Question

Without discussing whether or not it’s a good idea, What type of drawbacks (performance or otherwise) would one face if they were to encapsulate the built-in C++ data types into their own classes. For instance, similar to Java and C#, the int data type would have its own class called Int, overloading it with inline operators. Same with Single, Double, Long, etc.

Answer 1

There are no benefits. You cannot achieve identical behavior, nor performance.

---

Drawbacks:

• Performance will never be faster than built-in types. The best you’ll get is a class where everything is inlined to built-in type code, although what’s the point of that?
• Lots more code for no purpose
• Virtual calls have overhead
• Cannot achieve identical behavior, such as with operators (Casting primarily)
• Non-trivial constructors (Not necessarily in C++11)
• Not supported by C++ features such as template parameters
• Confusing. Nobody does this.

---

This is potentially useful when you don’t want to overload a Math class with many types when simply Integer or Real would suffice.

You have Java stuck in your head.

C++ takes the template approach:

template<typename A_type, typename B_type>
auto math_operation(A_type a, B_type b) -> decltype(a + b * 2) {
  return a + b * 2;
}

Now you have a function that works on any type that supports the correct operators. This will work for built-in types, and classes such as Int128.