haiku/docs/user/app/Handler.dox

/*
 * Copyright 2007, Haiku, Inc. All Rights Reserved.
 * Distributed under the terms of the MIT License.
 *
 * Authors:
 * 		Niels Sascha Reedijk, niels.reedijk@gmail.com
 * 
 * Corresponds to:
 *		/trunk/headers/os/app/Handler.h  rev 22577
 *		/trunk/src/kits/app/Handler.cpp  rev 21332
 */

/*!
	\file Handler.h
	\brief Provides the BHandler class.
*/

///// Globals /////

/*!
	\def B_OBSERVE_WHAT_CHANGE
	\brief Internal.
*/


/*!
	\def B_OBSERVE_ORIGINAL_WHAT
	\brief Constant for a message data field in observer messages.

	If you have called one of the flavors of BHandler::StartWachting(), and
	you receive a notification, sometimes there can be send a BMessage to go
	with that notification. The message you receive is a copy of that message,
	but with the what constant set to \c B_OBSERVER_NOTICE_CHANGE. The original
	\c what constant of the transmitted data message is stored behind the
	label defined by this constant.
*/


/*!
	\var B_OBSERVER_OBSERVE_ALL
	\brief Parameter to BHandler::StartWatching().
	
	\note Specifying this parameter as the \a what value, leads to the same
		results as calling BHandler::StartWatchingAll().
*/


///// BHandler /////


/*!
	\class BHandler
	\ingroup app
	\brief Handles messages that are passed on by a BLooper.

	The BHandler class implements two important pieces of functionality. It
	provides the foundations for <b>handling messages</b>, and it serves as a
	<b>state machine</b> that sends out notifications of the state changes.

	The most common use of this class is to <b>handle messages</b>. Handlers
	can be tied to loopers, which are the objects that send and receive
	messages. As soon as a message is received, the looper passes through its
	list of associated handlers and tries them in a certain order until the
	message is handled, or the options are exhausted.

	You should know that a looper is a subclass of a handler, and as such,
	loopers can be self-contained and do not need additional handlers. In many
	cases, this construction will suffice. You will simply subclass the looper,
	override its MessageReceived() hook and handle the messages you receive. In
	some cases, you might opt in for a more ingenious construction. A
	real-world example is the interface kit. Within that kit, the windows are
	represented by a BLooper, and all the views and controls in that kit are
	derived from BHandler. If you put a control in a window, then whenever
	messages such as clicks are received, the window loops the handlers until
	there is a handler that is at the screen position the click was in. It is
	not unlikely that you will some day want to use this functionality of the
	API.

	If your handler is limited to a certain type of messages, you can set a
	filter that the looper will apply to your message before passing it on to
	your overridden MessageReceived() method. The BMessageFilter class provides
	the framework for the flexible filtering options, and using AddFilter() you
	can apply filters to this handler. Note that a filter object should only be
	applied to one handler. They cannot be shared.

	For more information on the handling chain, have a look at the
	documentation of the BLooper class.

	Using BHandler as a <b>state machine</b> is a second area of functionality.
	Since handlers process messages, and perform actions associated with those,
	they are the center of keeping track on the current state of things within
	an application. If you want to synchronize these states between different
	parts of your application, you could perform this manually by sending
	messages to the interested components, or you can use the more flexible
	approach with observers.

	Observers watch a certain state. A handler can track one or more different
	states. Each state is represented by a four byte constant - just like the
	\c what property of a message. Using the StartWatching() methods, you can
	register observers both within your team, and in other applications. As an
	argument of that method, you can supply the state you want to watch, or you
	can register an observer using StartWatchingAll() to watch all the states
	the handler tracks. When the handler needs to emit a state change, you can
	use SendNotices(). You can specify the exact state change, and some data
	that you want to be send to the observers. This data is in the form of the
	very flexible BMessage, as such you are almost free to pass anything you
	want.

	Whenever SendNotices() is called, all interested observers will receive a
	message of the \a B_OBSERVER_NOTICE_CHANGE type. Please note that the
	constant that is associated with the state itself is not transmitted. If
	you require this information, consider using the message that is passed
	on to describe the state change.

