SocketNotifier Class Reference
[Callback and Event Handling Interfaces]

Environment-Independent Socket Notification Interface. More...

#include <events.h>

Inheritance diagram for SocketNotifier:

Inheritance graph
[legend]
Collaboration diagram for SocketNotifier:

Collaboration graph
[legend]

List of all members.

Public Member Functions

virtual void SetEnabled (bool enable)=0
 Set state of event delivery.
template<typename TargT>
void Register (TargT *targp, TRet(TargT::*mfp)(TA1, TA2, TA3, TA4, TA5, TA6))
 Register a simple object and method to be called.
void Register (BaseT const &src)
 Copy the state of an identical callback object.
TRet operator() (TA1 a1, TA2 a2, TA3 a3, TA4 a4, TA5 a5, TA6 a6)
 Invoke the callback method.
bool Registered (void) const
 Test whether an object and method are registered to be called.
void Unregister (void)
 Clear the currently registered object and method.
template<typename TargT>
void Bind (TargT *objp, TRet(TargT::*mfp)(void))
 Register a method with a nonconforming signature to the callback object.


Detailed Description

Environment-Independent Socket Notification Interface.

SocketNotifier provides an abstract mechanism to receive socket activity notifications independent of any specific main event loop. SocketNotifier is meant to be as simple as possible and is nothing more than a specialized derivative of Callback.

The first parameter passed to the callback points to the SocketNotifier object initiating the event. The second parameter identifies the file handle on which the event occurred.

SocketNotifier derived objects are instantiated by DispatchInterface::NewSocket(). Clients will then use the Callback::Register() method to set their object and function to be invoked on socket activity.

SocketNotifier is inherently single threaded and not meant for event loops that support multithreading.


Member Function Documentation

virtual void SetEnabled ( bool  enable  )  [pure virtual]

Set state of event delivery.

This method allows readiness reporting from a SocketNotifier object to be temporarily disabled.

Parameters:
enable Set to true to receive read or write readiness callbacks, false otherwise.

void Register ( TargT *  targp,
TRet(TargT::*)(TA1, TA2, TA3, TA4, TA5, TA6)  mfp 
) [inline, inherited]

Register a simple object and method to be called.

Parameters:
targp The object to receive the method call
mfp Pointer to the method of the object to be invoked
If the method to receive the callback invocation matches the signature of the callback, as specified by template parameters, this method may be used to register the target method.

For example, if the callback has signature:

 Callback<void, int, int, bool, const char *>

The method to be registered must have signature:

 void Method(int, int, bool, const char *).

Nonconforming methods may also be registered as the target using Bind(). However, Register() results in lower overhead callback invocations, and its use is advised when possible.

TRet operator() ( TA1  a1,
TA2  a2,
TA3  a3,
TA4  a4,
TA5  a5,
TA6  a6 
) [inline, inherited]

Invoke the callback method.

Note:
If no method is registered, this will cause a segfault/access violation.
Returns:
The return value from the called method

void Bind ( TargT *  objp,
TRet(TargT::*)(void)  mfp 
) [inline, inherited]

Register a method with a nonconforming signature to the callback object.

In the spirit of boost::bind, this module provides a generic argument remapping mechanism for target methods. This makes it possible to bind methods to callbacks where the method's signature does not match that of the callback, as long as a useful argument mapping can be created. This mechanism allows values to be stored at the time the method is bound to the callback, and passed to the target function each time the callback is invoked.

Note:
To register methods with matching signatures and no desire to remap arguments, use Callback::Register() instead. It results in lower overhead callback invocations.
Parameters:
objp Target object to receive the callback
mfp Target method to receive the callback. The signature of the method need not conform to the signature of the callback, but it must have the same return type.
Parameter 3 and on are arguments to pass to objp->mfp, and must match the signature of mfp. Arguments will be saved in the callback object and passed to the method when the callback is invoked. Special keywords Arg1, Arg2, ... Arg6 can be used to identify parameters passed to the callback by its caller, as potential arguments to the nonconforming method.

As an untypical example, suppose we want to log messages to stdout, and prefix them with either "error" or "warning" depending on which of two objects the message is submitted to.

 class Target {
        void LogValue(const char *source, const char *value) {
                printf("%s: %d\n", source, value);
        }
 };

 Target tgt;
 Callback<void, const char *> error;
 Callback<void, const char *> warning;

 error.Bind(&tgt, &Target::LogValue, "Error", Arg1);
 warning.Bind(&tgt, &Target::LogValue, "Error", Arg1);

 error("This is an error\n");
 warning("This is a warning\n");

The output would be:

 Error: This is an error
 Warning: This is a warning

This method has one specific advantage over boost::bind. It does not allocate memory under any circumstances. It will not unpredictably fail or throw an exception at runtime. On the down side, it is quite primitive compared to boost::bind. It does not support pointers to nonmember functions. Also, its stored parameters reside inside the Callback object itself. The amount of space reserved for stored parameters is larger than that set by boost::function, and Callback objects take up more space. Additionally, if a binding request is made that would exceed the space reserved for stored parameters, a compile time error will be generated.


The documentation for this class was generated from the following file:
Generated on Fri Jan 9 05:58:39 2009 for libhfp by  doxygen 1.5.4