Develop Reference

Semaphore with queue

public class SemaphoreWithQueues implements Semaphore {
private List<Object> queue;
private AtomicInteger current = new AtomicInteger(0);
private int permits;
public SemaphoreWithQueues(int permits) {
this.permits = permits;
this.queue = Collections.synchronizedList(new LinkedList<>());
}
#Override
public void enter() throws InterruptedException {
if (current.get() < permits) {
current.incrementAndGet();
} else {
Object block = new Object();
synchronized (block) {
queue.add(block);
block.wait();
current.incrementAndGet();
}
}
}
#Override
public void leave() {
if(queue.size() != 0) {
Object block = queue.get(0);
queue.remove(0);
synchronized (block) {
block.notify(); //Unblock quenue
}
}
current.decrementAndGet();
//current lessen and current thread have time come in block if(...)
// in enter() faster then another thread increased current
}
}
> The program usually output:
>
> 1 1 2 2 1 1 2 2 1 2
**Where run() of both threads is almost the same, such as:**
public void run(){
for (int i = 0; i <5; i++) {
try {
semaphore.enter();
} catch (InterruptedException e) {
System.err.println(e);
}
System.out.println(2);
semaphore.leave();
}
}
There are 2 threads using this semaphore. When 1 thread increases the queue, the second is waiting, the problem is that if we extracted the object from quene and unblocked it, then the thread that finished leave() start enter() faster and again increments the counter, while the awaked thread also increments the counter, current = 2, and the list is empty.
SORRY FOR BAD ENGLISH

There are many problems in the code.
Synchronization: Synchronization should be done for a shareable
resource. Why it is done for a local object which has scope only for
that method.
Object block = new Object();
synchronized (block) {
Both current and queue are independent properties, they should be
synchronized together.
Now let's come to point If you really want to create a semaphore using Queue. You do not need all this logic. You can use existing Java class e.g. BlockingQueue. Here is the implementation
class SemaphoreWithQueues implements Semaphore{
private BlockingQueue<Integer> queue;
public SemaphoreWithQueues(int permits) {
if(queue == null){
queue = new ArrayBlockingQueue<>(permits);
}
}
public void enter() {
queue.offer(1);
System.out.println(Thread.currentThread().getName() + " got a permit.");
}
public void leave() throws InterruptedException {
queue.take();
System.out.println(Thread.currentThread().getName() + " left the permit.");
}
}
And Task to use the semaphore
class Task implements Runnable {
private SemaphoreWithQueues semaphore;
public Task(SemaphoreWithQueues semaphore){
this.semaphore = semaphore;
}
public void run(){
for (int i = 0; i <5; i++) {
semaphore.enter();
try {
semaphore.leave();
} catch (InterruptedException e) {
e.printStackTrace();
}
}
}
}
public class Main {
public static void main(String[] args) {
SemaphoreWithQueues semaphoreWithQueues = new SemaphoreWithQueues(5);
Thread th1 = new Thread(new Task(semaphoreWithQueues));
Thread th2 = new Thread(new Task(semaphoreWithQueues));
Thread th3 = new Thread(new Task(semaphoreWithQueues));
th1.start();
th2.start();
th3.start();
}
}
But personally I do not like using Queue to create Semaphores, as it wastes memory unnecessary by creating elements in queue. Despite of this you can create a semaphore using single shareable object with permits using wait and notify mechanism. You can try with this approach. If you would like.

printing alternative output from 2 threads using semaphores

I am learning about the use of semaphores and multi threading in general but am kind of stuck. I have two threads printing G and H respectively and my objective is to alternate the outputs of each thread so that the output string is like this;
G
H
G
H
G
H
Each of the two classes has a layout similar to the one below
public class ClassA extends Thread implements Runnable{
Semaphore semaphore = null;
public ClassA(Semaphore semaphore){
this.semaphore = semaphore;
}
public void run() {
while(true)
{
try{
semaphore.acquire();
for(int i=0; i<1000; i++){
System.out.println("F");
}
Thread.currentThread();
Thread.sleep(100);
}catch(Exception e)
{
System.out.println(e.toString());
}
semaphore.release();
}
}
}
below is my main class
public static void main(String[] args) throws InterruptedException {
Semaphore semaphore = new Semaphore(1);
ClassA clasA = new ClassA(semaphore);
Thread t1 = new Thread(clasA);
ClassB clasB = new ClassB(semaphore);
Thread t2 = new Thread(clasB);
t1.start();
t2.join();
t2.start();
The output I am getting is way too different from my expected result. can anyone help me please? did I misuse the semaphore? any help?

