In this chapter, we learned how STM works and how to apply it in concurrent programs. We saw the advantages of using STM's transactional references and
atomic blocks over the
synchronized statements, and investigated their interaction with side effects. We studied the semantics of exception handling inside transactions and learned how to retry and conditionally re-execute transactions. Finally, we learned about transactional collections, which allow us to encode shared program data more efficiently.
These features together enable a concurrent programming model in which the programmer can focus on expressing the meaning of the program, without having to worry about handling lock objects, or avoiding deadlocks and race conditions. This is especially important when it comes to modularity. It is hard or near impossible to reason about deadlocks or race conditions in the presence of separate software components. STM exists to liberate the programmer from such concerns, and is essential...