Question

Possible Duplicate:
When should functions be member functions?

Are there situations when it is better to define functions outside of classes or should you use static functions inside a class?

Answer 1

There are some cases that have to be non-member functions:

• operator overloads can’t be static member functions (they can be non-static member functions), and in particular most binary operator overloads work better as non-member functions because you get implicit conversion on the LHS and RHS for free operator overloads but only on the RHS for member operator overloads.

• std::swap is conventionally called as using std::swap; swap(x,y); so that classes can “overload” it via ADL. Implementing swap conventionally therefore requires a non-member function, if only as a wrapper that calls a member function. The same would be true of other functions designed to be ADL-overloaded.

• Technically, static member functions can’t have “C” linkage and therefore aren’t suitable as callbacks when interfacing with other languages. In practice, C++ ABIs tend to make static functions call-compatible with C, provided of course that their parameters and return type exist in C.

So far I can think of one case that has to be a static member function rather than a free function:

• You want to use access specifier protected. Private static member functions are normally pointless, because it’s normally better to define a free function with internal linkage in the .cpp file, where nobody else can even see it let alone call it. But I suppose occasionally you’d want one.

Beyond that it’s really a style question, there isn’t very much practical difference between a static member function and a free function.