Generate Weights should give a dataset compatible with xr.DatasetWeighted

Generic Issue

xscen version: 0.7.24-beta
Python version: 3.11.7
Operating System: Linux (jupyter)

Description

As an example:

Trying to get the weighted mean using dsim.weighted(generate_weights(dsim_array,weights)).mean() currently results in NaN values, since the coordinate for weights does not correspond to the realization coordinate for dsim.

Generating weights currently gives me a dataset with the coordinate "realization" = [0,1,2,3,4,5,6], for example. In my case, it should give the coordinates "realization"=['ScenarioMIP_BCC_BCC-CSM2-MR_r1i1p1f1_ssp245', 'ScenarioMIP_BCC_BCC-CSM2-MR_r1i1p1f1_ssp370', 'ScenarioMIP_BCC_BCC-CSM2-MR_r1i1p1f1_ssp585', 'ScenarioMIP_CAS_FGOALS-g3_r1i1p1f1_ssp245', 'ScenarioMIP_CAS_FGOALS-g3_r1i1p1f1_ssp370', 'ScenarioMIP_CAS_FGOALS-g3_r1i1p1f1_ssp585']

What I Did

Generate a dictionary of datasets indexed by {realization: dataset}, with each dataset having the appropriate "cat:" attributes. (see #301)
Generate the weights with xs.ensemble.generate_weights
Be sad that you can't do ds.weighted(weights).mean()

ds_list = get_ds_dict(ds)
return xs.ensembles.generate_weights(ds_dict,independence_level="model",attribute_weights=attribute_weights,standardize=True)

Solution

weights = weights.assign_coords(realization=dsim.realization)

Ouranosinc / xscen