Generally speaking, the way to do this is to have each of the other threads periodically check a flag. Often background threads loop, waiting for work - they just have to check the flag each time they go round a loop. If they're using Object.wait()
or something similar to be told that there's more work, the same notification should be used to indicate that the thread should stop too. (Don't just spin until you're stopped - that will suck CPU. Don't just use sleep - that will delay termination.)
That allows all threads to terminate cleanly, releasing resources appropriately. Other options such as interrupt()
and the deprecated destroy()
method are much harder to control properly, IMO. (Interrupting a thread is better than hard-aborting it, but it has its own set of problems - such as the interruption is only processed at certain points anyway.)
EDIT: In code, it would look something like:
// Client code
for (Task task : tasks) {
task.stop();
}
// Threading code
public abstract class Task implements Runnable {
private volatile boolean stopped = false;
public void stop() {
stopped = true;
}
protected boolean shouldStop() {
return stopped;
}
public abstract void run();
}
Your tasks would then subclass Task
. You would need to make it slightly more complicated if you wanted the stop()
method to also notify a monitor, but that's the basic idea.
Sample task:
public class SomeTask extends Task {
public void run() {
while (!shouldStop()) {
// Do work
}
}
}
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…