JSR 166 Snapshot Introduction.

by Doug Lea

To join a mailing list discussing this JSR, go to: http://altair.cs.oswego.edu/mailman/listinfo/concurrency-interest .

Disclaimer - This prototype is experimental code developed as part of JSR166 and made available to the developer community for use as-is. It is not a supported product. Use it at your own risk. The specification, language and implementation are subject to change as a result of your feedback. Because these features have not yet been approved for addition to the Java language, there is no schedule for their inclusion in a product.

Package java.util.concurrent contains utility classes that are commonly useful in concurrent programming. Like package java.util, it includes a few small standardized extensible frameworks, as well as some classes that provide useful functionality and are otherwise tedious or difficult to implement. In this JSR, we have been conservative in selecting only those APIs and implementations that are useful enough to encourage nearly all concurrent programmers to use routinely. JSR 166 also includes a few changes and additions in packages outside of java.util.concurrent: java.lang, to address timing and uncaught exceptions, and java.util to better integrate queues, and to make Timers conform to new frameworks. The API covers:

The main rationale for JSR 166 is that threading primitives, such as synchronized blocks, Object.wait and Object.notify, are insufficient for many programming tasks. Currently, developers can use only the concurrency control constructs provided in the Java language itself. These are too low level for some applications, and are incomplete for others. As a result, application programmers are often forced to implement their own concurrency facilities, resulting in enormous duplication of effort creating facilities that are notoriously hard to get right and even harder to optimize. Offering a standard set of concurrency utilities will ease the task of writing a wide variety of multithreaded applications and generally improve the quality of the applications that use them.

Here are brief descriptions and rationales of the main components. For details see the javadocs at http://gee.cs.oswego.edu/dl/concurrent/index.html


A basic (nonblocking) Queue interface that is compatatible with java.util.Collections will be introduced into java.util. Also, although it is at the borders of being in scope of JSR-166, java.util.LinkedList will be adapted to support Queue, and a new non-thread-safe java.util.HeapPriorityQueue will be added.

Four implementations in java.util.concurrent support the extended BlockingQueue interface, that defines blocking versions of put and take: LinkedBlockingQueue, ArrayBlockingQueue, SynchronousQueue, and PriorityBlockingQueue. Additionally, java.util.concurrent.LinkedQueue supplies an efficient thread-safe non-blocking queue.

Since the target release is JDK1.5, and generics are slated to be in 1.5, Queues should be parametrized on element type. (Also some others below.) We are ignoring this for now.


Executors provide a simple standardized interface for defining custom thread-like subsystems, including thread pools, asynch-IO, and lightweight task frameworks. Executors also standardize ways of calling threads that compute functions returning results, via Futures. This is supported in part by defining java.lang.Callable, the argument/result analog of Runnable.

While the Executor framework is intended to be extensible (so includes for example, AbstractExecutor that simplifies construction of new implementations), the most commonly used Executor will be ThreadExecutor, which can be configured to act as all sorts of thread pools, background threads, etc. The class is designed to be general enough to suffice for the vast majority of usages, even sophisticated ones, yet also includes methods and functionality that simplify routine usage.

A few methods will also be added to the java.util.Timer to support Futures, and address other requests for enhancement.


The Lock interface supports locking disciplines that differ in semantics (reentrant, semaphore-based, etc), and that can be used in non-block-structured contexts including hand-over-hand and lock reordering algorithms. This flexibility comes at the price of more awkward syntax. Implementations include Semaphore, ReentrantMutex FIFOSemaphore, and CountDownLatch.

The Locks class additionally supports trylock-designs using builtin locks without needing to use Lock classes. This requires adding new capabilities to builtin locks inside JVMs.

A ReadWriteLock interface similarly defines locks that may be shared among readers but are exclusive to writers. For this release, only a single implementation, ReentrantReadWriteLock, is planned, since it covers all standard usage contexts. But programmers may create their own implementations to cover nonstandard requirements.


A Condition class provides the kinds of condition variables associated with monitors in other cocurrent languages, as well as pthreads condvars. Their support reduces the need for tricky and/or inefficient solutions to many classic concurrent problems. Conditions also address the annoying problem that Object.wait(msecs) does not return an indication of whether the wait timed out. This leads to error-prone code. Since this method is in class Object, the problem is basically unfixable.

To avoid compatibility problems, the names of Condition methods need to be different than Object versions. The downside of this is that people can make the mistake of calling cond.notify instead of cond.signal. However, they will get IllegalMonitorState exceptions if they do, so they can detect the error if they ever run the code.

The implementation requires VM magic to atomically suspend and release lock. But it is unlikely to be very challenging for JVM providers, since most layer Java monitors on top of posix condvars or similar low-level functionality anyway.

Atomic variables

Classes AtomicInteger, AtomicLong, AtomicDouble, AtomicFloat, and AtomicReference provide simple scalar variables supporting compareAndSwap (CAS) and related atomic operations. These are desparately needed by those performing low-level concurrent system programming, but much less commonly useful in higher-level frameworks.


Java has always supported sub-millisecond versions of several native time-out-based methods (such as Object.wait), but not methods to actually perform timing in finer-grained units. We address this by introducing java.lang.Clock, which provides multiple granularities for both accessing time and performing time-out based operations.


Barriers (multiway synchronization points) are very common in some styles of parallel programming, yet tricky to get right. The two most useful flavors (CyclicBarriers and Exchangers) don't have much of an interface in common, and only have one standard implementation each, so these are simply defined as public classes rather than interfaces and implementations.

Concurrent Collections

There are no new interfaces, but JSR 166 will supply a few Collection implementations designed for use in multithreaded contexts: ConcurrentHashTable, CopyOnWriteArrayList, and CopyOnWriteArraySet.

Uncaught Exception Handlers

The java.lang.Thread class will be modified to allow per-thread installation of handlers for uncaught exceptions. Ths optionally disassociates these handlers from ThreadGroups, which has proven to be too inflexible in many multithreaded programs. (Note that the combination of features in JSR 166 make ThreadGroups even less likely to be used in most programs. Perhaps they will eventually be deprecated.)

Additionally, Threads and ThreadLocals will now support a means to clear and remove ThreadLocals, which is needed in some thread-pool and worker-thread designs.

Doug Lea