Object slicing

In object-oriented programming, a subclass typically extends its superclass by defining additional member variables. If a superclass instance is assigned its value from a subclass instance, member variables defined in the subclass cannot be copied, since the superclass has no place to store them. This is a natural and unavoidable consequence of assignment by value from subclass objects. The term object slicing is sometimes used to refer to this aspect of assignment by value to a superclass instance.

Object slicing is also used to refer to a more subtle, problematic, case in which an object assignment by value appears to be to a superclass instance but is actually to a subclass instance. From the perspective of object memory layout, the member variables of the source instance can be thought of as having been "sliced off", leaving the corresponding member variables in the destination instance unchanged. It is this partial assignment (arguably a more apt term) that often surprises programmers and leads to unintended consequences.

Unexpected object slicing can happen in languages such as C++ in which assignment by value is not polymorphic. It is not possible in Java, which allows object assignment only by reference, or the D programming language, which allows object inheritance only through reference types.

For example, in C++:

struct A
{
    A(int a) : a_var(a) {}
    int a_var;
};

struct B : public A
{
    B(int a, int b) : A(a), b_var(b) {}
    int b_var;
};

B &getB()
{
    static B b(1, 2);
    return b;
}

void example()
{
    // Normal assignment by value to a
    A a(3);
    a = getB();
    // a.a_var == 1, b.b_var not copied to a

    B b2(3, 4);
    A &a2 = b2;
    // Partial assignment by value through reference to b2
    a2 = getB();
    // b2.a_var == 1, b2.b_var == 4!
}