Update global_mean_timeseries.py

[justin-richling]

Currently the time series plot is crashing when trying to check dimensions of the zm data. This is because the data was averaged before checking dims with pf.zm_validate_dims, which will fail because the 'lat' dim will no longer be present after spatial average. This PR will change order of checking zm dims first then it will average the data if it passes the dim check.

Also update LENS2 file grab since the files in the directory are not named exactly the same so add wildcard for searching.

[brianpm]

I think this line:
has_lat_ref, has_lev_ref = pf.zm_validate_dims(ref_ts_da)
could still be problematic if zm_validate_dims returns None.

We can either fix zm_validate_dims (return (None, None), for example), or we might just change this to:

valdims = pf.zm_validate_dims(ref_ts_da)
if valdims is not None:
  has_lat_ref, has_lev_ref = valdims
else:
  has_lat_ref, has_lev_ref = False, False # I *think* this is what we'd expect here ... double check

[justin-richling]

I think your'e right, if there is a scenario where there is not a lat dim, this would be a problem. I guess either change is sufficient because we want to shut it down if it's missing lat dim. @nusbaume any thoughts on which would be preferable? I can't seem to see an advantage of one over the other.

Either way, we should also probably put a check for when has_lat is None or False with a continue call too so that it doesn't try to get to ref_ts_da_ga = pf.spatial_average(ref_ts_da, weights=None, spatial_dims=None), obviously for the test case loop too.

[nusbaume]

@justin-richling If other scripts that are calling zm_validate_dims are checking for just a single None then I would vote to implement @brianpm's option. However, if none of them are doing this check then I would probably vote to modify zm_validate_dims itself to just return False for both variables in this case, as that should keep the interface the same everywhere while still protecting us from the situation where the latitude dimension or input data is invalid.

Anyways, that's my two cents. Hope that helps!

[justin-richling]

All test cases work as intended