This article explains how to implement connection points (formerly known as OLE connection points) using the MFC classes CCmdTarget and CConnectionPoint.
In the past, the Component Object Model (COM) defined a general mechanism (IUnknown::QueryInterface) that allowed objects to implement and expose functionality in interfaces. However, a corresponding mechanism that allowed objects to expose their capability to call specific interfaces was not defined. That is, COM defined how incoming pointers to objects (pointers to that object's interfaces) were handled, but it did not have an explicit model for outgoing interfaces (pointers the object holds to other objects' interfaces). COM now has a model, called connection points, that supports this functionality.
A connection has two parts: the object calling the interface, called the source, and the object implementing the interface, called the sink. A connection point is the interface exposed by the source. By exposing a connection point, a source allows sinks to establish connections to itself (the source). Through the connection point mechanism (the IConnectionPoint interface), a pointer to the sink interface is passed to the source object. This pointer provides the source with access to the sink's implementation of a set of member functions. For example, to fire an event implemented by the sink, the source can call the appropriate method of the sink's implementation. The following figure demonstrates the connection point just described.
MFC implements this model in the CConnectionPoint and CCmdTarget classes. Classes derived from CConnectionPoint implement the IConnectionPoint interface, used to expose connection points to other objects. Classes derived from CCmdTarget implement the IConnectionPointContainer interface, which can enumerate all of an object's available connection points or find a specific connection point.
For each connection point implemented in your class, you must declare a connection part that implements the connection point. If you implement one or more connection points, you must also declare a single connection map in your class. A connection map is a table of connection points supported by the ActiveX control.
The following examples demonstrate a simple connection map and one connection point. The first example declares the connection map and point; the second example implements the map and point. Note that CMyClass must be a CCmdTarget-derived class. In the first example, code is inserted in the class declaration, under the protected section:
The BEGIN_CONNECTION_PART and END_CONNECTION_PART macros declare an embedded class, XSampleConnPt (derived from CConnectionPoint), that implements this particular connection point. If you want to override any CConnectionPoint member functions or add member functions of your own, declare them between these two macros. For example, the CONNECTION_IID macro overrides the CConnectionPoint::GetIID member function when placed between these two macros.
In the second example, code is inserted in the control's implementation file (.cpp file). This code implements the connection map, which includes the connection point, SampleConnPt:
If your class has more than one connection point, insert additional CONNECTION_PART macros between the BEGIN_CONNECTION_MAP and END_CONNECTION_MAP macros.
Finally, add a call to EnableConnections in the class's constructor. For example:
Once this code has been inserted, your CCmdTarget-derived class exposes a connection point for the ISampleSink interface. The following figure illustrates this example.
Usually, connection points support "multicasting" — the ability to broadcast to multiple sinks connected to the same interface. The following example fragment demonstrates how to multicast by iterating through each sink on a connection point:
This example retrieves the current set of connections on the SampleConnPt connection point with a call to CConnectionPoint::GetConnections. It then iterates through the connections and calls ISampleSink::SinkFunc on every active connection.