Setup Variables (umbrella)

[dalonsoa]

          Ok, so I'm booking myself Thursday and Friday next week to work on this. My plan would be:

Step 1 (one PR)

• [x] Start from the simplified version created by @davidorme
• [x] Add a step somewhere in the Config class after the required models have been pulled out of the configuration to populate the initialised_by, updated_by, and used_by lists. The models might need updating.
• [x] Run some validation based initially on the following conditions, raising an error if they are broken:
  • [x] No variable is initialised by two models
  • [x] There is no variable used or updated that has not been initialised
  • [x] No two models update the same variable (or they do, I don't know if that is a use case)

Step 2 (another PR, after the above is merged)

• [ ] Update the get_model_sequence function to use this information, as well as (or in place of?) the explicit dependencies already in place, to try to come up with the right sequence.
• [x] Update the models so they initialise the variables when they are initialised and the registry of initialised variables is populated.

Step 3 (yet another PR)

• Update the models so they use the information from this registry -> TBC by each model at its own pace

I'm sure I'm missing something and I will need to improvise, but is there anything obvious I'm forgetting?

_Originally posted by @dalonsoa in https://github.com/ImperialCollegeLondon/virtual_rainforest/issues/371#issuecomment-1934300630_

[davidorme]

@dalonsoa Where are we on this meta issue? I think it's all done apart from the get_model_sequence replacement?

[dalonsoa]

Yes, that's all that is left, aside from using the variables, but that's more for each individual model to adopt.

I'll see if I can tackle the get_model_sequence before the end of this week.