Propagate labels through various Structure operations

[stefsmeets]

Summary

This PR follows up from #3137 and makes sure that labels are propagated correctly throughout structure, for example, when making supercells, inserting/deleting sites, applying translations/rotations, etc.

This PR also fixes a bug where the labels were incorrectly assigned when the CIF contains multiple species with different names.

Major changes:

• feature 1: Propagate labels in operations on structure
• fix 1: Correctly generate labels when cif contains multiple species with same name

Checklist

• Google format doc strings added. Check with ruff.
• Type annotations included. Check with mypy.
• Tests added for new features/fixes.
• If applicable, new classes/functions/modules have duecredit @due.dcite decorators to reference relevant papers by DOI (example)

Tip: Install pre-commit hooks to auto-check types and linting before every commit:

pip install -U pre-commit
pre-commit install

[janosh]

Thank you very much @stefsmeets, this is great! 🙏

Just to get an overview, are there more structure operations that are known to drop the site labels? Supercells, inserting/deleting sites, applying translations/rotations are all very important. Just want to leave a record here of what else might need doing in the future.

[stefsmeets]

@janosh Good point, I think these functions could still be updated:

• IStructure.from_spacegroup
• IStructure.from_magnetic_spacegroup
• IStructure.get_sites_in_sphere
• IStructure.get_all_neighbors
• IStructure.interpolate
• IMolecule.get_sites_in_sphere
• IMolecule.get_boxed_structure
• IMolecule.get_centered_molecule
• Everything to do with Neighbor/PeriodicNeighbor

I will update the PR with fixes for these today or tomorrow.

There is also the CIF writer which drops the labels. I looked at this before, but I found it a bit tricky to update.

[janosh]

There is also the CIF writer which drops the labels. I looked at this before, but I found it a bit tricky to update.

Feel free to just leave a todo comment for CIFParser so that we know it should still be done at some point.

[stefsmeets]

I had a go at updating the CifWriter with labels. Since the label defaults to Site.species_string, it turns out that it's a little bit more straightforward than when I first looked at this.

I also updated site.label to be a property. This helps generate the appropriate label for methods that update Site inplace (such as SitesCollection.add_oxidation_state_by_*).

Let me know what you think.

[janosh]

Almost ready to merge! Thanks for the great work. I think making label a @property was the right call.

[janosh]

@stefsmeets Thanks a lot for this PR! You really went the extra mile.

[mkhorton]

I noticed that performing a Structure.replace_species operation leaves the label intact, leading to a misleading label or a confusing end-user experience (for example, replacing Au with {"Au": 0.5, "Cu": 0.5} will leave the label as Au). Do you have any thoughts on this @stefsmeets?

[stefsmeets]

I noticed that performing a Structure.replace_species operation leaves the label intact, leading to a misleading label or a confusing end-user experience (for example, replacing Au with {"Au": 0.5, "Cu": 0.5} will leave the label as Au). Do you have any thoughts on this @stefsmeets?

I don't have any particular thoughts on this, but I can see how it would be confusing. What would you expect to happen?

[janosh]

What would you expect to happen?

maybe update the label to Composition({"Au": 0.5, "Cu": 0.5}).formula

[stefsmeets]

I can have a go at this next week. I created a new issue out of it to keep track.

[fxcoudert]

I think this caused a regression in CIF output: #3761