Error when trying to load only 'U' or 'V' from an llc dataset

I've been trying to run the following code to extract 'U' or 'V' from the llc4320 dataset.

import xmitgcm.llcreader as llcreader

model = llcreader.ECCOPortalLLC4320Model()
ds_vel = model.get_dataset(varnames=['U'], type='latlon', iter_start=10368, iter_stop=10369)

If varnames is set to either ['U'] or ['V'] an error message is received. If it is set to ['U', 'V'] the function runs as expected. This isn't a problem if loading other variables. on their own such as 'Eta' or 'W'.

The error message is below:

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-13-9a90c8e0ac18> in <module>
----> 1 ds_vel = model.get_dataset(varnames=['U'], type='latlon', iter_start=10368, iter_stop=10369)

~/anaconda3/envs/xmitgcm_env/lib/python3.8/site-packages/xmitgcm-0.4.1+20.g7c5305c-py3.8.egg/xmitgcm/llcreader/llcmodel.py in get_dataset(self, varnames, iter_start, iter_stop, iter_step, k_levels, k_chunksize, type, read_grid, grid_vars_to_coords)
    727 
    728         transformer = data_transformers[type]
--> 729         data = transformer(data_facets, _VAR_METADATA)
    730 
    731         # separate horizontal and vertical grid variables

~/anaconda3/envs/xmitgcm_env/lib/python3.8/site-packages/xmitgcm-0.4.1+20.g7c5305c-py3.8.egg/xmitgcm/llcreader/llcmodel.py in _all_facets_to_latlon(data_facets, meta)
    354             mate = meta[vname]['attrs']['mate']
    355             vector_pairs.append((vname, mate))
--> 356             vnames.remove(mate)
    357         except KeyError:
    358             pass

ValueError: list.remove(x): x not in list

There is also no problem if type is set to 'faces'

This is an interesting issue.

The challenge here is that U and V are actually tightly coupled together in the LLC4320 internal data storage format. When transforming to the latlon grid, we mix together U and V. So it's not possible to get one without the other.

I think the best path forward would be to add a quick check in get_dataset to ensure that all the metric vector pairs are present and raise an error if not. Would that satisfy you?

Ah I hadn't realised how U and V were coupled in that way. I think a check in get_dataset would be really useful in helping users identify what's going on!

The place to add this check would be be right after this:

xmitgcm/xmitgcm/llcreader/llcmodel.py

Lines 309 to 310 in 7c5305c

    
           all_vector_components = [inner for outer in (vector_pairs + metric_vector_pairs) 
        
                                    for inner in outer]

Would you feel comfortable implementing this yourself and making a pull request? That is certainly the fastest way to get the feature you desire! 😉

Yes, I'll have a crack at it somepoint this week 👍

	all_vector_components = [inner for outer in (vector_pairs + metric_vector_pairs)
	for inner in outer]