Discussion related to adopting a physical constants dictionary

[grantfirl]

We’ve been asked to adopt a standardized physical constants dictionary in the past: https://github.com/ESCOMP/PhysicalConstantsDictionary

• YAML file containing sets of self-consistent physical constants
• Each set is documented with references etc.

Can have more than one YAML file potentially corresponding to host models?

Pros:
More systematic and consistent way of setting constants
Provides way to standardize another aspect of models
CCPP-like in its approach (in terms of metadata, host agnosticity)

Cons:
Physical constants should be shared between hosts and physics, so shouldn’t this be up to hosts to adopt?
Should we encourage hosts to adopt?
Implementation overhead (more complicated than just setting constants in a file)
Is this being supported at all? Would we be adopting the project and do we have mandate/desire/time?

[ligiabernardet]

During the CCPP Visioning workshop we decided it is not a good idea for the CCPP to adopt the Constants Dictionary. The reason is that the CCPP is supposed to get physical constants from the host for consistency. It is up to the host if they want to adopt this dictionary. In other words, the dictionary has value, but it should not be imported via CCPP but via hosts.

It was also discussed that, for schemes that can also be run as subcomponents or otherwise standalone, it’s convenient for schemes to have their own constants – but should still pass through argument list from whatever code is calling the main entrypoint subroutine.