The complexity of models can be assessed in different ways. We can look at the complexity in terms of computational time it takes for simulation. We can also look at the complexity in terms of the underlying program size. There is a trade-off between these two forms and it is often unclear where the optimum is. How we perform this trade-off is determined by the way we choose to idealize the underlying physical system. To become better modelers, we will hence revisit the familiar schemes of idealization and then investigate a new approach that is favorable for many applications and offers new opportunities for code generation.