Everyone here seems to think that implementing Runnable is the way to go and I don't really disagree with them but there is also a case for extending Thread in my opinion, in fact you have sort of demonstrated it in your code.
If you implement Runnable then the class that implements Runnable has no control over the thread name, it is the calling code that can set the thread name, like so:
new Thread(myRunnable,"WhateverNameiFeelLike");
but if you extend Thread then you get to manage this within the class itself (just like in your example you name the thread 'ThreadB'). In this case you:
A) might give it a more useful name for debugging purposes
B) are forcing that that name be used for all instances of that class (unless you ignore the fact that it is a thread and do the above with it as if it is a Runnable but we are talking about convention here in any case so can ignore that possibility I feel).
You might even for example take a stack trace of its creation and use that as the thread name. This might seem odd but depending on how your code is structured it can be very useful for debugging purposes.
This might seem like a small thing but where you have a very complex application with a lot of threads and all of a sudden things 'have stopped' (either for reasons of deadlock or possibly because of a flaw in a network protocol which would be less obvious - or other endless reasons) then getting a stack dump from Java where all the threads are called 'Thread-1','Thread-2','Thread-3' is not always very useful (it depends on how your threads are structured and whether you can usefully tell which is which just by their stack trace - not always possible if you are using groups of multiple threads all running the same code).
Having said that you could of course also do the above in a generic way by creating an extension of the thread class which sets its name to a stack trace of its creation call and then use that with your Runnable implementations instead of the standard java Thread class (see below) but in addition to the stack trace there might be more context specific information that would be useful in the thread name for debugging (a reference to one of many queues or sockets it could processing for example in which case you might prefer to extend Thread specifically for that case so that you can have the compiler force you (or others using your libraries) to pass in certain info (e.g. the queue/socket in question) for use in the name).
Here's an example of the generic thread with the calling stack trace as its name:
public class DebuggableThread extends Thread {
private static String getStackTrace(String name) {
Throwable t= new Throwable("DebuggableThread-"+name);
ByteArrayOutputStream os = new ByteArrayOutputStream();
PrintStream ps = new PrintStream(os);
t.printStackTrace(ps);
return os.toString();
}
public DebuggableThread(String name) {
super(getStackTrace(name));
}
public static void main(String[] args) throws Exception {
System.out.println(new Thread());
System.out.println(new DebuggableThread("MainTest"));
}
}
and here's a sample of the output comparing the two names:
Thread[Thread-1,5,main]
Thread[java.lang.Throwable: DebuggableThread-MainTest
at DebuggableThread.getStackTrace(DebuggableThread.java:6)
at DebuggableThread.<init>(DebuggableThread.java:14)
at DebuggableThread.main(DebuggableThread.java:19)
,5,main]
interrupt()
. Again, it's an idea, it might be useful in the right case, however I do not recommend it. – Hungnamjava.nio.channels.InterruptibleChannel
(s) You can have a look at the impl. of Thread.interrupt(). Using a helping hand from theinterrupt()
caller is an idea and a useful approach, again not recommended for rookies. – Hungnamjava.nio.channels.spi.AbstractSelector
and overridewakeup()
but it's a way too dirty way to asses the case :) – HungnamRunnable
in OP) needs to execute logic in the caller. – Ineptitude