spglib 2.5 compatibility

[esoteric-ephemera]

Simple changes to accommodate updates to spglib's get_symmetry_dataset method in 2.5.0, which requires dot-access rather than dict-key access. Also pins the minimum version of spglib to account for this change.

[janosh]

spglib is a regular, not a build time dep. add it here

matminer/setup.py

Line 65 in 8e99cf6

"pymatgen >= 2023",

[esoteric-ephemera]

@janosh my bad, fixed!

[janosh]

thanks a lot @esoteric-ephemera! 👍

[janosh]

@esoteric-ephemera unfortunately it's installing 2.4 in CI, not 2.5 because of

matminer/requirements/ubuntu-latest_py3.10.txt

Line 293 in 8e99cf6

termcolor==2.4.0

need to update (or delete) those files as well

[janosh]

some tests are genuinely failing, apparently due to a discontinued/moved superconductivity2018 website. don't look that important, i think we can pytest.mark.skip(reason="...") these for now

_______________ MatminerDatasetsTest.test_superconductivity2018 ________________

self = <matminer.datasets.tests.test_datasets.MatminerDatasetsTest testMethod=test_superconductivity2018>

    def test_superconductivity2018(self):
        object_headers = ["composition"]
        numeric_headers = ["Tc"]
    
>       self.universal_dataset_check("superconductivity2018", object_headers, numeric_headers)

matminer/datasets/tests/test_datasets.py:643: 
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 
matminer/datasets/tests/test_datasets.py:[79](https://github.com/hackingmaterials/matminer/actions/runs/10530585976/job/29182138461?pr=943#step:8:80): in universal_dataset_check
    self.assertTrue(download_page.ok)
E   AssertionError: False is not true

[esoteric-ephemera]

Really weird - the tests all pass on my side, and for 3.10 (also manually checked the download links). Added the option to repeat trying the connection a few times in the test to see if any pass

[janosh]

in case you want to to exclude .DS_store to your global gitignore, something like this should do it (didn't test)

git config --global core.excludesfile ~/.gitignore_global && echo .DS_Store >> ~/.gitignore_global

[esoteric-ephemera]

Tests are passing for with the lower version pin of spglib, but not sure why auto-gen-release can't find a github token

[ml-evs]

Tests are passing for with the lower version pin of spglib, but not sure why auto-gen-release can't find a github token

There are many things I don't understand with the CI set up here, don't worry about it... otherwise, sounds great, if you re-commit the spglib 2.5 pin (and now leave it off of setup.py) then I'll merge!

[ml-evs]

Tests are passing for with the lower version pin of spglib, but not sure why auto-gen-release can't find a github token

There are many things I don't understand with the CI set up here, don't worry about it... otherwise, sounds great, if you re-commit the spglib 2.5 pin (and now leave it off of setup.py) then I'll merge!

Huh, actually looks like my latest auto-update PR also picked up the new spglib, without hitting this breakage... Perhaps we could just add a little test here too?

[esoteric-ephemera]

Huh, actually looks like my latest auto-update PR also picked up the new spglib, without hitting this breakage... Perhaps we could just add a little test here too?

Trying to see if setting the workflow permissions explicitly resolves this - don't think this is spglib related

[esoteric-ephemera]

It seems pretty consistent that the superconductivity2018 and tholander_nitrides datasets fail the URL status check - not sure why that is, but is it OK to skip these for now?

[ml-evs]

Sure, they seem unrelated, let's merge this for now

[ml-evs]

Thanks @esoteric-ephemera!