Psience.Data.ScanManager

Unified scan infrastructure

ScanManager is the single entry point for both halves of a scan:

• .generate(scan_iterator, ...) – drains a scan_iterator (an iterable of (index, values, atoms, coords) steps), writes one job file per step into {output_directory}/scan_data_{scan_id}/, and records a scan_info.json manifest mapping each step to its file.
• .parse(molecular_property_extractor, ...) – reads that manifest back, rebuilds a Molecule from each step’s electronic-structure output, runs molecular_property_extractor on it, and stacks the results into one tensor per property key, shaped like the scan grid.

The default implementation targets ORCA (OrcaJob for generation, .out files for parsing). To support another package, subclass and override default_job_builder and default_output_file_generator:

class GaussianScanManager(ScanManager):
    job_file_ext = ".gjf"
    output_file_ext = ".log"

    def default_job_builder(self, atoms, coords, charge, **opts):
        return GaussianJob(atoms=atoms, cartesians=..., charge=charge, **opts)

    # default_output_file_generator's extension swap (.gjf -> .log) already
    # works unchanged as long as job_file_ext/output_file_ext are set above

A few free functions are provided to build scan_iterators for the two scan types worked out previously (a local Cartesian atomic-position scan, and an internal-coordinate scan), via structure_scan_iterator, which zips an N-dimensional grid of coordinate values with a structure_generator into the (index, values, atoms, coords) stream generate expects.

Members