quantlab.tracking.base.flatten_metrics

quantlab.tracking.base.flatten_metrics(metrics)

Flatten nested metrics to outer/inner keys, keeping finite numbers only.

Booleans, NaN, infinities, strings, timestamps and every other non-numeric leaf are dropped, so every tracker accepts the result. Integers stay integers and floats (numpy ones included) become float.

Parameters:

metricsMapping

Metric names to values or to further mappings.

Returns:

dict[str, int | float]

The finite numeric leaves, keyed by their path joined with /.

Examples

>>> flatten_metrics({"whole": {"sharpe": 1.2, "worst": float("nan")}, "ic": 0.1, "note": "x"})
{'whole/sharpe': 1.2, 'ic': 0.1}