FYI, letting you know @esoteric-ephemera.

Did this not get fixed in #4431?

@esoteric-ephemera: Thanks for the redirect.

Yes, it looks like if the user installs Pymatgen today and runs the CLI entirely from scratch, everything should work due to #4431.

Upon further inspection, it seems like the issue @jinlhr542 and I ran into in #4456 was due to the breaking change. In previous versions of Pymatgen, the CLI would make a folder called POT_GGA_PAW_PBE_64. For whatever reason, this is now called POT_PAW_PBE_64. So, users who want to use the .64 PAW PBE potentials with Pymatgen or Atomate2 but used the Pymatgen CLI to generate the pseudopotential directory before #4431 was merged will run into compatibility issues.

@esoteric-ephemera: So I guess the question is --- do we revert the breaking name change, or do we just stick with what we have in master?

(I would, as usual, be happy about avoiding breaking changes. 😅)

Either way leads to a breaking change, since a user who set up the directory structure manually and doesn't use the CLI (like me) will have it break the other way

Only non breaking solution would be a check for either directory name

@esoteric-ephemera: Yes, this is also true.

I am okay with keeping things as is if that's preferred, but I want to document it here in case we get other reports. I will go ahead and close this PR, but anyone can comment if they want it re-opened.

Hey sorry I was unclear, I'm happy to add this check for either PBE.64 folder name. It's probably for the best if both are recognized by pymatgen

@esoteric-ephemera: That works and may reduce the headache in the long-term for MP staff 😉

Thanks for clarifying and offering to help! Certainly feel free to do so if you feel so inclined!

Looks good to me.

@shyuep FYI, I wouldn't merge as-is yet. @esoteric-ephemera suggested supporting both names. I'll leave that for others to deal with.