Set up dev tools about scikit-learn-knn-regression HOT 3 CLOSED

Good thought on the ArcGIS constraint! I hadn't considered that.

I agree that following sklearn's lead here makes sense. Adding a from __future__ import annotations gives access to typing with built-ins as early as 3.7 (see PEP 585), so I think we should be good to go with 3.8!

from scikit-learn-knn-regression.

@grovduck Any thoughts on Python version support? I'm setting up tox and need to figure out which environments to test against, but we don't currently list a python_requires in setup.py.

I usually loosely follow the numpy support schedule, which just dropped 3.8 a few days ago, but that seems a bit strict. sklearn still supports >=3.8. Any preferences?

from scikit-learn-knn-regression.

@aazuspan, this will start sounding repetitive, but I'll trust your judgement on this. A couple of random thoughts to consider, but probably not too strongly:

• It's possible that we might have ArcGIS users that aren't too familiar with Python and would therefore use the Python that ships with ArcGIS Pro. The default is a conda virtual environment using Python 3.9.
• When thinking about typing, it's really nice to just use list, dict, etc. and that came in at 3.9? We'd have to import Typing otherwise, correct? I'm guessing you know clever tricks around this.

At first blush, it seems to be reasonable to follow what sklearn supports, so I'm OK with 3.8.

from scikit-learn-knn-regression.