Python Minor Version Deprecation Policy

[StijnKas]

Here's an open question: according to the Numpy Python depcrecation policy, support for 3.9 will be dropped next month: https://numpy.org/neps/nep-0029-deprecation_policy.html. Currently, we support 3.8 (which is not even supported by numpy at this point). That policy doesn't guarantee it won't work on older versions, it just guarantees that versions above the minimum will be supported.

Do we want to tie our minimum supported version to the Numpy policy as well? Given that we have numpy as a dependency, I think it makes sense to hold onto their guidelines as the 'smallest' denominator. If they drop support and something breaks, there's nothing we can do to fix that. Thoughts @operdeck / @yusufuyanik1 ?

[operdeck]

On the "something breaks" - wouldn't we be covered through unit tests?

[StijnKas]

This now becomes more concrete: Numpy recently updated to 2.0, which supports 3.9 and up. There are backwards incompatible changes, so to 'support' the latest version of numpy, you need to tie to >=2.0. Our dependencies like Polars will surely do this soon, meaning by transitive requirements, we will only be compatible with versions <= 3.9. Just an fyi really that this will need to happen soon.