	BHandler is a part of the chain in the eloquent messaging structure. For a
	proper understanding of all its facets, have a look at the \ref app_messaging
	"messaging overview".
*/


/*!
	\fn BHandler::BHandler(const char* name = NULL)
	\brief Construct a new handler with a \a name.

	The newly constructed handler is not associated with a looper until you
	explicitly request this to happen. To associate this handler with a looper,
	use BLooper::AddHandler().
*/


/*!
	\fn BHandler::~BHandler()
	\brief Free the filters of this handler, as well as the list of observers.

	This method does not remove the handler from the looper to which this
	handler is associated. You should do this yourself, using
	BLooper::RemoveHandler().

	\warning This constructor does no type check whatsoever. Since you can pass
		any BMessage, you should - if you are not sure about the exact type -
		use the Instantiate() method, which does check the type.
*/


/*!
	\fn BArchivable *BHandler::Instantiate(BMessage *data)
	\brief Static method to instantiate a handler from an archived message.

	\return A pointer to the instantiated handler, or \c NULL if the \a data
		is not a valid archived BHandler object.
	\see BHandler(BMessage* data)
*/


///// Archiving /////

/*!
	\name Archiving

	BHandler inherits the BArchivable class, and as such implements support for
	archiving and unarchiving handlers.
*/


//! @{


/*!
	\fn BHandler::BHandler(BMessage* data)
	\brief Construct a handler from an archived message.

	This \a data has to be created using the BHandler::Archive() method.
	Note that only the name is stored. The filters, the associated looper and
	the observers are not stored, and should be manually added when you are
	using this object.
*/


/*!
	\fn status_t BHandler::Archive(BMessage *data, bool deep) const
	\brief Archive a handler to a message

	Currently, only the name is archived. The filters, the associated looper
	and the observers are not stored. 

	\param data The message to archive the object in.
	\param deep This parameter is ignored, as BHandler does not have children.
	\retval B_OK Archiving succeeded.
	\retval B_BAD_VALUE The \a data parameter is not a valid message.
	\see BHandler::Instantiate(BMessage *data)
*/


//! @}


///// The guts of BHandler /////


/*!
	\name Core Handler Functionality
*/


//! @{


/*!
	\fn void BHandler::MessageReceived(BMessage *message)
	\brief Handle a message that has been received by the associated looper.

	This method is reimplemented by subclasses. If the messages that have
	been received by a looper pass through the filters, then they end up in
	the MessageReceived() methods.

	The example below shows a very common way to handle message. Usually,
	this involves parsing the BMessage::what constant and then perform an
	action based on that. 

\code
void
ShowImageApp::MessageReceived(BMessage *message)
{
	switch (message->what) {
		case MSG_FILE_OPEN:
			fOpenPanel->Show();
			break;

		case B_CANCEL:
			// File open panel was closed,
			// start checking count of open windows.
			StartPulse();
			break;
		
		default:
			// We do not handle this message, pass it on to the base class.
			BApplication::MessageReceived(message); 
			break;
	}
}
\endcode

	If your handler cannot process this message, you should pass it on
	to the base class. Eventually, it will reach the base implementation,
	which will reply with \c B_MESSAGE_NOT_UNDERSTOOD.

	\attention If you want to keep or manipulate the \a message, have a
		look at BLooper::DetachCurrentMessage() to receive ownership of
		the message.

	\param message The message that needs to be handled.
*/


/*!
	\fn BLooper *BHandler::Looper() const
	\brief Return a pointer to the looper that this handler is associated with.

	\return If the handler is not yet associated with a looper, it will return
		\c NULL.
	\see BLooper::AddHandler()
	\see LockLooper()
*/


/*!
	\fn void BHandler::SetName(const char *name)
	\brief Set or change the name of this handler.
	\see Name()
*/


/*!
	\fn const char *BHandler::Name() const
	\brief Return the name of this handler.
	\see SetName()
*/


/*!
	\fn void BHandler::SetNextHandler(BHandler *handler)
	\brief Set the next handler in the chain that the message is passed on to
		if this \a handler cannot process it.

	This method has three requirements:
	-# This \a handler should belong to a looper.
	-# The looper needs to be locked. See LockLooper().
	-# The \a handler that you pass must be associated with the same looper.

	Failure to meet any of these requirements will result in your application
	crashing.