Semaphores can't help you solve such a task.
As far as I know, JVM doesn't promise any order in thread execution. It means that if you run several threads, one thread can execute several times in a row and have more processor time than any other. So, if you want your threads to execute in a particular order you can, for the simplest example, make a static boolean variable which will play a role of a switcher for your threads. Using wait() and notify() methods will be a better way, and Interface Condition will be the best way I suppose.
import java.io.IOException;
public class Solution {
public static boolean order;
public static void main(String[] args) throws IOException, InterruptedException {
Thread t1 = new ThreadPrint("G", true);
Thread t2 = new ThreadPrint("O", false);
t1.start();
t2.start();
t2.join();
System.out.println("Finish");
}
}
class ThreadPrint extends Thread {
private String line;
private boolean order;
public ThreadPrint(String line, boolean order) {
this.line = line;
this.order = order;
}
#Override
public void run() {
int z = 0;
while (true) {
try {
for (int i = 0; i < 10; i++) {
if (order == Solution.order) {
System.out.print(line + " ");
Solution.order = !order;
}
}
sleep(100);
} catch (Exception e) {
System.out.println(e.toString());
}
}
}
}
BTW there can be another problem cause System.out is usually an Operation System buffer and your OS can output your messages in an order on its own.
P.S. You shouldn't inherit Thread and implement Runnable at the same time
public class ClassA extends Thread implements Runnable{
because Thread class already implements Runnable. You can choose only one way which will be better for your purposes.
You should start a thread then join to it not vice versa.
t1.start();
t2.join();
t2.start();

As others have pointed out, locks themselves do not enforce any order and on top of that, you cannot be certain when a thread starts (calling Thread.start() will start the thread at some point in the future, but this might take a while).
You can, however, use locks (like a Semaphore) to enforce an order. In this case, you can use two Semaphores to switch threads on and off (alternate). The two threads (or Runnables) do need to be aware of each other in advance - a more dynamic approach where threads can "join in" on the party would be more complex.
Below a runnable example class with repeatable results (always a good thing to have when testing multi-threading). I will leave it up to you to figure out why and how it works.
import java.util.concurrent.*;
public class AlternateSem implements Runnable {
static final CountDownLatch DONE_LATCH = new CountDownLatch(2);
static final int TIMEOUT_MS = 1000;
static final int MAX_LOOPS = 10;
public static void main(String[] args) {
ExecutorService executor = Executors.newCachedThreadPool();
try {
AlternateSem as1 = new AlternateSem(false);
AlternateSem as2 = new AlternateSem(true);
as1.setAlternate(as2);
as2.setAlternate(as1);
executor.execute(as1);
executor.execute(as2);
if (DONE_LATCH.await(TIMEOUT_MS, TimeUnit.MILLISECONDS)) {
System.out.println();
System.out.println("Done");
} else {
System.out.println("Timeout");
}
} catch (Exception e) {
e.printStackTrace();
} finally {
executor.shutdownNow();
}
}
final Semaphore sem = new Semaphore(0);
final boolean odd;
AlternateSem other;
public AlternateSem(boolean odd) {
this.odd = odd;
}
void setAlternate(AlternateSem other) { this.other = other; }
void release() { sem.release(); }
void acquire() throws Exception { sem.acquire(); }
#Override
public void run() {
if (odd) {
other.release();
}
int i = 0;
try {
while (i < MAX_LOOPS) {
i++;
other.acquire();
System.out.print(odd ? "G " : "H ");
release();
}
} catch (Exception e) {
e.printStackTrace();
}
DONE_LATCH.countDown();
}
}

Java: How to put thread back to sleep INDEFINITELY after interrupt?

