When used correctly, transactional memory is a powerful tool for building concurrent applications that modify shared data. Nevertheless, no technology is a silver bullet, and neither is STM. In this section, we will study how to compose transactions in larger programs and learn how transactional memory interacts with other features of Scala. We investigate some of the caveats of STM, and go beyond transactional references and the
atomic blocks to show how to use STM more effectively.
Previously, we learned that an STM may roll back and retry a transaction. An attentive reader might notice that retrying a transaction means re-executing its side effects. Here, the side effects are arbitrary reads and writes to regular object fields and variables.