Why do commit methods require `&mut self`?

[helloskyline-a11y]

Hey there!

commit() requires &mut self, suggesting exclusive access enforced by the type system, while the implementation appears to rely on interior mutability and actor-style message passing (Mutex, atomics, channels) on database/mod.rs.

This makes the API look like a mix of static exclusivity and runtime synchronization models. Is there a specific reason for exposing commit() differently from the other operations?

[helloskyline-a11y]

Hi @poljar, sorry to bother you. But have you any answer?
From a user, having commit() require &mut self makes it a bit harder to share the database instance across threads. Even if the internal implementation relies on channels, we are forced to wrap the whole DB handle in a Mutex just to be able to call commit(), which doesn't happen often.
I'm not sure if I'm missing something, or if I can change the signature on my end. Thanks!

[poljar]

Hi there, sorry for the delay.

This is just a mistake. The initial version might not have used channels and thus required &mut self, but as the switch was made to a channel-based solution, mostly due to the requirements of the bindings if I remember correctly, this requirement disappeared.

The unused_mut lint might help us find all such instance.

[helloskyline-a11y]

Hey! No worries about the delay, I know it can be hard to maintain an open source lib 😀
Would you like me to go ahead and open a PR to fix them?

Just a small heads-up: relaxing &mut self to &self is backward compatible, but in some edge cases, it could technically be a breaking change. I think if we do that, we should upgrade seshat version.

[poljar]

A PR would be welcome.

Yeah, we can just bump the major version to be extra cautious. I don't think that's going to be a big deal.