Best practices for mocking dependencies?

[christianheussy]

I'm really enjoying statig but am facing challenges with unit testing state transitions. My goal is to:

1. Initialize the state machine to a specific state.
2. Mock dependencies within the state machine.
3. Verify state transitions after handling events.

I'm using serde for state initialization (as suggested in #25), but this prevents me from setting mock expectations beforehand. See a reproducible example here: #33.

The issue is that neither UninitializedStateMachine<M> nor InitializedStateMachine<M> implement core::ops::DerefMut, preventing mutable access for mock configuration before initialization.

A potential solution is implementing DerefMut (#32), but I'm unsure if providing mutable access at all times is ideal. Perhaps implementing it only for UninitializedStateMachine<M> would be better? I'm open to suggestions.

[christianheussy]

Interior mutability is one option, however this leads to a lot of extra code to deal with the requirement that everything captured in the async state methods must be Send. This leads to wrapping everything owned by the state machine with Arc<Mutex<_>> which is verbose and not ideal.

[christianheussy]

Ended up using Context instead, which permits passing in mutable references. I like the ergonomics of using items owned by the state machine, but this is an acceptable compromise.

[mdeloof]

I'm open to adding some way to get a mutable reference to the state machine's internal data, but we probably shouldn't use DerefMut for this, as the docs state that it should only be implemented on smart pointers.

An alternative would be to provide methods such as inner() and inner_mut() to access the internal data. But as you say, having mutable access at all times isn't ideal as it could break the state machine's internal invariants. It should be fine to use on UninitializedStateMachine, but maybe we should mark those methods unsafe if we were to implement them on InitializedStateMachine?