Smarter Dataset quantile method?

[ollie-bell]

What is your issue?

Calling ds.quantile(..., dim="x") result in all data variables in ds being modified to have a quantile dimension regardless of whether they had x as a dimension to begin with.

Is this intended? Can ds.quantile(..., dim="x") be called such that only data variables in ds that had x as a dimension end up with the x dimension being replaced by quantile dimension?

In my case I have a dimensionless data variable which I want to keep dimensionless. I know workaround are to either decode_coords="all" beforehand, or to do something like ds[my_dimensionless_var].isel(quantile=-1, drop=True) afterwards... but just checking if that is expected behaviour.

[mathause]

Thanks, this is certainly not desired behavior. Similar to #7027.