Difference in behavior while using dynamic_cast with reference and pointers

Question

I was checking the behavior of dynamic_cast and found that when it fails, std::bad_cast exception is thrown only if the destination is a reference type. If the destination is a pointer type then no exception is thrown from the cast. This is my sample code:

class A
{
    public:
     virtual ~A()
     {
     }
};

class B : public A
{
};

int  main()
{
    A* p = new A;

    //Using reference
    try
    {
    B& b = dynamic_cast<B&>(*p);
    }
    catch(std::bad_cast exp)
    {
    std::cout<<"Caught bad cast\n";
    }

    //Using pointer
      try
    {
    B* pB = dynamic_cast<B*>(p);

    if( pB == NULL)
    {
     std::cout<<"NULL Pointer\n";
    }
    }
    catch(std::bad_cast exp)
    {
    std::cout<<"Caught bad cast\n";
    }

    return 0;
}

Output is "Caught bad cast" and "NULL pointer". Code is compiled using VS2008. Is this the correct behavior ? If yes, then why there is a difference?

Answer 1

Yes, it is. Because dynamic_cast can't return NULL for a failed reference cast, an exception is the only way out.

That is, a reference can't be NULL, so there is nothing suitable to return.

Answer 2

Yes, this is correct behaviour. The reason is that you can have a null pointer, but not a null reference - any reference has to be bound to an object.

So when dynamic_cast for a pointer type fails it returns a null pointer and the caller can check for that, but when it fails for a reference type it can't return a null reference, so an exception is the only reasonable way to signal a problem.

Answer 3

Yes, 5.2.7/9

The value of a failed cast to pointer type is the null pointer value of the required result type. A failed cast to reference type throws bad_cast (18.5.2).

Answer 4

See the C++ Standard, section 5.2.7/9:

9 The value of a failed cast to pointer type is the null pointer value of the required result type. A failed cast to reference type throws bad_cast (18.5.2).

As to why - these are Stroustrup's words from the D & E book, section 14.2.2:

I use a reference cast when I want an assumption about a reference type checked and consider it a failure for my assumption to be wrong. If instead I want to select among plausible alternatives, I use a pointer cast and test the result.