automate the addition of default cf attributes?

[jerabaul29]

Should we automate the addition of default cf attributes when these are missing? Or should this be something the user has to call in manually? I.e. the addition of the attributes as done by:

ds = ds.traj.assign_cf_attrs()

Could look for the existence of the requested attributes in the class __init__ when a Traj object is created for example, and call this automatically if the attributes are not there? Any thoughts about this @gauteh @knutfrode ? :)

[knutfrode]

Maybe this could be added by default if we make a ds.traj.to_netcdf() (but still with possibility to skip)

[jerabaul29]

Agree that would be a good place to have such a call. Maybe having this call overwrite previous attributes would be good, so that also in case the dataset has been updated / sliced, the attributes are correct when writing to disk?

Would it make sense to call this actually quite often, for example both at opening if the attributes are missing (so the attributes are always present, they are convenient to always have available), and always even if present from before at writing to disk (so that we never write outdated attributes)? :)

[gauteh]

I don't think we should modify the dataset in init without any action by the user. In theory it needs to be updated in a lot of cases, seltime, gridtime, isel(trajectory=0) (which we don't control). Maybe in to_netcdf is a good place. I was also wondering if this should be normalize() method, which also fixes other inconsistencies (like missing role attributes etc).