Owning instance of object vs owning pointer

Question

Let's say I have two class A and B. There are 2 ways in which class B can use A.

First:

class B
{
  A *a;
}

Second:

class B
{
  A a;
}

Why most of the C++ libraries prefer to use First version as compared to Second. What could be the disadvantage of using Second approach. Does it relate to Stack vs Heap assignment? Please clarify.

I am not sure why you think the first version is better, I usually use the second. the first one, you should be carefull with class allocation, and deletion, — aah134, Aug 08 '13 at 18:40
"Why most of the C++ libraries prefer to use First version as compared to Second" So you looked into > 50% of all existing C++ libraries? Joking apart, they're different. E.g. [Why not use pointers for everything?](http://stackoverflow.com/q/1064325/420683) and many many other answers. — dyp, Aug 08 '13 at 18:43
It depends on how A will be used. For example, if you had a getter function on B for A that returned a pointer to it: `A* a = b->getA();` and then b gets deleted, then in the second example, you'd have a dangling pointer `*a` — crush, Aug 08 '13 at 18:45
You could also store a reference to the instance of class A. — Derek, Aug 08 '13 at 18:46
Five comments and no one has complained yet that [a raw pointer `A*` shouldn't be an owning pointer](http://stackoverflow.com/a/417635/420683)? I'm baffled ;) — dyp, Aug 08 '13 at 18:59

score 8 · Accepted Answer · answered Aug 08 '13 at 18:42

Some advantages of owning an instance (class B { A a; };):

No need to worry about creation and destruction of a because it happens automatically.
No need to worry that a might be a dangling or null pointer.
Memory locality: a lives where instances of B live. If you have a large array of Bs and access each B's A in turn, this could make a significant speed difference.
Memory efficiency: no storage for a pointer is needed.

To make a huge sweeping generalization, one could say that this approach is faster and safer.

Some advantages of owning a pointer (class B { A *a; };):

Polymorphism: a can actually point to a subclass of A.
a can be reassigned without needing to copy an instance of A.
a can live independently of B or even be owned by another object entirely.
a can be null, freeing up memory if it's not always needed.

To make another huge sweeping generalization, one could say that this approach is more flexible.

Polymorphism can also be achieved using a reference. All of the advantages of the "pointer" category plus some of the "non-pointer" category can be achieved using a smart pointer. — dyp, Aug 08 '13 at 18:45

Owning instance of object vs owning pointer

1 Answers1