Changelog¶

Version 0.3.4¶

Added clear CLI command to remove all ORCA files in the working directory (useful for cleaning up files that weren’t removed due to errors)
Added purge_cache CLI command to remove ORCA cache

Improved cache system to preserve original file extensions (.out, .log, .smd.out) when storing files
Enhanced NBO stabilization energy parsing with better error messages indicating NBOEXE environment variable requirement
Updated test for AM1 Mayer bond indices to account for different values compared to DFT methods

Removed ESP extrema descriptor (get_esp_extrema) - not implemented due to limitations in ORCA 6.0.1 for generating ESP cube files directly (would require orca_plot utility integration)

Fixed cache system issue where files were saved with .out extension regardless of original extension, causing file not found errors
Fixed NBO stabilization energy parsing to properly detect when NBOEXE environment variable is not set
Fixed NMR chemical shifts parsing to work correctly with cached files
Fixed test for AM1 Mayer bond indices to accept different value ranges compared to DFT

Cache now preserves original file extensions to ensure correct file retrieval
NBO analysis requires NBOEXE environment variable to be set to point to NBO executable (nbo6.exe or nbo5.exe)
ESP extrema calculation would require integration with orca_plot utility, which is not currently implemented
CLI commands clear and purge_cache help maintain clean working directories and cache management

The new modular structure makes it easier to extend functionality and maintain code
Descriptors are organized by category for better code navigation
All functionality remains backward compatible

Added new ORCABatchProcessing class for efficient batch processing of molecular descriptors with pandas compatibility
Added support for semi-empirical methods (AM1, PM3, PM6, PM7, RM1, MNDO, MNDOD, OM1, OM2, OM3)
Added pre_optimize parameter (default: True) for pre-optimizing molecular geometry using MMFF94 force field before ORCA calculations
Added multiprocessing support for parallel batch processing via parallel_mode="multiprocessing" parameter
Added automatic cleanup of all ORCA temporary files (input, output, and all temporary files) after calculations since results are cached
Added improved error parsing with brief summaries in logging.INFO and detailed information in logging.DEBUG
Added _pre_optimize_geometry() method for MMFF94 geometry optimization
Added _is_semi_empirical() method to detect semi-empirical methods

Refactored batch processing functionality from Orca.calculate_descriptors() into dedicated ORCABatchProcessing class
Orca.calculate_descriptors() now uses ORCABatchProcessing internally for backward compatibility
calculate_descriptors() now preserves original DataFrame columns (including ‘smiles’) instead of removing and re-adding them
Improved molecule hash calculation to include pre_optimize parameter for proper caching
Updated molecule hash calculation to exclude basis_set and dispersion_correction for semi-empirical methods
Enhanced input file generation to support semi-empirical methods (no basis set or dispersion correction needed)
Improved file cleanup to remove all ORCA files (including input and output files) since results are cached

Fixed DataFrame handling in batch processing to preserve all original columns
Fixed error handling to provide concise error messages in INFO level and detailed information in DEBUG level

ORCABatchProcessing supports three parallelization modes: “sequential”, “multiprocessing”, and “mpirun”
Pre-optimization uses RDKit’s MMFF94 force field for fast geometry optimization before quantum chemical calculations
All ORCA files are automatically cleaned up after successful calculations, with results stored in cache
Semi-empirical methods are automatically detected and handled differently from DFT methods
Batch processing now includes time estimation based on benchmark machine performance

Added numpy>=1.20.0 to project dependencies (numpy was used but not declared)
Added dynamic time estimation updates in batch processing - time estimates are now refined based on actual execution times of previous molecules
Added _get_available_descriptors() method to dynamically discover available descriptor methods

Updated dipole moment parser to prioritize gas-phase values when available (for calculations without solvation)
Improved time estimation algorithm: * Changed scaling exponent from O(N^3.5) to O(N^2.5) for more realistic estimates * Uses total_time from benchmark instead of scf_time as base unit * More realistic optimization step estimation (15-35 steps instead of 10-50) * Removed artificial 24-hour time cap
Refactored calculate_descriptors() method: * Removed redundant code duplication (replaced large if-elif chain with getattr-based method calls) * Removed redundant all_descriptors list - descriptors are now discovered dynamically * Removed unnecessary comments * Improved code maintainability and readability

Fixed dipole moment parser to correctly extract gas-phase values from ORCA output when available
Fixed time estimation showing unrealistic values (e.g., 47 hours for 2 molecules) - now provides accurate estimates based on actual benchmark data

Time estimator now uses exponential moving average for better prediction accuracy
Descriptor methods are called dynamically using getattr(self, desc_name)
Automatic descriptor discovery eliminates need to maintain manual descriptor lists