Code 2: Fix the error with Python >=3.9. Handle ruff and mypy

[andrewtarzia]

Note that the tests that fail in test_molecular seem like genuine changes in the numerical rebuild process. I wonder if changing versions actually has altered some of the math slightly? Any insight @marcinmiklitz ?

I do not think it is possible to fix mypy issues without significant changes to the data handling in pywindow, which is not in my timeframe at the moment. So I have ignored many with inline ignores.

To see an example of new docs: https://pywindow2--2.org.readthedocs.build/en/2/

[marcinmiklitz]

Hey @andrewtarzia,

I went through the PR and nothing stands out to me.

The 3 tests that are failing, are you more familiar with the nature of the mismatch. If these failures indicate a completely different answer, or simply out of 2 similar choice the numerical errors shifted the balance towards the other?

I haven't dig particularly deep yet. How worried we are here?

[andrewtarzia]

Hey @marcinmiklitz,

I think I have "solved" the tests. Basically, I think some math changed in the coordinate and ordering in the supercell generation and molecule (make_modular) process between numpy versions. So, I have updated the outputs based on current version, and checked the molecules made sense (CC3!), and output the expected structures to files for visualisation because comparing chemistry was hard with those dictionaries.

Please review the last two commits of changes.

[marcinmiklitz]

Hey @marcinmiklitz,

I think I have "solved" the tests. Basically, I think some math changed in the coordinate and ordering in the supercell generation and molecule (make_modular) process between numpy versions. So, I have updated the outputs based on current version, and checked the molecules made sense (CC3!), and output the expected structures to files for visualisation because comparing chemistry was hard with those dictionaries.

Please review the last two commits of changes.

Hey @andrewtarzia, thanks for solving this! I think this is a sensible explanation and solution. We've only had couple tests affected, not all of them, so it gives confidence that the differences were very small, hopefully not affecting results where larger samples are considered. Thanks! Merging this now.