Deterministic Core, Non-Deterministic Shell: Expanding on Functional Core/Imperative Shell

Revisiting Functional Core/Imperative Shell

The concept of "Functional Core, Imperative Shell," coined by Gary Bernhardt fourteen years ago, divides application code into two main parts. The Functional Core handles business logic without IO or destructive state updates, while the Imperative Shell manages state, external dependencies, and interactions with the outside world.

Characteristics of Core and Shell

The Functional Core is designed to be purely functional, meaning the same inputs consistently produce the same outputs. This isolation makes it highly amenable to testing, as there are no external dependencies to mock or stub. The Imperative Shell, conversely, has less complex logic but is responsible for coordinating with external systems like databases or APIs.

Determinism Beyond Pure Functions

The article argues that the key property making pure functions testable is determinism, not necessarily pure functional programming itself. Determinism ensures that given a specific sequence of inputs, the system's behavior is repeatable and predictable. This allows for reliable testing of the core's logic.

State Machines as Deterministic Cores

The author demonstrates that state machines can also provide deterministic behavior, similar to pure functions. An example of an `AddMachine` shows that even with imperative state updates, a consistent sequence of calls to its `transition` function will always result in the same final state. This suggests that state machines can serve as deterministic cores, offering the same testability benefits where pure functional approaches might be impractical due to language or performance constraints.