Opaque pointer

In computer programming, an opaque pointer is a special case of an opaque data type, a datatype declared to be a pointer to a record or data structure of some unspecified type.

Opaque pointers are present in several programming languages including Ada, C, C++ and Modula-2.

If the language is strongly typed, programs and procedures that have no other information about an opaque pointer type T can still declare variables, arrays, and record fields of type T, assign values of that type, and compare those values for equality. However, they will not be able to de-reference such a pointer, and can only change the object's content by calling some procedure that has the missing information.

Opaque pointers are a way to hide the implementation details of an interface from ordinary clients, so that the implementation may be changed without the need to recompile the modules using it. This benefits the programmer as well since a simple interface can be created, and most details can be hidden in another file. This is important for providing binary code compatibility through different versions of a shared library, for example.

This technique is described in Design Patterns as the Bridge pattern. It is sometimes referred to as "handle classes", the "Pimpl idiom" (for "pointer to implementation idiom"), "Compiler firewall idiom" or "Cheshire Cat", especially among the C++ community.

Examples

Ada

package Library_Interface is

   type Handle is limited private;

   -- Operations...

private
   type Hidden_Implementation;    -- Defined in the package body
   type Handle is access Hidden_Implementation;
end Library_Interface;

The type Handle is an opaque pointer to the real implementation, that is not defined in the specification. Note that the type is not only private (to forbid the clients from accessing the type directly, and only through the operations), but also limited (to avoid the copy of the data structure, and thus preventing dangling references).

package body Library_Interface is

   type Hidden_Implementation is record
      ...    -- The actual implementation can be anything
   end record;

   -- Definition of the operations...

end Library_Interface;

These types are sometimes called "Taft types"—named after Tucker Taft, the main designer of Ada 95—because they were introduced in the so-called Taft Amendment to Ada 83.

C

/* obj.h */

struct obj;

/*
 * The compiler considers struct obj an incomplete type. Incomplete types
 * can be used in declarations.
 */

size_t obj_size(void);

int obj_setid(struct obj *, int);

int obj_getid(struct obj *, int *);

/* obj.c */

#include "obj.h"

struct obj {
    int id;
};

/*
 * The caller will handle allocation.
 * Provide the required information only
 */

size_t
obj_size(void)
{
    return sizeof(struct obj);
}

int
obj_setid(struct obj *o, int i)
{
    if (o == NULL) return -1;
    o->id = i;
    return 0;
}

int
obj_getid(struct obj *o, int *i)
{
    if (o == NULL || i == NULL) return -1;
    *i = o->id;
    return 0;
}

This example demonstrates a way to achieve the information hiding (encapsulation) aspect of Object-Oriented Programming using the C language. If someone wanted to change the declaration of struct obj, it would be unnecessary to recompile any other modules in the program that use the obj.h header file unless the API was also changed.

C++

//header file:
class Handle {
public:
    Handle();                         // Constructor
    Handle(const Handle&);            // Copy constructor
    Handle(Handle&&);                 // Move constructor
    Handle& operator=(const Handle&); // Copy assignment operator
    ~Handle();                        // Destructor
    // Other operations...

private:
    struct CheshireCat;               // Not defined here
    CheshireCat* smile;               // Handle
};

//CPP file:
#include "handle.h"

struct Handle::CheshireCat {
    int a;
    int b;
};

Handle::Handle()
    : smile(new CheshireCat()) {
    // do nothing
}

Handle::Handle(const Handle& other)
    : smile(new CheshireCat(*(other.smile))) {
    // do nothing
}

Handle::Handle(Handle&& other) 
    : smile(0)
{
    std::swap(smile, other.smile);
}

Handle& Handle::operator=(const Handle &other) {
    *smile = *(other.smile);
    return *this;
}

Handle::~Handle() {
    delete smile;
}

One type of opaque pointer commonly used in C++ class declarations is the d-pointer. The d-pointer is the only private data member of the class and points to an instance of a struct. Named by Arnt Gulbrandsen of Trolltech, this method allows class declarations to omit private data members, except for the d-pointer itself. The result: (a) more of the class implementation is hidden from view; (b) adding new data members to the private struct does not affect binary compatibility; (c) the header file containing the class declaration only needs to #include those other files needed for the class interface, rather than for its implementation. One side benefit is that compilations are faster because the header file changes less often. The d-pointer is heavily used in the Qt and KDE libraries.

C#

See Private class data pattern

See also

• Application binary interface
• Handle (computing)
• Programming idiom

External links

• The Pimpl idiom
• Compilation Firewalls
• The Fast Pimpl Idiom
• D-Pointers — KDE TechBase
• When you "XOR the pointer with a random number"[1][2], the result is a "really opaque" pointer [3].
• Making Pimpl Easy, Vladimir Batov