhaps a typo changed > into <?)

If I change numpy>=1.22.3 to numpy, in environment.yml and let conda determine what version of Numpy to install, it picks 1.23.5 (latest version at the time of writing is 1.24.2). If I run the tests with this change, I have all tests passing.

---

_{Tip : you can vote for New Features on https://feathub.com/radis/radis
Also check for the Long Term Todo List}

[erwanp]

This was due to an old incompatibility with Vaex few months ago #490
In october we still didn't have confirmation of compatibility with vaex vaexio/vaex#2062 ; I just asked there for an update.

That your tests are working seems to suggest Vaex was also updated, and that we could lift the constraint