Config mapping issue with 2-level config options

[cbegeman]

In #435, we attempt to use replacements for two 2-level config options in Omega https://github.com/cbegeman/polaris/blob/df452e5cdd64281db14575c5152335018840c74a/polaris/tasks/ocean/single_column/forward.py#L143-L150

Only the first option appears in the config replacements (using quiet=False here:

polaris/polaris/model_step.py

Line 555 in 05586c3

quiet = True

[xylar]

Thanks @cbegeman. I'll take a look as soon as I can.

[xylar]

I have been able to reproduce the problem and I'm working on a fix.

[xylar]

Okay, this problem may be tricky to fix well. MPAS-Ocean had a requirement that namelist options had to be unique, regardless of their section. The ability to update model config options based on the option alone (and not the section) is based on this understanding.

Omega has opted not to have unique model config option names, meaning that after mapping, Omega config options also need to be able to identify their section(s).

We also always had in mind that Omega model config options would become the "default" at some point, and the mapping would be from Omega to MPAS-Ocean rather than the other way. With this change, would not be meaningful to make calls to the add_model_config_options() method.

I will think about a solution that is compatible with these longer-term concerns.

[xylar]

In #451, I decided not to address the more general issue of how to map from Omega to MPAS-Ocean when config option names are not unique. That solution just solves the MPAS-Ocean --> Omega mapping issue for now, with minimal disruption to existing code.