How to use the xgcm.models.mitgcm.utils.read_mds function in xgcm

"""

        fname_base = prefix

        # some special logic is required for grid variables
        if prefix in self._all_grid_variables:
            # grid variables don't have an iteration number suffix
            iternum = None
            # some grid variables have a different filename than their varname
            if 'filename' in self._all_grid_variables[prefix]:
                fname_base = self._all_grid_variables[prefix]['filename']
        else:
            assert iternum is not None

        # get a dict of variables and their data
        vardata = read_mds(os.path.join(self.dirname, fname_base), iternum,
                           endian=self.endian)
        for vname, data in vardata.items():
            # we now have to revert to the original prefix once the file is read
            if fname_base != prefix:
                vname = prefix
            try:
                metadata = (self._all_grid_variables[vname]
                            if vname in self._all_grid_variables
                            else self._all_data_variables[vname])
            except KeyError:
                if self._ignore_unknown_vars:
                    # we didn't find any metadata, so we just skip this var
                    continue
                else:
                    raise KeyError("Couln't find metadata for variable %s "
                                   "and `ignore_unknown_vars`==False." % vname)