I am writing Java software, that has a single thread, which listens to external buttons being pressed. If the button is pressed, the thread informs other threads, but otherwise it just sleeps.
My model is to use interrupt-driven design. Ideally I would like to make
a thread sleep as long as no button is pressed. When the button is pressed I would like the thread to do some work and go back to sleep.
Could anyone confirm / correct the following implementation?
// This is a code that interrupt-driven thread will execute
public void run() {
while (true) {
try {
Thread.sleep(1000); // Sleeps only for 1s. How to sleep indefinitely?
} catch (InterruptedException exception) {
process(exception); // Doing some work
// then going back to sleep using the while loop
}
}
}
Also, after each button click in the terminal I get a message
I/O exception raised from stop()
What does this message mean (i.e why is it printed if I catch the exception)? Can I avoid the terminal to print it?

It is generally considered a code smell if you use exceptions to control your program flow.
The correct solution to this problem is to use a BlockingQueue of events that the event handler reads from. This is commonly called a producer/consumer.
public class TwoThreads {
public static void main(String args[]) throws InterruptedException {
System.out.println("TwoThreads:Test");
new TwoThreads().test();
}
// The end of the list.
private static final Integer End = -1;
static class Producer implements Runnable {
final BlockingQueue<Integer> queue;
public Producer(BlockingQueue<Integer> queue) {
this.queue = queue;
}
#Override
public void run() {
try {
for (int i = 0; i < 1000; i++) {
queue.add(i);
Thread.sleep(1);
}
// Finish the queue.
queue.add(End);
} catch (InterruptedException ex) {
// Just exit.
}
}
}
static class Consumer implements Runnable {
final BlockingQueue<Integer> queue;
public Consumer(BlockingQueue<Integer> queue) {
this.queue = queue;
}
#Override
public void run() {
boolean ended = false;
while (!ended) {
try {
Integer i = queue.take();
ended = i == End;
System.out.println(i);
} catch (InterruptedException ex) {
ended = true;
}
}
}
}
public void test() throws InterruptedException {
BlockingQueue<Integer> queue = new LinkedBlockingQueue<>();
Thread pt = new Thread(new Producer(queue));
Thread ct = new Thread(new Consumer(queue));
// Start it all going.
pt.start();
ct.start();
// Wait for it to finish.
pt.join();
ct.join();
}
}
Don't let yourself be confused by how much code this is - most of it is just wrapping. The core functionality is:
At start - create a BlockingQueue and share it between the two threads.
BlockingQueue<Integer> queue = new LinkedBlockingQueue<>();
Thread pt = new Thread(new Producer(queue));
Thread ct = new Thread(new Consumer(queue));
When an event happens, post to the queue.
queue.add(i);
The event handler feeds off the queue.
while (!ended) {
try {
Integer i = queue.take();
Note that take here will block until an event is posted or an interrupt occurrs.

You can use
Thread.sleep(Long.MAX_VALUE); // more than the life of your computer
or
synchronized(this) {
wait();
}
or this wake on interrupt but doesn't throw an exception
LockSupport.park();
However a more elegant solution is likely to be to use an ExecutorService is designed to be a sleeping thread pool which wakes when you give it work to do.
ExecutorsService executor = Executors.newSingleThreadExecutor();
// when you want it to do something
executor.submit(this::process);
Note: you should consider how you want to handle exceptions. In the example in your question, an exception or error will kill the thread and it will stop running. In my example it won't kill the thread pool but the actual exception could be lost. For this reason I suggest you write it like this.
executor.submit(() -> {
try {
process();
} catch(Throwable t) {
LOGGER.warning(t);
}
});
Note: instead of just calling process and it having to figure out what you want to do you can write it like this.
doSomething(string, number, pojo);
That way you can see what data you expect the background thread to work on.
For comparison, here is the TwoThread example using the current thread as a producer and an Executor Service.
public class TwoThreadsJava5 {
public static void main(String args[]) throws InterruptedException {
System.out.println("TwoThreads:Test - Java 5.0 style");
ExecutorService executor = Executors.newSingleThreadExecutor();
for (int i = 0; i < 1000; i++) {
final int finalI = i;
executor.submit(() -> {
try {
System.out.println(finalI);
} catch (Throwable t) {
t.printStackTrace();
}
});
}
executor.shutdown();
executor.awaitTermination(1, TimeUnit.MINUTES);
}
}
And in Java 8 you could write
public class TwoThreadsJava8 {
public static void main(String args[]) throws InterruptedException {
System.out.println("TwoThreads:Test - Java 8 style");
IntStream.range(0, 1000)
.parallel()
.forEach(System.out::println);
}
}

why object.wait(value) is not accurate?

