0.4.1

This is a minor release for compatibility with OpenFF Toolkit version 0.12.x.

Note the API is still under active development and may change in future.

Bug fixes

• #210 Toolkit 0.12 compatibility @mattwthompson

Full Changelog: 0.4.0...0.4.1

0.4.0

This is a minor release for compatibility with OpenFF Toolkit version 0.11.x.

Note the API is still under active development and may change in future.

New Features

• #204 Toolkit 0.11 support @Yoshanuikabundi

Full Changelog: 0.3.2...0.4.0

0.3.2

This is a micro release that contains new features.

Note the API is still under active development and may change in future.

New Features

• #198 Updated documentation, including migration to use Markdown files and openff-sphinx-theme @Yoshanuikabundi
• #202 Support multi-dimensional TorsionDrive grid indices @chapincavender
• #206 Support more QC programs provided by QCEngine @mattwthompson

Full Changelog: 0.3.1...0.3.2

0.3.1

This is a micro release that contains bug fixes, QoL improvements and new features.

Note the API is still under active development and may change in future.

BugFixes

#183 Fixed a bug that stopped molecules from being added to datasets via attributes passed to Dataset.add_molecule. @dotsdl
#184 Correctly create results from collections with missing inchi_key fields @jthorton
#187 Now able to make results from single point datasets with Nan records @jthorton
#192 Basic datasets created from an OptimizationResultCollection now keep all conformers for molecules. @jthorton

New Features

#186 Improved performance for submissions of large datasets by reducing the number of save calls @dotsdl
#193 Fragmentation components now support user supplied torsion target SMARTS @jthorton

Full Changelog: 0.3.0...0.3.1

0.3.0

This is a major release that contains bug fixes, QoL improvements and new features.

Note the API is still under active development and may change in future.

BugFixes

#175 DFTD3 specs on datasets are now dropped in favour of the combined single spec when using the update_specification_and_metadata function.
#178 Derealisation now raises the correct error type when the file can not be found.
#180 Multi-component systems are now split, reordered and recombined to make sure each component has a continuous atom ordering to stop run time errors. Molecule attributes are no longer required when adding a molecule to a dataset as they can be generated on the fly for the molecule. In the case of reordering, this helps the attributes and molecule object stay in sync.
#181 Dataset factories now raise the correct error type when an input file can not be found.

New Features

#177 Torsiondrive datasets now also support constraints using the same API as optimisation datasets.

0.2.4

This is a minor release that contains bug fixes and QoL improvements.

Note the API is still under active development and may change in future.

BugFixes

#173 QCSpec keywords can now be a list of floats which allows for perturbed calculations in psi4.

New Features

#165 This adds support for HDF5 files as input to dataset factories acting as the main input for multicomponent systems, see docs for more details.

0.2.3

This is a minor release that contains bug fixes and QoL improvements.

Note the API is still under active development and may change in future.

BugFixes

#164 allows Psi4 QC specifications to use a None basis set which is not validated. This is useful for shorthand CBS calculations.
#163 switches the results filters to use the fixed hydrogen layer inchi key when deduplicating molecules to distinguish between tautomers.
#160 adds support for the new openmm import paths for versions >7.6.0 and keeps backwards support for versions <7.6.0.
#159 now uses keywords to query for results to improve speed and reliability.
#157 dihedrals chosen for torsion drives are also now deduplicated based on molecule symmetry.
#156 the conformers from a torsion drive results object are now ordered from lowest to highest angle to aid visualisation.

New Features

#162 adds a new results filter based on the minimum number of conformers.
#158 changes the way toolkits are passed to dataset factories, now toolkit registries can be used and are respected by the workflow rather than configuring each workflow component separately.
#154 adds the ability to convert a basic or optimisation dataset directly to a list of QCEngine JSON tasks ready for execution. Torsion drives will be added at a later date.

0.2.2 Conda-Forge

This marks the first stable release to be put on conda-forge, it contains bugfixes and some QoL improvements.

Note the API is still under active development and may change in future.

BugFixes

• #150 allows extra calculation keywords to be passed to the qcspec such as the integration grid size.
• #148 automatically retries to connect to a local QCFractal instance when verify should be set to False
• #137 stips any atom maps from the input molecules before they enter the workflow.
• #108 removes the threaded submissions which were introduced in the last release and caused a lot of issues. A process pool is now used instead but is switched off by default.
• #103 fixes molecule enumeration to make sure input molecules are not lost.

New Features

• Documentation is now available here
• The results classes have been totally reworked to now only store references to data in a QCFractal instance rather than copies of it.
• Results classes now have their own custom filters which can be used to refine datasets see an example here.

0.2.1

This is a minor bug fix release.

Bugfixes

• PR #96: Fixes a bug introduced by the threading update where no tasks are generated if a blank dataset is used to expand the compute of an existing dataset. This is a temporary fix until a more explicit pathway to compute expansions is implemented.
• RP #95: Fixes a bug in qcspecification validation which stopped any smirnoff99frosst forcefields from being used. This requires a special version of the openforcefield toolkit which can be installed using conda install openforcefield -c omnia/label/benchmark. This is a temporary fix until we move to the openff-toolkit package on conda-forge.
• PR #94: Fixes a bug introduced by the use of literal types in the geometric procedure schema which would raise an error when creating results from a dataset that had non-default geometric settings due to case sensitive settings. This is now fixed by a validator that runs before the literal type check to convert the input to the correct case.

0.2.0

Namespace migration

This major release includes the package namespace migration to conform with other openforcefield projects and will break all import paths. The package should now be imported as from openff.qcsubmit import X.

Bugfixes

• PR #86: Stops QCSubmit from changing the index given to an entry in aBasicDataset or OptimizationDataset if the entry has only one conformer.
• PR #84: Updates the geometric optimization schema to use fields and fixes a bug in the default settings for optimizations where reset was set to False rather than True.
• PR #69: Updates the logic used to check basis set names in the basis_set_exchange package to correctly handle month and heavy basis sets.
• PR #73: This fixes a molecule deduplication bug where after identifying two molecules with the same inchikey a graph isomorphism check would fail to find a mapping between the molecules due to differences in bond order and formal charge.

New Features

• PR #81: This updates all schemas to use pydantic fields which will help with documentation and also introduces the openff.qcsubmit.common_structures.SCFProperties enum to help validation and make it easier for users to check what options are available. We also introduce openff.qcsubmit.common_structures. MoleculeAttributes class which helps cmiles validation but is also extendable to contain any extra information that users want to add to an entry.