	By default, the handlers are chained in order that they were associated to
	a looper with BLooper::AddHander().

	\see NextHandler()
*/


/*!
	\fn BHandler *BHandler::NextHandler() const
	\brief Return the next hander in the chain to which the message is passed
		on.
	\see SetNextHandler()
*/


//! @}


///// Message Filtering /////


/*!
	\name Message Filtering
*/


//! @{


/*!
	\fn void BHandler::AddFilter(BMessageFilter *filter)
	\brief Add a filter as a prerequisite to this handler.

	If the handler is associated with a looper, this looper needs to be locked
	in order for this operation to succeed.

	Note that the filter is not copied, rather a pointer to the filter is
	stored. As such, you need to make sure that the filter object exists as
	long as it is added to this handler.

	\see RemoveFilter(), SetFilterList()
*/


/*!
	\fn bool BHandler::RemoveFilter(BMessageFilter *filter)
	\brief Remove a filter from the filter list.

	If the handler is associated with a looper, this looper needs to be locked
	in order for this operation to succeed.

	Note that the filter is not deleted, merely removed from the list. You need
	to take care of the memory yourself.

	\retval true The filter was in the filter list and is removed.
	\retval false The filter was not found in the filter list.

	\see AddFilter(), FilterList()
*/


/*!
	\fn void BHandler::SetFilterList(BList* filters)
	\brief Set the internal list of filters to \a filters.

	If the handler is associated with a looper, this looper needs to be locked
	in order for this operation to succeed.

	The internal list will be replaced with the new list of \a filters. All the
	existing filters will be \b deleted.

	\see AddFilter(), FilterList()
*/


/*!
	\fn BList *BHandler::FilterList()
	\brief Return a pointer to the list of filters.

	\return A pointer to the list of filters. Do not manipulate the list of
		filters directly, but use the methods provided by this class, in order
		to maintain internal consistency.

	\see AddFilter(), RemoveFilter(), SetFilterList().
*/


//! @}


///// Locking /////


/*!
	\name Locking

	This class provides some utility functions to look the looper associated
	with this handler.
*/


//! @{


/*!
	\fn bool BHandler::LockLooper()
	\brief Lock the looper associated with this handler.

	\retval true The looper is locked.
	\retval false There was an error acquiring the lock.

	\see LockLooperWithTimeout(), UnlockLooper()
*/


/*!
	\fn status_t BHandler::LockLooperWithTimeout(bigtime_t timeout)
	\brief Lock the looper associated with this handler, with a time out value.

	\param timeout The time to wait for acquiring the lock in microseconds. You
		may also use B_INFINITE_TIMEOUT, in which this method will wait as long
		as it takes to acquire the lock.

	\retval B_OK Locking succeeded.
	\retval B_BAD_VALUE This handler is not associated with a looper (anymore).
	\retval B_TIMED_OUT The time specified in \a timeout has passed without
		locking the looper.

	\see LockLooper(), UnlockLooper()
*/


/*!
	\fn void BHandler::UnlockLooper()
	\brief Unlock the looper.
*/


//! @}


///// Scripting //////


/*!
	\name Scripting
*/


//! @{


/*!
	\fn BHandler * BHandler::ResolveSpecifier(BMessage *msg, int32 index,
		BMessage *specifier, int32 form, const char *property)
	\brief Determine the proper handler for a scripting message.
*/


/*!
	\fn status_t BHandler::GetSupportedSuites(BMessage *data)
	\brief Reports the suites of messages and specifiers that derived classes
		understand.
*/


//! @}


///// Observing /////


/*!
	\name Observing

	Handlers can function as state machines, which emit messages to observers
	when the state changes. Use the following methods to subscribe to these
	notifications.

	Note that there is a semantic difference between the two StartWatching()
	methods. The overloaded method that accepts a BHandler, expects as
	argument an \a observer that watches this handler. The method that
	accepts a BMessenger, expects a \a target that emits the state changes
	to this handler.
*/


//! @{


/*!
	\fn status_t BHandler::StartWatching(BMessenger target, uint32 what)
	\brief Subscribe this handler to watch a specific state change of a
		\a target.

	Use this method to subscribe messengers to watch state changes in this
	handler, this also means that observers from other teams can be
	subscribed.

