I am not referring to the punk next door whom at 2 years old has too much to say but none can be understood; not even the links on your social network account whose friends are really just the ones I want to see; or the people I usually get drank with. No, this is not about them. C++ talks of access and scope that are available when something is a friend [Stroustrup, 1997].
Why am I using it?
There is a limited framework that I have come across where the only easy way to hide controls is to make them delete themselves, and to be able to reuse such controls, I have to create another. Reuse here is an inappropriate term but it is used because in the context I am referring, they are class members. An issue arises then when a member calls delete this; and the compiler does not guarantee to set that pointer to null. Compilers today would be more standards compliant but this particular port of GCC I am using is not.
When am I using it?
There is no way then for checking for P’s validity if it were a member of a class C and when P deletes itself.
class P{};
class C {
P* p;
public:
void m()
{
if (p) /* is undefined */
{/*...*/}
}
//...
};
Usually, this situation warrants setting P to null after deleting it. However, if this should happen in one of P’s methods this = NULL; is semantically incorrect. Making C::p public is easier but it will be exposing data unnecessarily.
Here comes friend. Declaring P to be a friend will make C’s private parts accessible to P:
class C {
friend class P;
P* p;
public:
void m();
//...
};
Now I can do something like this:
class P {
C* c;
public:
void doSomething()
{
//...
c->p = 0; // c->p is actually this
delete this;
}
};
Should you use it too?
Honestly, I don’t know if what I am doing is right. It works for me and as I have said previously there is no other easy way. Well, there might be… 😉
References:
[Stroustrup, 1997] Bjarne Stroustrup: The C++ Programming Language
Walk The Dog 