Retrying transactions

In sequential computing, a single thread is responsible for executing the program. If a specific value is not available, the single thread is responsible for producing it. In concurrent programming, the situation is different. When a value is not available, some other thread, called a producer, might eventually produce the value. The thread consuming the value, called a consumer, can either block the execution until the value becomes available, or temporarily execute some other work before checking for the value again. We have seen various mechanisms for achieving this relationship, ranging from monitors and the synchronized statement from Chapter 2, Concurrency on the JVM and the Java Memory Model, concurrent queues from Chapter 3, Traditional Building Blocks of Concurrency; futures and promises in Chapter 4, Asynchronous Programming with Futures and Promises; to event-streams in Chapter 6, Concurrent Programming with Reactive Extensions.

Syntactically, the atomic statement...