Threads

All the changes we've made in HelloWeb4 have to do with setting up a separate thread of execution to make the text of our applet blink. Java is a multithreaded language, which means there can be many threads running at the same time. A thread is a separate flow of control within a program. Conceptually, threads are similar to processes, except that unlike processes, multiple threads share the same address space, which means that they can share variables and methods (but they have their own local variables). Threads are also quite lightweight in comparison to processes, so it's conceivable for a single application to be running hundreds of threads concurrently.

Multithreading provides a way for an application to handle many different tasks at the same time. It's easy to imagine multiple things going on at the same time in an application like a Web browser. The user could be listening to an audio clip while scrolling an image, and in the background the browser is downloading an image. Multithreading is especially useful in GUI-based applications, as it can improve the interactive performance of these applications.

Unfortunately for us, programming with multiple threads can be quite a headache. The difficulty lies in making sure routines are implemented so they can be run by multiple concurrent threads. If a routine changes the value of a state variable, for example, then only one thread can be executing the routine at a time. Later in this section, we'll examine briefly the issue of coordinating multiple thread's access to shared data. In other languages, synchronization of threads can be an extremely complex and error-prone endeavor. You'll see that Java gives you a few simple tools that help you deal with many of these problems. Java threads can be started, stopped, suspended, and prioritized. Threads are preemptive, so a higher priority thread can interrupt a lower priority thread when vying for processor time. See Chapter 6, Threads for a complete discussion of threads.

The Java run-time system creates and manages a number of threads. I've already mentioned the AWT thread, which manages repaint() requests and event processing for GUI components that belong to the java.awt package. A Java-enabled Web browser typically has at least one separate thread of execution it uses to manage the applets it displays. Until now, our example has done most of its work from methods of the Applet class, which means that is has been borrowing time from these threads. Methods like mouseDragged() and actionPerformed() are invoked by the AWT thread and run on its time. Similarly, our init() method is called by a thread in the Web browser. This means we are somewhat limited in the amount of processing we do within these methods. If we were, for instance, to go into an endless loop in our init() method, our applet would never appear, as it would never finish initializing. If we want an applet to perform any extensive processing, such as animation, a lengthy calculation, or communication, we should create separate threads for these tasks.

The Thread Class

As you might have guessed, threads are created and controlled as Thread objects. We have added a new instance variable, blinkThread, to our example to hold the Thread that handles our blinking activity:

Thread blinkThread; 

An instance of the Thread class corresponds to a single thread. It contains methods to start, control, and stop the thread's execution. Our basic plan is to create a Thread object to handle our blinking code. We call the Thread's start() method to begin execution. Once the thread starts, it continues to run until we call the Thread's stop() method to terminate it.

But Java doesn't allow pointers to methods, so how do we tell the thread which method to run? Well, the Thread object is rather picky; it always expects to execute a method called run() to perform the action of the thread. The run() method can, however, with a little persuasion, be located in any class we desire.

We specify the location of the run() method in one of two ways. First, the Thread class itself has a method called run(). One way to execute some Java code in a separate thread is to subclass Thread and override its run() method to do our bidding. In this case, we simply create an instance of this subclass and call its start() method.

But it's not always desirable or possible to create a subclass of Thread to contain our run() method. In this case, we need to tell the Thread which object contains the run() method it should execute. The Thread class has a constructor that takes an object reference as its argument. If we create a Thread object using this constructor and call its start() method, the Thread executes the run() method of the target object, rather than its own. In order to accomplish this, Java needs to have a guarantee that the object we are passing it does indeed contain a compatible run() method. We already know how to make such a guarantee: we use an interface. Java provides an interface named Runnable that must be implemented by any class that wants to become a Thread.

The Runnable Interface

The second technique I described for creating a Thread object involved passing an object that implements the Runnable interface to the Thread constructor. The Runnable interface specifies that the object contains a run() method that takes no arguments and returns no value. This method is called automatically when the system needs to start the thread.

