incidentally, one way to get around the need to manually approve CI runs on every commit would be to merge a small PR instantly. if you want you could fix a few of the DOC errors

ruff check . --select DOC --preview

and submit a PR for that which we merge right away

why are only the windows tests failing? is the value of dataset.magmoms platform dependent?

You really had me here :) I haven't looked into the source code yet (thought that's a known issue), would have a look tomorrow.

Looks like dataset.magmoms is missing for windows for some reason:

assert len(dataset_dict["magmom"]) == len(dataset_dict["force"])
E       assert 0 == 1

Again:

reverse_readfile

your current arch nemesis strikes again 🤣

That's sad, looks like we cannot merge this until reverse readline is fixed, because this wouldn't be a working version for windows users.

your current arch nemesis strikes again 🤣

OK! my time off is over, going back to work tomorrow :)

Any comment on the following message (seemingly something terrible)?

tests\test_encoders.py .................
!!!!!! the two directed edges are equal but this operation is not supposed to happen
!!!!!! the two directed edges are equal but this operation is not supposed to happen
tests\test_graph.py ........

Any comment on the following message (seemingly something terrible)?

i think those messages are expected (maybe from a test meant to check that error message)

i'm fine with merging this and having a main branch where the parse_vasp_dir has issues on windows. we can wait with a package release until those are resolved

Thanks a lot for the migration to np 2! @DanielYang59 @janosh
I just fixed the message test_graph.py and will merge this PR now