BUG: Problems handling data consisting of a single time slice

[spencerkclark]

There are numerous problems caused (in many parts of the code) if we load in data that has only one time value; I've run into this in trying to construct a test case for #129 with a file consisting of a single monthly mean.

I think it mostly has to do with the fact that in that case time is treated as a scalar coordinate and not an indexable dimension. Since there are numerous places where we try to do da.sel or da.isel on the time dimension throughout the code, we run into a lot of problems.

@spencerahill is there a way to enforce that a coordinate is always indexable (in other words make sure that it is always in ds.dims even if it only has one value)?

[spencerahill]

What immediately comes to mind for me is for us to just check upon loading if there is only a single time value.

I.e. so data_loader._load_data_from_disk would become:

    ...
    apply_preload_user_commands(file_set)
    ds =xr.open_mfdataset(file_set, preprocess=rename_grid_attrs,
                             concat_dim=internal_names.TIME_STR,
                             decode_cf=False)
    return _ensure_indexable_time(ds)

def  _ensure_indexable_time(ds):
    """"Force dataset to have indexable time coordinate.

    The primary use-case is for datasets with a single time-value, which
    results in a scalar coordinate that is therefore not indexable.
    """
    if <insert needed bool here>:
        return <insert needed logic here>
    else:
        return ds

Do you think something like that would work?

[spencerahill]

Unless this becomes a major can of worms, let's get this into the v0.1.1 bugfix release; it's a not-too-uncommon use-case to have a single time value.

[spencerahill]

Closed by #135