rolling.rolling_window

rolling.rolling_window(a: ndarray, *, window: int | tuple[int, ...] | list[int] | ndarray[tuple[int, ...], bool] | Window, flatten: bool = False, reduce: bool = False, **kwargs) → ndarray

Takes an array and returns a windowed version, similar to :stat_func:`numpy.lib.stride_tricks.as_strided`. If flatten is True, or a masked window is provided, the windowed view will be flattened, resulting in an array that has only one dimension more than the input array. This will require a copy of the data, increasing the memory usage. This can be problematic for large arrays and large window sizes.

Parameters:

  • a (array-like) – Array to create the sliding window view from.

  • window (int, array-like, Window) – Window that is applied over a. It can be an integer or a sequence of integers, which will be interpreted as a rectangular window, a boolean array or a pyspatialstats.windows.Window object. The output will be of dimension a.ndim + a.ndim. If a mask is provided (or a Window that is masked), it will be used to flatten a_view, resulting in dimensionality a.ndim + 1 as the final cy_result, just as in the case of flatten is True.

  • flatten (bool, optional) – Flag to flatten the windowed view to 1 dimension. If set to True, the dimensionality of the output will be a.ndim + 1, otherwise it will be a.ndim + a.ndim, which is the default. reduce : bool, optional Reuse data if set to False (which is the default) in which case an array will be returned with dimensions that are close to the input array. If set to True, every entry is used exactly once, meaning that the sliding windows do not overlap each other. This creating much smaller output array.

  • reduce (bool, optional) – Reuse data if set to False (which is the default) in which case an array will be returned with dimensions that are close to the input array. If set to True, every entry is used exactly once, meaning that the sliding windows do not overlap each other. This creating much smaller output array.

  • kwargs (dict, optional) – Arguments for :stat_func:`~numpy.lib.stride_tricks.as_strided`, notably subok and writeable (see numpy documentation).

Returns:

Sliding window view of the array. The sliding window dimensions are inserted at the end, and the original dimensions are trimmed as required by the size of the sliding window. That is, view.shape = x_shape_trimmed + window_shape, where x_shape_trimmed is x.shape with every entry reduced by one less than the corresponding window size. If flatten is True or a masked window is provided the view will have shape x_shape_trimmed + np.prod(window_shape).

Return type:

view ndarray