\code
// Handler B watches Handler A
BHandler A, B;
BMessenger messengerA(&A)

B.StartWatching(messengerA, kNetworkConnection);
\endcode

	\param target The messenger from which the notifications would be
		received.
	\param what The state that needs to be watched.
	\return During the call of this method, a notification will be transmitted
		using the \a target. If this works, then this method will return
		\c B_OK.

	\see StartWatchingAll(BMessenger), StopWatching(BMessenger, uint32)
*/


/*!
	\fn status_t BHandler::StartWatchingAll(BMessenger target)
	\brief Subscribe this handler to watch a \a target for all events.

	This method performs the same task as StartWatching(BMessenger, uint32),
	but it will subscribe to all the state changes the \a target knows. 

	\see StartWatching(BMessenger, uint32), StopWatchingAll(BMessenger)
*/


/*!
	\fn status_t BHandler::StopWatching(BMessenger target, uint32 what)
	\brief Unsubscribe this handler from watching a specific state.

	This method will unsubscribe this handler from watching a specific event
	in a \a target.

	\see StartWatching(BMessenger, uint32)
*/


/*!
	\fn status_t BHandler::StopWatchingAll(BMessenger target)
	\brief Unsubscribe this handler from watching all states.

	This method will unsubscribe the \a target from watching all state changes.

	\see StartWatchingAll(BMessenger)
*/


/*!
	\fn status_t BHandler::StartWatching(BHandler* observer, uint32 what)
	\brief Subscribe an \a observer for a specific state change of this handler.

	Use this method to subscribe observers to watch this handler. State changes
	of this handler that match the \a what argument, will be sent.

\code
// Handler B wants to observe Handler A
BHandler A, B;

A.StartWatching(&B, kNetworkConnection);
\endcode

	Since pointers to handlers can only
	exist in the local namespace, have a look at
	StartWatching(BMessenger, uint32) for inter-team watching. 

	\param observer The observer for this handler.
	\param what The state that needs to be watched.
	\return During the call of this method, a notification will be transmitted
		using the \a observer. If this works, then this method will return
		\c B_OK.

	\see StartWatchingAll(BHandler*), StopWatching(BHandler*, uint32)	
*/


/*!
	\fn status_t BHandler::StartWatchingAll(BHandler* observer)
	\brief Subscribe an \a observer for a all state changes.

	This method performs the same task as StartWatching(BHandler, uint32),
	but it will subscribe the \a observer to all the state changes this handler
	tracks.

	\see StartWatching(BHandler*, uint32), StopWatchingAll(BHandler*)	
*/


/*!
	\fn status_t BHandler::StopWatching(BHandler* handler, uint32 what)
	\brief Unsubscribe an observer from watching a specific state.

	This method will unsubscribe the \a handler from watching a specific event.

	\see StartWatching(BHandler*, uint32)
*/


/*!
	\fn status_t BHandler::StopWatchingAll(BHandler* handler)
	\brief Unsubscribe an observer from watching all states.

	This method will unsubscribe the \a handler from watching all state changes.

	\see StartWatchingAll(BHandler*)
*/


//! @}


///// State changes /////


/*!
	\name Emitting State Changes

	If your handler functions as a state machine, and it has observers (which
	subscribed using the StartWatching() method), you can emit these state
	changes.
*/


//! @{


/*!
	\fn void BHandler::SendNotices(uint32 what, const BMessage *msg)
	\brief Emit a state change to the observers.

	The actual state (specified by \a what) will not be transmitted. This is
	merely for internal bookkeeping. It is not entirely unimaginable that you
	still want to inform the observers of what actually took place. You can
	use the \a msg to transmit this, and any other data you want. Note that the
	message will be copied and slightly altered: the \c what member of the
	message will be \c B_OBSERVER_NOTICE_CHANGE, and the \c what constant you
	specified will be stored in the #B_OBSERVE_ORIGINAL_WHAT label.

	\param what The identifier of the state.
	\param msg Any data associated with the state change. You retain ownership
		of this data, so make sure you dispose it when you are done.
*/


/*!
	\fn bool BHandler::IsWatched() const
	\brief Check if there are any observers watching this handler.
*/


//! @}