OptimizationDataset doesn't inherit GeometricProcedure from OptimizationDatasetFactory

[amcisaac]

When you create an OptimizationDataset using OptimizationDatasetFactory.create_dataset(), it doesn't inherit the optimization settings specified in GeometricProcedure from the OptimizationDatasetFactory.

Given that there's the option to specify the GeometricProcedure at the factory level, I would expect it to transfer that to the dataset, but I think it doesn't since the BaseDataset and BasicDataset don't have this property.

Minimal example:

from openff.qcsubmit.factories import OptimizationDatasetFactory
from openff.qcsubmit.procedures import GeometricProcedure
from openff.toolkit.topology import Molecule

opt_procedure_notdefault = GeometricProcedure(program='geometric', 
                  coordsys='cart', 
                  convergence_set='TURBOMOLE', 
                  constraints={})

dataset_factory = OptimizationDatasetFactory(optimization_program=opt_procedure_notdefault)

dataset_factory.optimization_program == opt_procedure_notdefault # Returns True

mols = [
    Molecule.from_smiles(smiles)
    for smiles in [
        "CN(C)O",
        "CC",
    ]
]

dataset = dataset_factory.create_dataset(
    dataset_name="Test",
    tagline="testing geometric procedure",
    description=(
        "A dataset to test geometric procedure"
    ),
    molecules=mols
)

dataset.optimization_procedure == opt_procedure_notdefault # Returns False

dataset.optimization_procedure # Returns the default

# GeometricProcedure(program='geometric', coordsys='dlc', enforce=0.0, epsilon=1e-05, reset=True, 
# qccnv=False, molcnv=False, check=0, trust=0.1, tmax=0.3, maxiter=300, convergence_set='GAU', constraints={})

dataset._get_specifications() # Returns the default-- see `keywords` at the end for the optimization part

# {'default': OptimizationSpecification(program='geometric', qc_specification=QCSpecification(program='psi4', 
# driver=<SinglepointDriver.deferred: 'deferred'>, method='b3lyp-d3bj', basis='dzvp', keywords={'maxiter': 200, 'scf_properties': 
# [<SCFProperties.Dipole: 'dipole'>, <SCFProperties.Quadrupole: 'quadrupole'>, <SCFProperties.WibergLowdinIndices: 
# 'wiberg_lowdin_indices'>, <SCFProperties.MayerIndices: 'mayer_indices'>]}, protocols=AtomicResultProtocols(wavefunction=
# <WavefunctionProtocolEnum.none: 'none'>, stdout=True, error_correction=ErrorCorrectionProtocol(default_policy=True, 
# policies=None), native_files=<NativeFilesProtocolEnum.none: 'none'>)), keywords={'coordsys': 'dlc', 'enforce': 0.0, 'epsilon': 1e-05, 
# 'reset': True, 'qccnv': False, 'molcnv': False, 'check': 0, 'trust': 0.1, 'tmax': 0.3, 'maxiter': 300, 'convergence_set': 'GAU'}, 
# protocols=OptimizationProtocols(trajectory=<TrajectoryProtocolEnum.all: 'all'>))}

The user can fix this by manually setting the optimization settings after creating a dataset:

dataset.optimization_procedure = opt_procedure_notdefault

[amcisaac]

It won't let me upload the yaml with my full environment, but I'm using QCSubmit 0.54.0, installed today so everything should be up to date. I'm pretty sure the environment's not the problem.

name: openff_min
channels:
  - conda-forge
dependencies:
  - python
  - openff-toolkit
  - openff-qcsubmit
  - jupyter