Adds support for an MSONAtoms class that's an MSONable form of an ASE Atoms object

[Andrew-S-Rosen]

Summary

Here, I introduce the MSONAtoms class, which is simply an MSONable version of the ASE Atoms object, introducing an .as_dict() and from_dict() method that uses ASE's JSON encode/decoder to do the serialization. By default, the object returned from ASEAtomsAdaptor.get_atoms() is now an MSONableAtoms object. The original Atoms class can be returned via the msonable=False keyword argument, although this should rarely be needed in practice since the functionality of MSONAtoms is otherwise identical to Atoms. This is a follow-up to materialsvirtuallab/monty#619.

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)

[shyuep]

MSONAbleAtoms is a mouthful.How about just call it AtomsM?

[Andrew-S-Rosen]

MSONAbleAtoms is a mouthful.How about just call it AtomsM?

My personal opinion is that descriptive is better than concise. It's not immediately clear what an AtomsM means at a glance. Having it be unambiguous is particularly helpful with type hinting, among other things.

That said, I agree it's somewhat long. I'm open to other alternatives.

[JaGeo]

I would also vote for MSONAbleAtoms or something descriptive.

[shyuep]

Then MSONAtoms?

[Andrew-S-Rosen]

Works for me. I'll make the change when I'm at my laptop.

[mkhorton]

Looks good to me.

I do have a question, perhaps @Andrew-S-Rosen knows, is lossless conversion to/from Structure and Atoms possible? Or are some attributes potentially lost?

[Andrew-S-Rosen]

@mkhorton Good question. I have spent a fair bit of time on this module to ensure that it is lossless, and there are back-and-forth tests for this as well. If you spot an edge case, please let me know and I will patch it!

[janosh]

Or are some attributes potentially lost?

@mkhorton there's no remaining known loss of information during Atoms<->Structure round trips. many edge cases were reported in the past which have all been addressed, most recently in

#3359
#3338
#3270
#3151

unknown remaining edge cases are possible but likely rare

[shyuep]

Thanks @Andrew-S-Rosen . This is great!

[shyuep]

I just pushed a change that ensures MSONAtoms is returned by from_dict. This is supposed to be the case. Otherwise it loses functionality with deserializaation.

[ml-evs]

Something here breaks imports in cases where the user doesn't have ASE installed. This seems to be breaking the latest mp_api as a result:

from mp_api.client.mprester import MPRester
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/mp_api/client/__init__.py", line 8, in <module>
    from .mprester import MPRester
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/mp_api/client/mprester.py", line 10, in <module>
    from emmet.core.electronic_structure import BSPathType
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/emmet/core/electronic_structure.py", line 11, in <module>
    from pymatgen.analysis.magnetism.analyzer import (
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/pymatgen/analysis/magnetism/__init__.py", line 5, in <module>
    from pymatgen.analysis.magnetism.analyzer import (
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/pymatgen/analysis/magnetism/analyzer.py", line 24, in <module>
    from pymatgen.transformations.advanced_transformations import MagOrderingTransformation, MagOrderParameterConstraint
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/pymatgen/transformations/advanced_transformations.py", line 33, in <module>
    from pymatgen.io.ase import AseAtomsAdaptor
  File "/home/mevans/repos/re2fractive/re2fractive/experiments/2_featurizing_refractive_index_data/.venv-mp-api/lib/python3.11/site-packages/pymatgen/io/ase.py", line 44, in <module>
    class MSONAtoms(Atoms, MSONable):
                    ^^^^^
NameError: name 'Atoms' is not defined

Will raise an issue

[ml-evs]

see #3644