Sticking with our technique for implementing our applet in a single class, we have opted to add the run() method for blinkThread to our HelloWeb4 class. This means that HelloWeb4 needs to implement the Runnable interface. We indicate that the class implements the interface in our class declaration:

public class HelloWeb4 extends Applet
    implements MouseMotionListener, ActionListener, Runnable {...}

At compile time, the Java compiler checks to make sure we abide by this statement. We have carried through by adding an appropriate run() method to our applet. Our run() method has the task of changing the color of our text a couple of times a second. It's a very short routine, but I'm going to delay looking at it until we tie up some loose ends in dealing with the Thread itself.

start( ) and stop( )

Now that we know how to create a Thread to execute our applet's run() method, we need to figure out where to actually do it. The start() and stop() methods of the Applet class are similar to init(). The start() method is called when an applet is first displayed. If the user then leaves the Web document or scrolls the applet off the screen, the stop() method is invoked. If the user subsequently returns, the start() method is called again, and so on. Unlike init(), start() and stop() can be called repeatedly during the lifetime of an applet.

The start() and stop() methods of the Applet class have absolutely nothing to do with the Thread object, except that they are a good place for an applet to start and stop a thread. An applet is responsible for managing the threads that it creates. It would be considered rude for an applet to continue such tasks as animation, making noise, or performing extensive calculations long after it's no longer visible on the screen. It's common practice, therefore, to start a thread when an applet becomes visible and stop it when the applet is no longer visible.

Here's the start() method from HelloWeb4:

public void start() { 
    if ( blinkThread == null ) { 
        blinkThread = new Thread(this); 
        blinkThread.start(); 
    } 
} 

The method first checks to see if there is an object assigned to blinkThread; recall that an uninitialized instance variable has a default value of null. If not, the method creates a new instance of Thread, passing the target object that contains the run() method to the constructor. Since HelloWeb4 contains our run() method, we pass the special variable this to the constructor to let the thread know where to find the run() method it should run. this always refers to our object. Finally, after creating the new Thread, we call its start() method to begin execution.

Our stop() method takes the complimentary position:

public void stop() { 
    if ( blinkThread != null ) { 
        blinkThread.stop(); 
        blinkThread = null; 
    } 
} 

This method checks to see if blinkThread is empty. If not, it calls the thread's stop() method to terminate its execution. By setting the value of blinkThread back to null, we have eliminated all references to the thread object we created in the start() method, so the garbage collector can dispose of the object.

run( )

Our run() method does its job by setting the value of the variable blinkState. We have added blinkState, a boolean value, to represent whether we are currently blinking on or off:

boolean blinkState; 

The setColor() line of our paint() method has been modified slightly to handle blinking. The call to setColor() now draws the text in white when blinkState is true:

gc.setColor( blinkState ? Color.white : currentColor() ); 

Here we are being somewhat terse and using the C-like ternary operator to return one of two alternate color values based on the value of blinkState.

Finally, we come to the run() method itself:

public void run() { 
    while ( true ) { 
        blinkState = !blinkState; 
        repaint(); 
        try { 
            Thread.sleep(500); 
        } 
        catch (InterruptedException e ) { 
        } 
    } 
} 

At its outermost level, run() uses an infinite while loop. This means the method will run continuously until the thread is terminated by a call to the controlling Thread object's stop() method.

The body of the loop does three things on each pass:

• Flips the value of blinkState to its opposite value using the not operator, !.

• Calls repaint() so that our paint() method can have an opportunity to redraw the text in accordance with blinkState.

• Uses a try/catch statement to trap for an error in our call to the sleep() method of the Thread class. sleep() is a static method of the Thread class. The method can be invoked from anywhere and has the effect of putting the current thread to sleep for the specified number of milliseconds. The effect here is to give us approximately two blinks per second.

