`iris.util.rolling_window` does not support Dask arrays

[bouweandela]

🐛 Bug Report

According to the docstring of iris.util.rolling_window it maintains laziness when called, but actually passing in a Dask array leads to a crash.

How To Reproduce

Steps to reproduce the behaviour:

In [1]: import dask.array as da

In [2]: from iris.util import rolling_window

In [3]: x = da.arange(10).reshape((2, 5))

In [4]: rolling_window(x, 3).compute()
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[4], line 1
----> 1 rolling_window(x, 3)

File ~/src/scitools/iris/lib/iris/util.py:337, in rolling_window(a, window, step, axis)
    334 num_windows = (a.shape[axis] - window + step) // step
    335 shape = a.shape[:axis] + (num_windows, window) + a.shape[axis + 1 :]
    336 strides = (
--> 337     a.strides[:axis]
    338     + (step * a.strides[axis], a.strides[axis])
    339     + a.strides[axis + 1 :]
    340 )
    341 rw = np.lib.stride_tricks.as_strided(a, shape=shape, strides=strides)
    342 if ma.isMaskedArray(a):

AttributeError: 'Array' object has no attribute 'strides'

Expected behaviour

Instead of a crash, I would expect an output array

In [1]: import dask.array as da

In [2]: from iris.util import rolling_window

In [3]: x = da.arange(10).reshape((2, 5))

In [4]: rolling_window(x, 3).compute()
Out[4]: 
array([[[0, 1, 2],
        [1, 2, 3],
        [2, 3, 4]],

       [[5, 6, 7],
        [6, 7, 8],
        [7, 8, 9]]])

Environment

• OS & Version: Ubuntu 23.10
• Iris Version: 3.8.0.dev125