NeoFOAM workflow

[greole]

This issue is intended for the discussion and planning of a NeoFOAM workflow. Here, workflow means how to execute simulations and parameter studies. This would be useful for a continuous benchmarking setup. We have several options:

• snakeMake
• obr (maybe rename obr to a better name)
• ...

@HenningScheufler and I agreed on having a python based approach since the data analysis ecosystem is very exetensive.

Ideally one can have a yaml/json file which encodes a workflow/ parameterstudy. In the case of obr it would look like

case:
    type: CaseOnDisk # other options can be gitrepo, tutorialcase
    solver: pimpleFoam
    origin: ${{yaml.location}}/../basicSetup
variation:
    - operation: fvSolution
      schema: "linear_solver/{solver}{preconditioner}{executor}"
      values: # several other options exists to define parameter sweeps, akin to generator expressions
        - set: solvers/p
          preconditioner: none
          solver: GKOCG
          forceHostBuffer: 1
          verbose: 1
          executor: ${{env.GINKGO_EXECUTOR}}
        - set: solvers/p
          preconditioner: IC
          solver: GKOCG
          forceHostBuffer: 1
          verbose: 1
          executor: ${{env.GINKGO_EXECUTOR}}

[bevanwsjones]

I agree that Python + yaml is a very good lang + config file system to work with. I looked into snakeMake previously and it is cool, but I find the mixing of languages etc made it a big meh. Could be a skill thing though.