Thread: Help designing data structure

Originally Posted by kotoko

Knowing this, do smart pointers make more sense? (Yes a database makes even more but I want to put that aside for now)

It depends. If the actual object is guaranteed not to go out of scope*, a raw pointer is sufficient. Otherwise, use a smart pointer; managing memory directly (via new/delete) is *never* acceptable- violating RAII is the surest path to undefined behavior! Anyway, I'd recommend using std::tr1::shared_ptr or the like.

*That can be trickier than it sounds. For example, if you store your objects in an std::vector, a pointer to one of it's elements could become invalidated once another object is added to the vector. On the other hand, an std::list would work fine. Point is, choose your data structure carefully, and be sure to RTFM.

Originally Posted by kotoko

@tabstop I think we might be having what we call in Portuguese a deaf men conversation. What I'm saying is I do not want to duplicate. What I understand from your suggestion is that I will have to duplicate. What I'm asking is is this really what you are suggesting (to have duplicate).

I have said, and will continue to say, exactly the opposite. My suggestions have been built around not duplicating any objects. (Or, I suppose, at least are consistent with them -- you may have to consistently take the pointer option when a choice was presented.)