Releases · oasci/reptar

Logging capabilities with sparse implementation.
Basic zarr File handling.
Geometry validation during sampling using validate_and_fill_geometry from
QCElemental. Also added utils.validate_geometry function that checks this for
multiple structures.
Tests for xtb energy+gradient calculations.
Tests for psi4 energy and energy+gradient calculations.
Pylint section in documentation.

reptar-write-xyz now supports all file types.
Handle 2D arrays in XYZ writers.
References to string files are renamed to xyz.
System information (i.e., atomic numbers and coordinates) are moved to
driver.run() instead of during initialization.
Unified specification of ray workers and number of cores.
n_workers always defaults to 1 which n_cpus_per_worker specifies the number
of CPU cores per worker (also defaults to 1).

Ability to remove data with the File object.
Option (missing_is_none) to return None from File.get() if the key does not exist.
Saver with sampling.
with_md5_update option for putting data in reptar files.
File.copy() method to help copying data between groups.
driverOPT and psi4_opt worker.
reptar-write-xyz script for writing xyz files from command line.
driverENERGY for just energy calculations.
psi4_energy worker.
Saver class for long calculations.
Parse final electronic energies from ORCA.
Support for CREST conformer and rotamer searches.
Support for Atomic Simulation Environment (ASE) trajectories.
Determine structure indices with respect to structure provenance specifications.
Simple structural descriptors.
Quick documentation for sampler.
Example script for GDML npz files.
Writer for standard XYZ files with optional comments.
Writer for extended XYZ files used in the Gaussian approximation potentials (GAP) code.
Documentation for writers.
Method to store arrays in ASE and schnetpack databases.

Pylint and Black formatting; thus, lots of naming changes were made.
Renamed creator class to Creator to avoid namespace conflicts.
Moved reptar.calculators.save.Saver to reptar.Saver.
Condensed and cleaned sampling routines into Sampler class.
Use qcelemental.periodictable for atomic numbers, symbols, and masses instead of manual dictionaries.
All drivers have a use_ray parameter that defaults to False.
Calculators now use keyword arguments.
Cleaned up ray calculations.
Convert scf energies from cclib to Hartree by default.
Moved example calculations to reptar-data.
Manual api documents with sphinx-multiversion.
Raise RuntimeError when key does not exist in exdir.
PDB writer numbers each atom name.
Renamed File.add() to File.put().
Return data from exdir as np.ndarray instead of np.memmap.
creator.group is now creator.from_calc.
Specify array dimensions in docstrings.
Do not import every submodule in reptar.
That way, optional calculation dependencies are not required when importing reptar.
Upgrades to README.
PDB writer is now a function instead of class.
PDB writer requires arrays instead of reptar files and group keys.
Rename data class to File.
This clears up previous ambiguous usage of "data" to refer to both a file and value of a key.
Rename reptarWriter to textWriter (more specific).
Require setting the memory for Psi4 worker.

Previously, coordinates were sliced based on a boolean mask during sampling.
The specific order of entity_ids in r_prov_specs is no longer correct.
For example, if the selection as [0, 382, 82, 45, 100] then the actual coordinates would be in order of 45, 82, and then 100, which is incorrect.
Any sampled structures before this commit do not have correct r_prov_specs with respect to the source.
psi4_engrad no longer translates and rotates molecules to origin (which would result in incorrect gradients).
_generate_structure_samples when quantity is all.
Would incorrectly generate combinations with replacement.

Provide feedback