A manageable version of a java.util.concurrent.ExecutorService.

A ManagedExecutorService extends the Java™ SE ExecutorService to provide methods for submitting tasks for execution in a Jakarta™ EE environment. Implementations of the ManagedExecutorService are provided by a Jakarta EE Product Provider. Application Component Providers use the Java Naming and Directory Interface™ (JNDI) to look-up instances of one or more ManagedExecutorService objects using resource environment references. ManagedExecutorService instances can also be injected into application components through the use of the Resource annotation.

The Jakarta Concurrency specification describes several behaviors that a ManagedExecutorService can implement. The Application Component Provider and Deployer identify these requirements and map the resource environment reference appropriately.

The most common uses for a ManagedExecutorService is to run short-duration asynchronous tasks such as for processing of asynchronous methods in Jakarta Enterprise Beans or for processing async tasks for Servlets that supports asynchronous processing.

Tasks are run in managed threads provided by the Jakarta EE Product Provider and are run within the application component context that submitted the task. All tasks run without an explicit transaction (they do not enlist in the application component's transaction). If a transaction is required, use a jakarta.transaction.UserTransaction instance. A UserTransaction instance is available in JNDI using the name: "java:comp/UserTransaction" or by requesting an injection of a jakarta.transaction.UserTransaction object using the Resource annotation.

Example:

 public run() {
   // Begin of task
   InitialContext ctx = new InitialContext();
   UserTransaction ut = (UserTransaction) ctx.lookup("java:comp/UserTransaction");
   ut.begin();

   // Perform transactional business logic

   ut.commit();
 }
 

Tasks can optionally provide an ManagedTaskListener to receive notifications of lifecycle events, through the use of ManagedTask interface.

Example:

 public class MyRunnable implements Runnable, ManagedTask {
   ...
   public void run() {
     ...
   }

   public ManagedTaskListener getManagedTaskListener() {
     return myManagedTaskListener;
   }
   ...
 }

 MyRunnable task = ...;
 ManagedExecutorService executor = ...;

 executor.submit(task); // lifecycle events will be notified to myManagedTaskListener
 

Asynchronous tasks are typically submitted to the ManagedExecutorService using one of the submit methods, each of which return a java.util.concurrent.Future instance. The Future represents the result of the task and can also be used to check if the task is complete or wait for its completion.

If the task is canceled, the result for the task is a java.util.concurrent.CancellationException exception. If the task is unable to run due to a reason other than cancellation, the result is a AbortedException exception.

Example:

 /**
  * Retrieve all accounts from several account databases in parallel.
  * Resource Mappings:
  *  type:      jakarta.enterprise.concurrent.ManagedExecutorService
  *  jndi-name: concurrent/ThreadPool
  */
 public List<Account> getAccounts(long accountId) {
   try {
       javax.naming.InitialContext ctx = new InitialContext();
       ManagedExecutorService mes = (ManagedExecutorService)
           ctx.lookup("java:comp/env/concurrent/ThreadPool");

       // Create a set of tasks to perform the account retrieval.
       ArrayList<Callable<Account>> retrieverTasks = new ArrayList<Callable<Account>>();
       retrieverTasks.add(new EISAccountRetriever());
       retrieverTasks.add(new RDBAccountRetriever());

       // Submit the tasks to the thread pool and wait for them
       // to complete (successfully or otherwise).
       List<Future<Account>> taskResults= mes.invokeAll(retrieverTasks);

       // Retrieve the results from the resulting Future list.
       ArrayList<Account> results = new ArrayList<Account>();
       for(Future<Account> taskResult : taskResults) {
           try {
               results.add(taskResult.get());
           } catch (ExecutionException e) {
               Throwable cause = e.getCause();
               // Handle the AccountRetrieverError.
           }
       }

       return results;

   } catch (NamingException e) {
       // Throw exception for fatal error.
   } catch (InterruptedException e) {
       // Throw exception for shutdown or other interrupt condition.
   }
 }


 public class EISAccountRetriever implements Callable<Account> {
     public Account call() {
         // Connect to our eis system and retrieve the info for the account.
         //...
         return null;
   }
 }

 public class RDBAccountRetriever implements Callable<Account> {
     public Account call() {
         // Connect to our database and retrieve the info for the account.
         //...
   }
 }

 public class Account {
     // Some account data...
 }
 

ManagedExecutorService provides various methods which correspond to the static methods of java.util.concurrent.CompletableFuture and its constructor/newIncompleteFuture method, enabling you to create completion stages that are backed by the ManagedExecutorService as the default asynchronous execution facility, both for those stages as well as all dependent stages that are created from those, and so on. This allows you to create pipelines of completion stage actions that run with consistent and predictable thread context, regardless of which thread each dependent action ends up running on.

Example:

 ManagedExectorService executor = InitialContext.doLookup("java:comp/DefaultManagedExecutorService");
 ...
 CompletableFuture<Integer> future = executor
    .supplyAsync(supplier)
    .thenApply(function1)
    .thenApplyAsync(function2)
    ...
 
 

Context propagation to completion stages that are backed by a ManagedExecutorService must be done in a consistent and predictable manner, which is defined as follows,

• If the supplied action is already contextual, (for example, contextService.createContextualProxy(action, Runnable.class)), then the action runs with the already-captured context.
• Otherwise, each type of thread context is either propagated from the thread that creates the completion stage or the context is marked to be cleared, according to the configuration of the ManagedExecutorService that is the default asynchronous execution facility for the new stage and its parent stage. In the case that a ManagedExecutorService is supplied as the executor argument to a *Async method, the supplied ManagedExecutorService is used to run the action, but not to determine the thread context propagation and clearing.

Each type of thread context is applied (either as cleared or previously captured) to the thread that runs the action. The applied thread context is removed after the action completes, whether successfully or exceptionally, restoring the thread's prior context.

When dependent stages are created from the completion stage, and likewise from any dependent stages created from those, and so on, thread context is captured or cleared in the same manner. This guarantees that the action performed by each stage always runs under the thread context of the code that creates the completion stage, unless the user explicitly overrides this by supplying a pre-contextualized action.

Completion stages that are backed by a ManagedExecutorService must raise java.lang.IllegalArgumentException if supplied with an action that implements ManagedTask.