The title pretty much says it all. Basically, is it legal to do this:
class Base {
//stuff
}
class Derived: public Base {
//more stuff
}
vector<Base> foo;
Derived bar;
foo.push_back(bar);
Based on other posts I've seen, the following is okay, but I don't want to use pointers in this case because it's harder to make it thread safe.
vector<Base*> foo;
Derived* bar = new Derived;
foo.push_back(bar);
No, the Derived
objects will be sliced: all additional members will be discarded.
Instead of raw pointers, use std::vector<std::unique_ptr<Base> >
.
It's legal but suffers from object slicing. Basically, you'll have a vector of Base
objects. No polymorphism, type info will be lost for derived objects... It's as if you'd just be adding Base
objects to the vector.
You can use smart pointers instead.
vector<Base> foo;
Derived bar;
foo.push_back(bar);
This is equal to pushing Base object, because push_back
is declared like this:
void push_back ( const T& x );
So, compiler will do implicit downgrading conversion and do copy into vector memory pool.
No, it is not possible to contain Derived
inside vector<Base>
. They will be Base
.
If you add some virtual function to Base
then override it in Derived
, create Derived
object, push it into vector<Base>
and then call it from vector's new object, you will see that Base
implementation is called