fix bug with nmr set writing units

[jacksund]

resolves #2508

[shyuep]

Please add a Unittest to catch this bug and prevent it from recurring.

[jacksund]

Yep, I'm working on that now 😄

[coveralls]

Coverage decreased (-0.7%) to 83.279% when pulling 08daf15 on jacksund:fix-nmr-set into 6c84eb5 on materialsproject:master.

[jacksund]

I was trying to add a lower-level check to all float-type INCAR settings, but am unsure the best place to do this. Maybe this would be in the get_string method of the Incar class? Might be something to consider in order to avoid issues with other vasp sets.

pymatgen/pymatgen/io/vasp/inputs.py

Line 693 in 6c84eb5

def get_string(self, sort_keys: bool = False, pretty: bool = False) -> str:

And the addition would be something like...

elif isinstance(self[k], float):
    # note the added float() which ensures FloatWithUnit conversion
    lines.append([k, " ".join([str(float(i)) for i in self[k]])])
# plus another check for a list of floats

EDIT: currently this PR only implements a check on the NMR set. This lower-level check would need to be a separate PR

[jacksund]

@shyuep a unittest is in place and this is ready for review

[shyuep]

Thanks.