Searching by standard_name doesn't refine the resulting dataset by standard_name

[anton-seaice]

Is your feature request related to a problem? Please describe.

When searching a datastore by standard_name, it returns one catalog dataset:

e.g.
esm_ds.search(variable_standard_name="sea_surface_height_above_geoid")

However when .to_dask() is run, the resulting xarray dataset has all variables in the source files included:

Describe the feature you'd like

I would like the resulting dataset to only include the variable which has the requested variable_standard_name

Describe alternatives you've considered

Ignore it, usign the cf standard name on the resulting dataset works fine, e.g.:

ds.cf["sea_surface_height_above_geoid"]

Additional context

This is mostly an issue of neatness.

As the dataset is still a dask object, and the source files need opening anyway, the performance benefit of only including the requested variables in the returned dataset is probably small.

I think this is an issue with intake-esm upstream ? ping @dougiesquire and @charles-turner-1

[dougiesquire]

Yeah, we've talked about this before @anton-seaice. This is a "feature" of Intake-esm datastores, which keep track of a single column for the dataset variables, defined by the variable_column_name. Only searches on this column (which is "variable" for most of our datastores) will refine the dataset returned by to_dask().

This possibly wouldn't be too hard to change in Intake-ESM

[charles-turner-1]

I'm also wondering if we might be able to use a DerivedVariableRegistry to do this on the fly - I'm not sure whether we might run into difficulties knowing all the different possible variants of variable_standard_name to open the dataset using.

But yeah, it could be a nice feature to add to intake-esm for sure.

[anton-seaice]

Oh and the same / similar bevause for variable_long_name would be good for consistency