[MamboJambo]

Hello to all,


I would like to create a class representing a model. Since there will be different models, it needs to be runTime selectable. Each model, will need a subModel defined for the system of equations which should also be runTime selectable.


The main idea is illustrated in the image bellow:





I was thinking of doing this based on the structure of the viscosity model.

Foam::autoPtr<Foam::viscosityModel> Foam::viscosityModel::New
(
    const word& name,
    const dictionary& dict,
    const volVectorField& U,
    const surfaceScalarField& phi
)
{
    const word modelType(dict.get<word>("transportModel"));

    Info<< "Selecting incompressible transport model " << modelType << endl;

    auto cstrIter = dictionaryConstructorTablePtr_->cfind(modelType);

    if (!cstrIter.found())
    {
        FatalIOErrorInLookup
        (
            dict,
            "viscosityModel",
            modelType,
            *dictionaryConstructorTablePtr_
        ) << exit(FatalIOError);
    }

    return autoPtr<viscosityModel>(cstrIter()(name, dict, U, phi));
}

So inside my createFields, I would have a autoPtr<BaseClass> and inside the base class a member of type autoPtr<subModelBaseClass>.

Is this a suitable strategy? Do you know of a better approach? Does the use of autoPtr [unique_ptr] have any influence in parallel processing?

Thanks