should cf-xarray change the meaning of .keys?

[dcherian]

After merging #104, I remembered that xarray Datasets have .keys method which returns a list of variable names in Dataset.data_vars.

#104 added .cf.keys that replaced .cf.get_valid_keys (i.e. .cf.keys returns a List of valid "special keys"). This changes cf_xarray's usual position of not changing the "meaning" of any existing xarray method.

I am unsure of whether we actually want to do this.

1. Something like .cf.get_valid_keys() is useful, I think; though it may be less useful now that we have .cf.axes and friends. Is there a better name for this function?
2. Should .cf.keys() instead return a list of "special keys" in ds.data_vars? I don't know if this is that useful either...

Thoughts?

[kthyng]

If you don't want to overlap with xarray commands, then probably shouldn't use cf.keys() anymore. I lean slightly toward skipping it altogether since the individual categories could be combined together to get the same thing and I don't see it as being used enough to justify a shortcut.