Instance of an object within class definition

[RedUnderTheBed]

I want to have an instance of a class used within its own class definition. Such as:

class A{
private:
A a;

public:
A GetA(){return a;}
};

This is a simple thing to do in java with its mighty multipass compiling, but how could I achieve this in C++?

[JVene]

First, it's not because of the multipass compilation so much as the means by which objects are created in Java.

In C++, this is done with pointers. I'd suggest smart pointers (they operate something like Java's use of variables).

[Muthuveerappan]

First of all thanks for bringing up this question, I didn't even realize this wasn't possible.

As Jvene suggested, I was so used to declaring "a" as a pointer to the class "A" in such cases.

Question
-----------
Just wonder why C++ permits the declaration of a pointer to the same class while not an object of the class.

[RedUnderTheBed]

Ah yes, good point. I've actually realised that this isn't the issue that I'm having. I have two classes, each in their own header files. A uses B in its definition and B wants to use A in its definition. It seems the resolution for this is to have a dummy B class defined before A. So nevermind, thanks for your help.

[Russco]

because a pointer has a fixed size, 32 bits on a 32 bit machine and 64 bits on a 64 bit machine, whereas a full blown object has indeterminate size until the full declaration is seen.

[STLDude]

...I have two classes, each in their own header files. A uses B in its definition and B wants to use A in its definition. It seems the resolution for this is to have a dummy B class defined before A.

Look up forward declaration, here, here and here are a good start.

[Muthuveerappan]

because a pointer has a fixed size, 32 bits on a 32 bit machine and 64 bits on a 64 bit machine, whereas a full blown object has indeterminate size until the full declaration is seen.

Thanks Russco, I didn't know that. I think I understand better.

Look up forward declaration, here, here and here are a good start.

I don't think forward declaration might help in this case, correct me if I am wrong,

Forward declaration wouldn't solve the problem of incomplete type,
As Russco stated Objects would require the complete class specification, and for pointers forward declaration would be sufficient.

I have given below an example:

class ClassB;

class ClassA
{
    public:
        ClassB objB;
};

class ClassB
{
    public:
        ClassA objA;
};


int main()
{

    return(0);
}

Compilation Error:

test.cpp:6: error: field 'objB' has incomplete type

[laserlight]

I don't think forward declaration might help in this case, correct me if I am wrong,

Forward declaration wouldn't solve the problem of incomplete type,
As Russco stated Objects would require the complete class specification, and for pointers forward declaration would be sufficient.

You're right. Plus even if this was allowed, its meaning would be rather "interesting": a class A object owns a class A object which owns a class A object which...