allow kwargs in map_over_datasets?

[mathause]

Is your feature request related to a problem?

datatree.map_over_subtree allowed passing kwargs but xr.map_over_datasets does not. I think it probably should?

import numpy as np
import xarray as xr
import datatree

example_dict = {"set1": xr.Dataset({"var1": xr.DataArray([1, 2, 3], dims = "time")})}
dt = datatree.DataTree.from_dict(example_dict)
datatree.map_over_subtree(np.var)(dt, ddof=0) # works

import numpy as np
import xarray as xr

example_dict = {"set1": xr.Dataset({"var1": xr.DataArray([1, 2, 3], dims = "time")})}
dt = xr.DataTree.from_dict(example_dict)
xr.map_over_datasets(np.std, dt, ddof=1) # fails

Describe the solution you'd like

No response

Describe alternatives you've considered

Of course partial could be used or other such constructs but that's not so convenient...

Additional context

No response

[mathause]

Partially (🤷‍) discussed in #9693 (comment)

[TomNicholas]

I think adding **kwargs would be fine. As @shoyer said in that linked comment it would follow the pattern of xr.apply_ufunc.

The rabbit hole to avoid IMO is passing different kwargs to different datasets.

Do you want to submit a PR for this?

[shoyer]

There is an ambiguity between keyword arguments to xarray.map_over_datasets and the wrapped function.

For this reason, my preference would be to add keyword argument support into an explicit argument, e.g., xr.map_over_datasets(np.std, dt, kwargs=dict(ddof=1)), similar to xarray.apply_ufunc.