问题
I just knew std::enable_shared_from_this
form this link.
But after reading the code below, I don't know when to use it.
try {
Good not_so_good;
std::shared_ptr<Good> gp1 = not_so_good.getptr();
} catch(std::bad_weak_ptr& e) {
// undefined behavior (until C++17) and std::bad_weak_ptr thrown (since C++17)
std::cout << e.what() << '\n';
}
The code above is "not so good" because there is no existing shared_ptr
before calling getptr()
. So the good thing should be:
std::shared_ptr<Good> gp1 = std::make_shared<Good>(); // having a shared_ptr at the beginning
std::shared_ptr<Good> gp2 = gp1->getptr();
However, if I have already had a shared_ptr
object, why don't I just simply code like this: std::shared_ptr<Good> gp2 = gp1;
, meaning that I don't need std::enable_shared_from_this
at all.
In my opinion, using std::enable_shared_from_this
is to make sure that more than one shared_ptr
objects have the same control block so that we can avoid the double-delete problem. But if I must remind myself to create a shared_ptr
at the beginning, why don't I just remind myself to use shared_ptr
object to create a new one, instead of using raw pointer?
回答1:
The hint about when std::enable_shared_from_this<T>
is useful is in its name: when yielding objects based on some requests it may be necessary to return a pointer to an object itself. If the result should be a std::shared_ptr<T>
it becomes necessary to return such a pointer from within a member function where there is generally no std::shared_ptr<T>
accessible.
Having derived from std::enable_shared_from_this<T>
provides a way to get hold of a std::shared_ptr<T>
given just a pointer of type T
. Doing so does, however, assume that the object is already managed via a std::shared_ptr<T>
and it would create mayhem if the object is allocated on the stack:
struct S: std::enable_shared_from_this<S> {
std::shared_ptr<S> get_object() {
return this->shared_from_this();
};
}
int main() {
std::shared_ptr<S> ptr1 = std::make_shared<S>();
std::shared_ptr<S> ptr2 = ptr1->get_object();
// ...
}
In a realistic scenario there is probably some condition under which a std::shared_ptr<T>
to the current object is returned.
回答2:
There're some use case which you can't use the template std::shared_ptr<T>
like opaque pointer.
In that case, it's useful to have this:
In some_file.cpp
struct A : std::enable_shared_from_this<A> {};
extern "C" void f_c(A*);
extern "C" void f_cpp(A* a) {
std::shared_ptr<A> shared_a = a->shared_from_this();
// work with operation requires shared_ptr
}
int main()
{
std::shared_ptr<A> a = std::make_shared<A>();
f_c(a.get());
}
In some_other.c
struct A;
void f_cpp(struct A* a);
void f_c(struct A* a) {
f_cpp(a);
}
回答3:
Let's say I want to represent a computation tree. We'll have an addition represented as a class deriving from expression with two pointers to expressions, so an expression can be evaluated recursively. However, we need to end the evaluation somewhere, so let's have numbers evaluate to themselves.
class Number;
class Expression : public std::enable_shared_from_this<Expression>
{
public:
virtual std::shared_ptr<Number> evaluate() = 0;
virtual ~Expression() {}
};
class Number : public Expression
{
int x;
public:
int value() const { return x; }
std::shared_ptr<Number> evaluate() override
{
return std::static_pointer_cast<Number>(shared_from_this());
}
Number(int x) : x(x) {}
};
class Addition : public Expression
{
std::shared_ptr<Expression> left;
std::shared_ptr<Expression> right;
public:
std::shared_ptr<Number> evaluate() override
{
int l = left->evaluate()->value();
int r = right->evaluate()->value();
return std::make_shared<Number>(l + r);
}
Addition(std::shared_ptr<Expression> left, std::shared_ptr<Expression> right) :
left(left),
right(right)
{
}
};
Live on Coliru
Note that the "obvious" way of implementing Number::evaluate()
with return std::shared_ptr<Number>(this);
is broken because it will result in double delete.
来源:https://stackoverflow.com/questions/41364700/when-should-we-use-stdenable-shared-from-this