Question

I consider writing a class that behaves differently, depending on its construction method:

Example:

// #1 - Own an object
class MyClass {
    MyClass(std::string const& str) : m_str(str) {}

    std::string m_str;
}

// #2 - Share an object (Can be done by pointer as well)
class MyClass {
    MyClass(std::string& str) : m_str(str) {}

    std::string& m_str;
}

1. Is this a good idea to have both behaviors in the same class?
2. What is the best way to implement such a class?

Answer 1

Try this:

struct MyClass
{
    std::unique_ptr<std::string> dont_touch_me;
    std::string & str;

    explicit MyClass(std::string const & s)
    : dont_touch_me(new std::string(s))
    , str(*dont_touch_me)
    { }

    explicit MyClass(std::string & r)
    : str(r)
    { }
};

Now just make sure you always use MyClass::str for everything.