iterate_subsets_by_window#

openstef_beam.evaluation.window_iterators.iterate_subsets_by_window(subset: T, window: Window, reference_date: datetime | None = None) → Iterator[tuple[datetime, T]][source]#

Yields evaluation subsets for each window with sufficient data coverage.

Guarantees:

  • Windows with coverage less than window.minimum_coverage are skipped

  • Each yielded subset contains only timestamps present in window, ground truth and predictions

  • All yielded subsets maintain the original sample interval

Parameters:

  • subset (TypeVar(T, bound= TimeSeriesDataset)) – The evaluation subset to iterate over

  • window (Window) – Window specification with size, stride, and minimum coverage parameters

  • reference_date (datetime | None) – Reference date for single window evaluation. When provided, yields a single windowed subset based on the reference date instead of all available windows.

  • subset

  • window

  • reference_date

Yields:

Iterator of (window end time, windowed evaluation subset) tuples

Return type:

Iterator[tuple[datetime, TypeVar(T, bound= TimeSeriesDataset)]]