Add to_xgcm_grid_dataset function to help read into XGCM

[mgrover1]

It would be helpful to have a helper function to prepare the data for XGCM

[jthielen]

Or, another option could be altering the Dataset metadata to meet the CF & COMODO conventions so that xgcm can handle it natively.

[mgrover1]

We could use something like the following

def to_xgcm(ds):
    """
    Converts an XGCM Grid
    """
    _required_coords = ['XLAT_M', 'XLAT_C', 'XLONG_M', 'XLONG_C']
    
    if not set(_required_coords).issubset(list(ds.coords)):
        raise ValueError(f'Missing Required Coordinates: {_required_coords}')
        
    grid = xgcm.Grid(ds, coords={"X":{"inner":"XLAT_M",
                                      "outer":"XLAT_C"},
                                 "Y":{"inner":"XLONG_M",
                                      "outer":"XLONG_C"}
                                }
                    )
    return grid

This does require those coordinates to be included in the dataset...

[lpilz]

This doesn't work for all projections though, right? Since XGCM interpolates the data based on the coordinates, I think that you have to be careful with anything other than lat/lon runs. E.g. for lambert runs, IMO one would have to first convert the XLAT*/XLONG* coordinates to the lambert grid and then use this grid with XGCM.

Also, XLAT_M and XLAT_C are WPS/met_em specific coordinates, are they not?

[jthielen]

This doesn't work for all projections though, right? Since XGCM interpolates the data based on the coordinates, I think that you have to be careful with anything other than lat/lon runs. E.g. for lambert runs, IMO one would have to first convert the XLAT*/XLONG* coordinates to the lambert grid and then use this grid with XGCM.

Also, XLAT_M and XLAT_C are WPS/met_em specific coordinates, are they not?

I concur that that using the lats and lons only works for a lat/lon projection. Since derived dimension coords are a needed feature anyway, I strongly lean towards just using CF & COMODO attributes conventions (which is what I've worked towards in #14) instead of directly providing an xgcm.Grid from xwrf.

Also, unsure on XLAT_M and XLAT_C...I think we need to collect more sample WRF files alongside user stories so we know the full extent of what we're trying to handle!