ODE solvers uses NULL to represent unconstrained domains

[Akash504-ai]

ODE solvers currently use std::vector<Polytope*> to represent domains, where a NULL pointer is used to indicate the unconstrained domain R^d.

This mixes the domain semantic with pointer validity and makes solver behavior hard to reason. In several places also solver steps proceed differently depending on whether this is a NULL pointer without exposing this information to the caller.

We should avoid using NULL as a sentinel, instead we should make the domain representation explicit. The handling of the unconstrained domain should live inside the corresponding polytope or domain abstraction not in the solver logic.

[arnavsharma990]

This makes sense, using NULL as a sentinel mixes semantic meaning with pointer validity and makes solver logic harder to reason about.

Would it be reasonable to introduce an explicit Domain abstraction (like i mean a small wrapper or tagged type distinguishing unconstrained vs polytope domain) and migrate the solver logic gradually?

I’d be interested in exploring a minimal refactor here if this direction aligns with the intended design.

[Akash504-ai]

Thanks for the suggestion @arnavsharma990 ,I agree that introducing an explicit Domain abstraction is the right longterm direction.
However since this affects solver architecture and touches core design assumptions, it would be good to first align with the maintainers before starting implementation.
I’m also waiting for guidance from the mentors on how they’d like to approach this. Once we have clarity on scope and design direction, we can move forward in a structured way.