Add map between MPAS-Ocean and Omega variables

[xylar]

Checklist

• Developer's Guide has been updated
• API documentation in the Developer's Guide (api.md) has any new or modified class, method and/or functions listed
• Documentation has been built locally and changes look as expected
• Testing comment in the PR documents testing used to verify the changes

[cbegeman]

Thanks for getting to this so quickly!

If I remember correctly, in our conversation yesterday we concluded that for IO streams we would just manually produce different sections in the yaml file for each model. Thus, we don't need a public function that has an MPAS-O var name as an input and the Omega var name as output. We would only be using these functions to rename initial state and output dataset variables.

[cbegeman]

From a brief look at the code, this looks good. Let me know when you would like me to test it, as I see this is marked draft.

[xylar]

@cbegeman, I'm still working on this. I was about to write you to have a look but then I realized I'm not happy with it. I then revised it and I'm still not happy, but I'm getting there.

[xylar]

Okay, @cbegeman, this is now ready for you to take another quick look. The main difference from an hour ago is that the map_input_dataset() and map_output_dataset() methods don't require the calling code to worry about which model is being run. This should be more convenient, as it mean we can hopefully just call the methods right before writing datasets out or right after reading them in. If it looks okay, I'll update the manufactured solution test to use the new methods.

[xylar]

Are we aware of others?

[xylar]

Darn it! I realized immediately after trying to use this in manufactured_solution that this won't work because it's precisely for ocean steps that are not model steps that we need this mapping. I'll think about it some more and give it another go.

[xylar]

@cbegeman, I now have a solution I'm happy with. I need to test it out, hopefully tomorrow.

In the meantime, let me know if it looks okay to you.

[cbegeman]

I think this looks great! I like the IOStep idea. Somehow the map_to_model_dataset and map_from_model_dataset names aren't super intuitive to me but given that we'll mostly be calling the read/write method it doesn't matter much

[xylar]

Somehow the map_to_model_dataset and map_from_model_dataset names aren't super intuitive to me

@cbegeman, I'd appreciate suggestions.

[xylar]

Somehow the map_to_model_dataset and map_from_model_dataset names aren't super intuitive to me

@cbegeman, how about map_to_native_model_vars() and map_from_native_model_vars()?

[xylar]

Testing

I successfully ran the pr suite and the manufactured_solution test case on Chrysalis with MPAS-Ocean.

I set up the manufactured_solution test case with Omega and successfully ran the init/200km step. The forward/200km step fails because output.nc is not produced (as usual). I confirmed that SshCell occurs in initial_condition.nc instead of ssh, as expected.

[cbegeman]

@cbegeman, how about map_to_native_model_vars() and map_from_native_model_vars()?

Sure, that's clearer to me. Or map_dataset_from_model_vars

[xylar]

@cbegeman, I could go either way. I guess I don't see either a obviously clearer so I'm tempted to stick with what I have.

Anything else you'd like to see in terms of code changes or testing before this goes in?

[xylar]

I'm not actually sure this is correct. We will see once Omega is actually writing output.

[cbegeman]

Nothing more from me. Feel free to merge as-is. Thanks!

[xylar]

Thanks @cbegeman!