The classic example of cyclic references is where you have two classes A
and B
where A
has a reference to B
which has a reference to A
:
#include <memory>
#include <iostream>
struct B;
struct A {
std::shared_ptr<B> b;
~A() { std::cout << "~A()
"; }
};
struct B {
std::shared_ptr<A> a;
~B() { std::cout << "~B()
"; }
};
void useAnB() {
auto a = std::make_shared<A>();
auto b = std::make_shared<B>();
a->b = b;
b->a = a;
}
int main() {
useAnB();
std::cout << "Finished using A and B
";
}
If both references are shared_ptr
then that says A
has ownership of B
and B
has ownership of A
, which should ring alarm bells. In other words, A
keeps B
alive and B
keeps A
alive.
In this example the instances a
and b
are only used in the useAnB()
function so we would like them to be destroyed when the function ends but as we can see when we run the program the destructors are not called.
The solution is to decide who owns who. Lets say A
owns B
but B
does not own A
then we replace the reference to A
in B
with a weak_ptr
like so:
struct B {
std::weak_ptr<A> a;
~B() { std::cout << "~B()
"; }
};
Then if we run the program we see that a
and b
are destroyed as we expect.
Live demo
Edit: In your case, the approach you suggested looks perfectly valid. Take ownership away from A
and something else owns the A
s.
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…