Periodic option for DataArray.rolling?

[yt87]

An example:

import xarray as xr
import pandas as pd

time = pd.date_range(start='2020-01-01', end='2020-12-31', freq='M')
da = xr.DataArray(np.arange(time.size), coords=[time], dims='time')
da.rolling(time=3, min_periods=2, center=True).mean().values

array([ 0.5,  1. ,  2. ,  3. ,  4. ,  5. ,  6. ,  7. ,  8. ,  9. , 10. , 10.5])

I would like to have a 3-month average instead:

array([4.,  1. ,  2. ,  3. ,  4. ,  5. ,  6. ,  7. ,  8. ,  9. , 10. , 7.])

A solution would be (if implemented):

da.rolling(time=3, min_periods=2, center=True, periodic=True).mean().values

Or, is there a workaround? In real life, da is a dask array.

[keewis]

This looks like the common boundary condition, for which I'd use DataArray.pad:

da.pad({"time": 1}, mode="wrap").rolling(time=3, min_periods=2, center=True).mean().isel(time=slice(1, -1))

This should work with any kind of duck array, including dask and numpy.

[yt87]

Thanks, this works for me.