The nature of the relationship between datasets and models is a many to many one. Thus my suggestion would be to put the emphasis at the level of model. Basically any one dataset can be linked to many models, and each model registers which datasets its linked to.
The idea is that later on, when doing the fit, the model knows against which data it has to fit.
Of course, as an alternative, we could keep datasets and models very independent of each other and give to the fit routine a model and a set of datasets.
The nature of the relationship between datasets and models is a many to many one. Thus my suggestion would be to put the emphasis at the level of model. Basically any one dataset can be linked to many models, and each model registers which datasets its linked to.
The idea is that later on, when doing the fit, the model knows against which data it has to fit.
Of course, as an alternative, we could keep datasets and models very independent of each other and give to the fit routine a model and a set of datasets.
What do you think @francescorandi ?