Best practices or principles for sharing objects between threads in Java

Question

I tried to search but couldn't find exact answer I was looking for hence putting up a new question.

If you wish to share any mutable object(s) between multiple threads, are there any best practices/principles/guidelines to do it ?

Or will it simply vary case by case ?

Answer 1

Sharing mutable objects between threads is risky.

The safest way is to make the objects immutable, you can then share them freely.

If they must be mutable then each of the objects each needs to ensure their own thread safety using the usual methods to do so. ( synchronized , AtomicX classes, etc).

The ways to protect the individual objects will vary a lot though depending on how you are using them and what you are using them for.

Answer 2

In java, you should synchronize any method that changes/reads the state of shared object, it is the easiest way.

other strategies are:

• make use of thread safe classes (ConcurrentHashMap) for example
• use of locks
• use of volatile keyword, to avoid stale objects (sometimes could be used as lightweight synchronizer)

they key is sync your updates/reads to guarantee consistent state, the way you do it, could vary a lot.

Answer 3

The problems with sharing objects between threads are caused by having the two threads access the same data structure at the same time, with one mutating the structure while the other depends on the structure to be complete, correct or stable. Which of these cause the problem is important and should be considered when choosing the strategy.

These are the strategies I use.

1. Use immutable objects as much as possible.

This removes the issue of changing the data structure altogether. There are however a lot of useful patterns that can not be written using this approach. Also unless you are using a language/api which promotes immutability it can be inefficient. Adding a entry to a Scala list is much faster than making a copy of a Java list and adding a entry to the copy.

2. Use the synchronize keyword.

This ensures that only one thread at a time is allowed to change the object. It is important to choose which object to synchronize on. Changing a part of a structure might put the hole structure in an illegal state until another change is made. Also synchronize removes many of the benefits of going multithreaded in the first place.

3. The Actor model.

The actor model organizes the world in actors sending immutable messages to each other. Each actor only has one thread at once. The actor can contain the mutability. There are platforms, like Akka, which provide the fundamentals for this approach.

4. Use the atomic classes. (java.util.concurrent.atomic)

These gems have methods like incrementAndGet. They can be used to achieve many of the effects of synchronized without the overhead.

5. Use concurrent data structures.

The Java api contains concurrent data structures created for this purpose.

5. Risk doing stuff twice.

When designing a cache it is often a good idea to risk doing the work twice instead of using synchronize. Say you have a cache of compiled expressions from a dsl. If an expression is compiled twice that is ok as long as it eventually ends up in the cache. By allowing doing some extra work during initialization you may not need to use the synchronize keyword during cache access.

Answer 4

There is example. StringBuilder is not thread safe, so without synchronized (builder) blocks - result will be broken. Try and see.

Some objects are thread safe (for example StringBuffer), so no need to use synchronized blocks with them.

    public static void main(String[] args) throws InterruptedException {
        StringBuilder builder = new StringBuilder("");

        Thread one = new Thread() {
            public void run() {

                for (int i = 0; i < 1000; i++) {
                    //synchronized (builder) {
                        builder.append("thread one\n");
                    //}
                }
            }
        };
        Thread two = new Thread() {
            public void run() {

                for (int i = 0; i < 1000; i++) {
                    //synchronized (builder) {
                        builder.append("thread two\n");
                    //}
                }
            }
        };
        one.start();
        two.start();
        one.join();
        two.join();

        System.out.println(builder);
    }

Answer 5

Although there are some good answers already posted, but here is what I found while reading Java Concurrency in Practice Chapter 3 - Sharing Objects.

Quote from the book.

The publication requirements for an object depend on its mutability:

• Mutable objects can be published through any mechanism;
• Effectively immutable objects (whose state will not be modified after publication) must be safely published;
• Mutable objects must be safely published, and must be either threadsafe or guarded by a lock.

Book states ways to safely publish mutable objects:

To publish an object safely, both the reference to the object and the object's state must be made visible to other threads at the same time. A properly constructed object can be safely published by:

• Initializing an object reference from a static initializer;
  
• Storing a reference to it into a volatile field or AtomicReference;
  
• Storing a reference to it into a final field of a properly constructed object; or
  
• Storing a reference to it into a field that is properly guarded by a lock.
  

The last point refers to using various mechanisms like using concurrent data structures and/or using synchronize keyword.