Precursor: The Event Dispatch Thread (EDT).
Swing is single-threaded. What does this mean?
All processing in a Swing program begins with an event. The EDT is a thread that processes these events in a loop along the following lines (but more complicated):
class EventDispatchThread extends Thread {
Queue<AWTEvent> queue = ...;
void postEvent(AWTEvent anEvent) {
queue.add(anEvent);
}
@Override
public void run() {
while (true) {
AWTEvent nextEvent = queue.poll();
if (nextEvent != null) {
processEvent(nextEvent);
}
}
}
void processEvent(AWTEvent theEvent) {
// calls e.g.
// ActionListener.actionPerformed,
// JComponent.paintComponent,
// Runnable.run,
// etc...
}
}
The dispatch thread is hidden from us through abstraction: we generally only write listener callbacks.
- Clicking a button posts an event (in native code): when the event is processed,
actionPerformed
is called on the EDT.
- Calling
repaint
posts an event: when the event is processed, paintComponent
is called on the EDT.
- Calling
invokeLater
posts an event: when the event is processed, run
is called on the EDT.
- Everything in Swing begins with an event.
Event tasks are processed in sequence, in the order they are posted.
The next event can only be processed when the current event task returns. This is why we cannot have an infinite loop on the EDT. actionPerformed
(or run
, as in your edit) never returns, so the calls to repaint
post paint events but they are never processed and the program appears to freeze.
This is what it means to "block" the EDT.
There are basically two ways to do animation in a Swing program:
Use a Thread
(or a SwingWorker
).
The benefit to using a thread is that the processing is done off the EDT, so if there is intensive processing, the GUI can still update concurrently.
Use a javax.swing.Timer
.
The benefit to using a timer is that the processing is done on the EDT, so there is no worry about synchronization, and it is safe to change the state of the GUI components.
Generally speaking, we should only use a thread in a Swing program if there's a reason to not use a timer.
To the user, there is no discernible difference between them.
Your call to revalidate
indicates to me that you are modifying the state of the components in the loop (adding, removing, changing locations, etc.). This is not necessarily safe to do off the EDT. If you are modifying the state of the components, it is a compelling reason to use a timer, not a thread. Using a thread without proper synchronization can lead to subtle bugs that are difficult to diagnose. See Memory Consistency Errors.
In some cases, operations on a component are done under a tree lock (Swing makes sure they are thread-safe on their own), but in some cases they are not.
We can turn a loop of the following form:
while ( condition() ) {
body();
Thread.sleep( time );
}
in to a Timer
of the following form:
new Timer(( time ), new ActionListener() {
@Override
public void actionPerformed(ActionEvent evt) {
if ( condition() ) {
body();
} else {
Timer self = (Timer) evt.getSource();
self.stop();
}
}
}).start();
Here is a simple example demonstrating animation both with a thread and a timer. The green bar moves cyclically across the black panel.
import javax.swing.*;
import java.awt.*;
import java.awt.event.*;
class SwingAnimation implements Runnable{
public static void main(String[] args) {
SwingUtilities.invokeLater(new SwingAnimation());
}
JToggleButton play;
AnimationPanel animation;
@Override
public void run() {
JFrame frame = new JFrame("Simple Animation");
JPanel content = new JPanel(new BorderLayout());
play = new JToggleButton("Play");
content.add(play, BorderLayout.NORTH);
animation = new AnimationPanel(500, 50);
content.add(animation, BorderLayout.CENTER);
// 'true' to use a Thread
// 'false' to use a Timer
if (false) {
play.addActionListener(new ThreadAnimator());
} else {
play.addActionListener(new TimerAnimator());
}
frame.setContentPane(content);
frame.pack();
frame.setLocationRelativeTo(null);
frame.setDefaultCloseOperation(JFrame.EXIT_ON_CLOSE);
frame.setVisible(true);
}
abstract class Animator implements ActionListener {
final int period = ( 1000 / 60 );
@Override
public void actionPerformed(ActionEvent ae) {
if (play.isSelected()) {
start();
} else {
stop();
}
}
abstract void start();
abstract void stop();
void animate() {
int workingPos = animation.barPosition;
++workingPos;
if (workingPos >= animation.getWidth()) {
workingPos = 0;
}
animation.barPosition = workingPos;
animation.repaint();
}
}
class ThreadAnimator extends Animator {
volatile boolean isRunning;
Runnable loop = new Runnable() {
@Override
public void run() {
try {
while (isRunning) {
animate();
Thread.sleep(period);
}
} catch (InterruptedException e) {
throw new AssertionError(e);
}
}
};
@Override
void start() {
isRunning = true;
new Thread(loop).start();
}
@Override
void stop() {
isRunning = false;
}
}
class TimerAnimator extends Animator {
Timer timer = new Timer(period, new ActionListener() {
@Override
public void actionPerformed(ActionEvent ae) {
animate();
}
});
@Override
void start() {
timer.start();
}
@Override
void stop() {
timer.stop();
}
}
static class AnimationPanel extends JPanel {
final int barWidth = 10;
volatile int barPosition;
AnimationPanel(int width, int height) {
setPreferredSize(new Dimension(width, height));
setBackground(Color.BLACK);
barPosition = ( width / 2 ) - ( barWidth / 2 );
}
@Override
protected void paintComponent(Graphics g) {
super.paintComponent(g);
int width = getWidth();
int height = getHeight();
int currentPos = barPosition;
g.setColor(Color.GREEN);
g.fillRect(currentPos, 0, barWidth, height);
if ( (currentPos + barWidth) >= width ) {
g.fillRect(currentPos - width, 0, barWidth, height);
}
}
}
}
while(true)
for test – Sounderrevalidate()
andpaint()
from? What thread is the main game loop running on? All AWT stuff needs to happen on the EDT, and the rest of your work needs to happen off it. Sounds like an AWT threading issue. Reviewing an AWT/Swing tutorial should help; they all cover the relevant threading stuff. – Warfield