ADF Variable Defaults

[andrewgettelman]

@brianpm with help from @nusbaume led the charge and added the ability to add variable defaults to easily customize plotting (color tables, variable unit conversions, ranges, etc). I hooked this up with the map and zonal mean (line, haven't done the zonal mean lat-height yet) plotting.

So it's an easy exercise to play with new variables and get them looking 'pretty' in the code now. A good 'Easiest (green)' hackathon project for the next phases.

I made a google sheet for listing variables people want to look at: ADF Variable Priority

It would be good to get comments on this, and in particular: get a list of variables people want to look at. So please add key variables, and also which plots you want to look at (mostly it will apply to all the plots).

Especially from @cecilehannay : what would make the ADF useful? What variables do you need. Plot types (e.g. tables, Taylor diagram) could be useful too.

[brianpm]

Hijacking this old discussion, I'd like to raise a new point about adf_variable_defaults.yaml. Right now that file is >1200 lines, and with the PR #339, I think it is going to be more than 2000 lines.

I'd like to suggest that we break this file into individual files for each variable. Yes, that's a lot of file, but they would be very small, and we could just combine them at run time with a small function like:

def load_var_defaults():
    config = {}
    yaml_files = glob.glob("var_defaults/*.yml")
    for yaml_file in yaml_files:
        with open(yaml_file, "r") as f:
            config.update(yaml.safe_load(f))
    return config

This would be a small code change and would not impact the functionality of ADF.

[nusbaume]

I completely agree that splitting the current adf_variable_defaults.yaml file is a good idea, but I might vote to have a YAML file for every "category" of variable (e.g. each option listed in the category attribute in the defaults file), instead of one for every single variable. This way we can hopefully have a nice balance between having files that are too large versus files that may be too numerous.

Of course I am happy to be overruled if other people who are more active with the ADF think differently.