RecursiveTask

RecursiveTask is an abstract class in the Java ForkJoin framework (java.util.concurrent) that represents a task which returns a result. It extends ForkJoinTask and is the counterpart to RecursiveAction, which does not return a value. Subclasses implement the compute() method to perform the task’s work.

A typical use pattern involves either computing directly for small problems or splitting the problem into subtasks, forking them, and then combining the results. In practice, a RecursiveTask’s compute() often checks a termination condition or threshold; if the problem is large, it creates two or more subtasks, forks them (either explicitly or via invokeAll), and then gathers their results with join() and combines them.

Tasks are executed within a ForkJoinPool, which can be the common pool or a user-created pool. The framework employs work-stealing to balance load among threads, allowing tasks to be executed in parallel across multiple cores.

Common applications include parallel reductions (such as summing elements of an array), divide-and-conquer algorithms, and other computations that can be broken into independent subproblems whose results can be combined. As with other parallel tasks, careful attention is needed to overhead: excessive task creation or fine-grained subtasks can degrade performance. Avoid shared mutable state and blocking calls within compute() to maintain scalability.

Usage outline: create a subclass of RecursiveTask<V>, implement compute() to either perform direct computation or split into subtasks, submit the top-level task to a ForkJoinPool, and obtain the result via join() or through the pool’s invoke method.