sweep.load_job_results

sweep.load_job_results(
    cli,
    job,
    registry,
    export_paths,
    *,
    run_id,
    export_type='patch',
    quiet=False,
    load_config=None,
    succeeded_before=0,
)

Load CSV results for a single job into the registry.

Extracts result loading logic from recover_sweep_results() for reuse in adaptive sweep runners. Includes retry logic for robustness against filesystem delays and partial writes.

Parameters

Name	Type	Description	Default
cli	JoshCLI	JoshCLI instance (needed for path resolution).	required
job	ExpandedJob	The completed job to load results for.	required
registry	RunRegistry	Registry to load results into.	required
export_paths	ExportPaths	Export path configuration from inspect_exports().	required
run_id	str	The run_id to attribute cell_data rows to. Must be the run_id from the specific CLI invocation that produced these results (from registry.start_run() or RegistryCallback.record()).	required
export_type	str	Type of export (“patch”, “meta”, “entity”).	`'patch'`
quiet	bool	Suppress output.	`False`
load_config	LoadConfig \| None	Retry and timing configuration. Uses defaults if None.	`None`
succeeded_before	int	Number of jobs that succeeded before this one. Used for error context in ResultLoadError.	`0`

Returns

Name	Type	Description
	int	Number of rows loaded.

Raises

Name	Type	Description
	ResultLoadError	If raise_on_missing=True and CSV not found after retries, or if CSV load fails after retries.

Examples

>>> # Basic usage
>>> run_id = registry.start_run(run_hash=job.run_hash)
>>> result = cli.run(to_run_config(job))
>>> registry.complete_run(run_id, exit_code=result.exit_code)
>>> export_paths = cli.inspect_exports(InspectExportsConfig(...))
>>> rows = load_job_results(cli, job, registry, export_paths, run_id=run_id)