221

Can you make a std::shared_ptr point to an array? For example,

std::shared_ptr<int> sp(new int[10]);

If not, then why not? One reason I am already aware of is one can not increment/decrement the std::shared_ptr. Hence it cannot be used like a normal pointer to an array.

2
  • 2
    FWIT, you may also consider just using std::vector. You'd have to be careful to pass the array around using references so that you don't make copies of it. Syntax for accessing data is cleaner than shared_ptr, and resizing it is very very easy. And you get all the STL goodness should you ever want it. Commented Nov 30, 2012 at 11:57
  • 6
    If the size of the array is determined at compile time, you might also consider using std::array. It is almost the same as a raw array, but with proper semantics for use in most library components. Especially objects of that type are destroyed with delete, not delete[]. And unlike vector, it stores the data directly in the object, so you get no extra allocation. Commented Apr 19, 2015 at 13:23

2 Answers 2

335

With C++17, shared_ptr can be used to manage a dynamically allocated array. The shared_ptr template argument in this case must be T[N] or T[]. So you may write

shared_ptr<int[]> sp(new int[10]);

From n4659, [util.smartptr.shared.const]

  template<class Y> explicit shared_ptr(Y* p);

Requires: Y shall be a complete type. The expression delete[] p, when T is an array type, or delete p, when T is not an array type, shall have well-defined behavior, and shall not throw exceptions.
...
Remarks: When T is an array type, this constructor shall not participate in overload resolution unless the expression delete[] p is well-formed and either T is U[N] and Y(*)[N] is convertible to T*, or T is U[] and Y(*)[] is convertible to T*. ...

To support this, the member type element_type is now defined as

using element_type = remove_extent_t<T>;

Array elements can be access using operator[]

  element_type& operator[](ptrdiff_t i) const;

Requires: get() != 0 && i >= 0. If T is U[N], i < N. ...
Remarks: When T is not an array type, it is unspecified whether this member function is declared. If it is declared, it is unspecified what its return type is, except that the declaration (although not necessarily the definition) of the function shall be well formed.


Prior to C++17, shared_ptr could not be used to manage dynamically allocated arrays. By default, shared_ptr will call delete on the managed object when no more references remain to it. However, when you allocate using new[] you need to call delete[], and not delete, to free the resource.

In order to correctly use shared_ptr with an array, you must supply a custom deleter.

template< typename T >
struct array_deleter
{
  void operator ()( T const * p)
  { 
    delete[] p; 
  }
};

Create the shared_ptr as follows:

std::shared_ptr<int> sp(new int[10], array_deleter<int>());

Now shared_ptr will correctly call delete[] when destroying the managed object.

The custom deleter above may be replaced by

  • the std::default_delete partial specialization for array types

    std::shared_ptr<int> sp(new int[10], std::default_delete<int[]>());
    
  • a lambda expression

    std::shared_ptr<int> sp(new int[10], [](int *p) { delete[] p; });
    

Also, unless you actually need share onwership of the managed object, a unique_ptr is better suited for this task, since it has a partial specialization for array types.

std::unique_ptr<int[]> up(new int[10]); // this will correctly call delete[]

Changes introduced by the C++ Extensions for Library Fundamentals

Another pre-C++17 alternative to the ones listed above was provided by the Library Fundamentals Technical Specification, which augmented shared_ptr to allow it to work out of the box for the cases when it owns an array of objects. The current draft of the shared_ptr changes slated for this TS can be found in N4082. These changes will be accessible via the std::experimental namespace, and included in the <experimental/memory> header. A few of the relevant changes to support shared_ptr for arrays are:

— The definition of the member type element_type changes

typedef T element_type;

 typedef typename remove_extent<T>::type element_type;

— Member operator[] is being added

 element_type& operator[](ptrdiff_t i) const noexcept;

— Unlike the unique_ptr partial specialization for arrays, both shared_ptr<T[]> and shared_ptr<T[N]> will be valid and both will result in delete[] being called on the managed array of objects.

 template<class Y> explicit shared_ptr(Y* p);

Requires: Y shall be a complete type. The expression delete[] p, when T is an array type, or delete p, when T is not an array type, shall be well-formed, shall have well defined behavior, and shall not throw exceptions. When T is U[N], Y(*)[N] shall be convertible to T*; when T is U[], Y(*)[] shall be convertible to T*; otherwise, Y* shall be convertible to T*.

Sign up to request clarification or add additional context in comments.

11 Comments

+1, remark: There is also Boost's shared-array.
@tshah06 shared_ptr::get returns a pointer to the managed object. So you can use it as sp.get()[0] = 1; ... sp.get()[9] = 10;
ALT: std::shared_ptr<int> sp( new int[10], std::default_delete<int[]>() ); see also en.cppreference.com/w/cpp/memory/default_delete
@Jeremy If the size is known at compile time there's no need to write a class for that, std::shared_ptr<std::array<int,N>> should be enough.
Why does unique_ptr get that partial specialization but shared_ptr does not?
|
32

A possibly easier alternative that you might be able to use is shared_ptr<vector<int>>.

5 Comments

Yes it is. Or a vector is a superset of an array - it has the same in-memory representation (plus metadata) but is resizable. There aren't really any situations where you want an array but can't use a vector.
The difference, here, is that the vector size is anymore static, and the access to the data will be done with a double indirection. If performance is not the critical issue, this works, otherwise sharing an array may have its own reason.
Then you can probably use shared_ptr<array<int, 6>>.
The other difference is that it's slightly bigger and slower than a raw array. Generally not really an issue but let's not pretend that 1 == 1.1.
There are situations where the source of the data in the array means that it is unwieldily or unnecessary to convert to a vector; such as when getting a frame from a camera. (Or, that's my understanding, anyway)

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.