Exceptions

The try/catch statement in Java is used to handle special conditions called exceptions. An exception is a message that is sent, normally in response to an error, during the execution of a statement or a method. When an exceptional condition arises, an object is created that contains information about the particular problem or condition. Exceptions act somewhat like events. Java stops execution at the place where the exception occurred, and the exception object is said to be thrown by that section of code. Like events, an exception must be delivered somewhere and handled. The section of code that receives the exception object is said to catch the exception. An exception causes the execution of the instigating section of code to abruptly stop and transfers control to the code that receives the exception object.

The try/catch construct allows you to catch exceptions for a section of code. If an exception is caused by a statement inside of a try clause, Java attempts to deliver the exception to the appropriate catch clause. A catch clause looks like a method declaration with one argument and no return type. If Java finds a catch clause with an argument type that matches the type of the exception, that catch clause is invoked. A try clause can have multiple catch clauses with different argument types; Java chooses the appropriate one in a way that is analogous to the selection of overloaded methods.

If there is no try/catch clause surrounding the code, or a matching catch clause is not found, the exception is thrown up the call stack to the calling method. If the exception is not caught there, it's thrown up another level, and so on until the exception is handled. This provides a very flexible error-handling mechanism, so that exceptions in deeply nested calls can bubble up to the surface of the call stack for handling. As a programmer, you need to know what exceptions a particular statement can generate, so methods in Java are required to declare the exceptions they can throw. If a method doesn't handle an exception itself, it must specify that it can throw that exception, so that the calling method knows that it may have to handle it. See Chapter 4, The Java Language for a complete discussion of exceptions and the try/catch clause.

So, why do we need a try/catch clause around our sleep() call? What kind of exception can Thread's sleep() method throw and why do we care about it, when we don't seem to check for exceptions anywhere else? Under some circumstances, Thread's sleep() method can throw an InterruptedException, indicating that it was interrupted by another thread. Since the run() method specified in the Runnable interface doesn't declare it can throw an InterruptedException, we must catch it ourselves, or the compiler will complain. The try/catch statement in our example has an empty catch clause, which means that it handles the exception by ignoring it. In this case, our thread's functionality is so simple it doesn't matter if it's interrupted. All of the other methods we have used either handle their own exceptions or throw only general-purpose exceptions that are assumed to be possible everywhere and don't need to be explicitly declared.

A Word About Synchronization

At any given time, there can be a number of threads running in the Java interpreter. Unless we explicitly coordinate them, these threads will be executing methods without any regard for what the other threads are doing. Problems can arise when these methods share the same data. If one method is changing the value of some variables at the same time that another method is reading these variables, it's possible that the reading thread might catch things in the middle and get some variables with old values and some with new. Depending on the application, this situation could cause a critical error.

In our HelloWeb examples, both our paint() and mouseDrag() methods access the messageX and messageY variables. Without knowing the implementation of our particular Java environment, we have to assume that these methods could conceivably be called by different threads and run concurrently. paint() could be called while mouseDrag() is in the midst of updating messageX and messageY. At that point, the data is in an inconsistent state and if paint() gets lucky, it could get the new x value with the old y value. Fortunately, in this case, we probably would not even notice if this were to happen in our application. We did, however, see another case, in our changeColor() and currentColor() methods, where there is the potential for a more serious "out of bounds" error to occur.

The synchronized modifier tells Java to acquire a lock for the class that contains the method before executing that method. Only one method can have the lock on a class at any given time, which means that only one synchronized method in that class can be running at a time. This allows a method to alter data and leave it in a consistent state before a concurrently running method is allowed to access it. When the method is done, it releases the lock on the class.

Unlike synchronization in other languages, the synchronized keyword in Java provides locking at the language level. This means there is no way that you can forget to unlock a class. Even if the method throws an exception or the thread is terminated, Java will release the lock. This feature makes programming with threads in Java much easier than in other languages. See Chapter 6, Threads for more details on coordinating threads and shared data.

Whew! Now it's time to say goodbye to HelloWeb. I hope that you have developed a feel for the major features of the Java language, and that this will help you as you go on to explore the details of programming with Java.