consider this code which basically has an object(WaitedObject) and two threads(SomeTask and SomeTaskWithWait) compete to call the methods (longRunningTask() and withWaitTask() respectively) of the object synchronously
package closerLookAtWait;
class WaitedObject
{
int i=0;
synchronized void longRunningTask()
{
System.out.println(i++);
for(long j=999; j>0; j--)
{}
}
synchronized void withWaitTask()
{
System.out.println("Now Waiting");
long time1 = System.currentTimeMillis();
try {
//Thread.sleep(500);
wait(50);
} catch (InterruptedException e) {
// TODO Auto-generated catch block
e.printStackTrace();
}
long time2 = System.currentTimeMillis() - time1;
System.out.println("Done Waiting for "+time2);
}
}
class SomeTask implements Runnable
{
WaitedObject wo;
SomeTask(WaitedObject wo)
{
this.wo = wo;
}
#Override
public void run() {
while(true)
wo.longRunningTask();
}
}
class SomeTaskWithWait implements Runnable{
WaitedObject wo;
SomeTaskWithWait(WaitedObject wo)
{
this.wo = wo;
}
#Override
public void run() {
while(true)
wo.withWaitTask();
}
}
public class SomeWaitingWithLong {
public static void main(String[] args) {
WaitedObject wo = new WaitedObject();
new Thread(new SomeTask(wo)).start();
new Thread(new SomeTaskWithWait(wo)).start();
}
}
sample output:
well i got output as 54,54,50,65,51,52,..,78,..84,..50,52,52.
now my question is why such inaccuracy? (even 65 is ok, but why 84?)

One of the reasons is, OS puts that thread in suspended mode for the time(ms) you provide in wait(). When the time completes it isn't guarrented that your thread will be executed at once because OS has assigned another thread with a higher priority in your process to be executed by the processor or some other higher priority process is being assigned to the processor for execution. Even if your thread was at highest priority, even then there will be some delay sometimes because of context switching & in Java's case, GC.

Simple answer: Android is not a real time OS.

Java ThreadPool reporting

I have a worker threadpool set up that executes a bit of work which I want to log in a central place.
To be more precise, I've extended the Thread class into a worker class, which checks the status of a concurrent queue. If it's empty, then it waits. As elements are added by another thread, notify() wakes the workers. Once they've completed the task, they wait for the next element in the queue.
What's the best practice to have each of the threads report their status at the end of each of their tasks?
public class PoolWorker extends Thread {
public ConcurrentLinkedQueue<Device> q;
public PoolWorker(ConcurrentLinkedQueue<Device> q, String type){
this.q = q;
this.type = type;
}
#Override
public void run(){
while (true)
{
Device d = null;
try{
synchronized(q){
while(q.isEmpty())
{
q.wait(); // wait for a notify()
}
d = q.remove();
}
// do some work
// report status of work completed
}
}

Try to do something like this
ExecutorService exec = Executors.newFixedThreadPool(10);
Runnable runn = new Runnable()
{
#Override
public void run()
{
System.out.println("");
}
};
exec.execute(runn);

As mentioned best way is to use BlockingQueue. Below is the sample code:
public class PoolWorker extends Thread {
public ArrayBlockingQueue<String> q;
public String type;
public PoolWorker(ArrayBlockingQueue<String> q, String type) {
this.q = q;
this.type = type;
}
#Override
public void run() {
while(true){
String work = null;
try {
System.out.println("PoolWorker.run:waiting .............");
work = q.take();
} catch (InterruptedException e) {
e.printStackTrace();
}
System.out.println("PoolWorker.run..work: " + work);
}
}
public static void main(String[] args) throws InterruptedException {
ArrayBlockingQueue<String> pool = new ArrayBlockingQueue<String>(100);
PoolWorker worker = new PoolWorker(pool, "Something");
worker.start();
addWork(pool, "work1");
addWork(pool, "work2");
addWork(pool, "work3");
addWork(pool, "work4");
addWork(pool, "work5");
//Just give enough time to run
Thread.sleep(5000);
}
private static void addWork(ArrayBlockingQueue<String> pool, String work) throws InterruptedException {
System.out.println("PoolWorker.addWork: " + work);
pool.put(work);
}
}
There is nice sample code available in Java documentation as well:
http://docs.oracle.com/javase/7/docs/api/java/util/concurrent/BlockingQueue.html