Add new tutorial video

[dcherian]

No description provided.

[mathause]

Nice! I did not know about coarsen(...).construct(...) - interesting. I usually use set_index for that - could be interesting to test which approach is faster.

def_to_seasonal(ds):
year=ds.time.dt.year.dataseason=ds.time.dt.season.data# assign new coordsds=ds.assign_coords(year=("time", year), season=("time", season))
# reshape the array to (..., "season", "year")returnds.set_index(time=("year", "season")).unstack("time")

[max-sixty]

Very good video @dcherian !

[dcherian]

@mathause I suspect coarsen.construct is faster because it's a simple reshape (ignoring the copies associated with handling the boundary condition):

xarray/xarray/core/variable.py

Line 2373 in 95bb9ae

returnvariable.data.reshape(shape), tuple(axes)

Unstacking allocates a new array and then assigns all values. It is definitely the more general solution though.

We should discuss this in the proposed cookbook (#1790)