1. A good API will provide easy to use interfaces but also provide hard to miss-use interfaces. Usually the later one is more fundamental than that of the previous one. Consider you want to write a data class, there are thousands ways to write it. Here is one example:
class Date { public: Date(int month, int day, int year); ... }; Date(3, 4, 2014); Date(4, 3, 2014);
Both are OK, but only one is logical right.
Under such situation, a better way to implement that is to constrict clients what they can do and force them to right direction:
class Month { public: static Month Jan() { return Month(1); } ... private: explicit Month(int m); }; Data d(Month::Mar(), Day(10), Year(2011));
2. Let's see another example, suppose your interface returns a dynamic allocated resource and will be deleted after all. There are chances that a programmer will forget to delete it or delete it multi times. So, return a smart pointer would be a great idea.
std::tr1::shared_ptr<Investment> createInvestment() { std::tr1::shared_ptr<Investment> retVal(static_cast<Investment>(0), getRidOfInvestment()); retVal = ...; //let retVal point to right object return retVal; }
Besides, smart pointer have another advantage that it will use default or assigned deleter, and you don't need to worry about "cross-DLL problems".