Question

Assume there is a method which returns a std::set:

std::set<string> MyClass::myMethod() const
{
   std::set<string> result;
   // ... some code to fill result ...
   return result;
}

In the caller side we can store result of myMethod in two ways:

void MyClass::test()
{
const std::set<string>  s1 = myMethod();  // 1
const std::set<string>& s2 = myMethod();  // 2 (using reference)
// ... some code to use s1 or s2 ...
}

My questions are:

• Is there any difference between them?
• Which way is better and
  efficient?

Answer 1

Theoretically, the second version is better because it doesn’t require the extra copy.

In practice, RVO will likely kick in so they should be about the same.