Define what needs to be parsed

[ligerzero-ai]

Let's have a clear scope of what parsing needs to return:

A dict of objects (1 object per file)?

A dict of generic DFT outputs (stresses, magmoms, forces, energies etc.)?

The goal is to minimise any custom parsing that we might have to do ourselves, and leverage existing parsers where possible (ase, pymatgen etc.)

[liamhuber]

If possible, I feel it would be nice to see the IO formatted in terms of a dataclass dataclasses rather than a raw dict

[jan-janssen]

For compatibility with pyiron_atomistics I would aim for reproducing the output of parse_vasp_output():

from pyiron_atomistics.vasp.output import parse_vasp_output
parse_vasp_output(working_directory="pyiron_vasp/tests/static/vasp_test_files/full_job_sample")
parse_vasp_output(working_directory="pyiron_vasp/tests/static/vasp_test_files/full_job_corrupt_potcar")

[jan-janssen]

The ASE parser for atomistic structures is better than expected:

atoms = read(filename="pyiron_vasp/tests/static/vasp_test_files/poscar_samples/POSCAR_no_species", format="vasp")
atoms

>>> Atoms(symbols='HZn32', pbc=True, cell=[7.9, 7.9, 7.9], constraint=FixAtoms(indices=[0, 29, 30, 31, 32]))

[ligerzero-ai]

The ASE parser for atomistic structures is better than expected:

atoms = read(filename="pyiron_vasp/tests/static/vasp_test_files/poscar_samples/POSCAR_no_species", format="vasp")
atoms

>>> Atoms(symbols='HZn32', pbc=True, cell=[7.9, 7.9, 7.9], constraint=FixAtoms(indices=[0, 29, 30, 31, 32]))

In what way? speed?

[jan-janssen]

In what way? speed?

It is picking up the selective dynamics and converts those into the corresponding constraint.

[jan-janssen]

If possible, I feel it would be nice to see the IO formatted in terms of a dataclass dataclasses rather than a raw dict

We are working on the dataclasses in https://github.com/pyiron-dev/pyiron-hdf5-format/blob/main/pyiron_io/dataclasses/generic.py