One strength of Scala is how convenient it is for building Domain-Specific Languages (DSLs). Many Scala libraries involve some kind of DSL, and a popular application of DSLs is to offer non-default forms of control flow to the developer - we call this sort of DSL a control flow DSL. They are commonly used to control concurrency, parsing, to express distributed data transformers, or simply to provide general tail-calling in Scala programs.

The key feature of control flow DSLs, programmatically replacing the relationship between program logic and the JVM callstack, is also the reason behind why debugging such programs is a significant challenge. Common debugging approaches, such as looking at the stacktrace from a program crash, or attaching a debugger, no longer work since the corresponding artifacts do not accurately represent the program’s logical state.

In this talk, we propose a lightweight in-process implementation strategy for the Debug Adapter Protocol, which can bring modern debugging convenience to the land of control flow DSLs. We describe what we learned from prototyping this strategy in our own DSL, and open a broader discussion on evaluating the applicability and usefulness of this technique.