MP2020Compatibility not applying correct anion correction

[bowen-bd]

Version: 2023.10.11

from pymatgen.entries.compatibility import MaterialsProject2020Compatibility
from pymatgen.entries.computed_entries import ComputedStructureEntry

params = {"hubbards": {"Mn": 3.9, "O": 0, "Li": 0}, "run_type": "GGA+U"}

cse = ComputedStructureEntry(lmo, vasp_raw_energy, parameters=params)
MaterialsProject2020Compatibility(check_potcar=False).process_entries(cse)

The anion correction is missing for these O2-

[bowen-bd]

Code from here

[shyuep]

@janosh @kristinpersson feel free to tag whoever is supposed to be dealing with this in MP.

[janosh]

@shyuep The person to tag would be @rkingsbury or myself. I thought I fixed this problem in #3184 but looks like there's another bug related to Composition indexing using Elements/str vs. Species. Currently:

comp = Composition({"Li+": 2, "Mn3+": 2, "O2-": 4})
comp["Li"] == 0

I would argue comp["Li"] should return 2 instead, i.e. return all species with matching symbol, regardless of oxidation state. That would fix the problem @BowenD-UCB encountered since currently comp["O"] returns zero and so no oxide correction is applied.

But this would be breaking, and in particular might conflict with user expectations in the (rarely used) case of allow_negative=True since e.g. currently

comp = Composition({"Mg": 1, "Mg2+": -1})
comp["Mg"] == 1

which would change to return 0.

[shyuep]

I think it is fine for you to handle it for now. But a more permanent person in charge needs to be defined.