Immutable Convergency

[JP-Ellis]

I've noticed that the root finding algorithms all take a mutable reference to the Convergency trait, which seems odd. In particular, I was hoping to have a const Convergency set for a module, but this is impossible to achieve because it must be mutable.

Are there any use case for a mutable convergency test other than in DebugConvergency? I would have thought that if DebugConvergency is the only reason why this is used, would it not be better to make it internally mutable through a RwLock (or similar)? As that is for debugging, the extra overhead should not be too much of a concern.

In fact, would it be better if the functions to use an associated type instead of &mut dyn Convergency? This would allow for mutable cases as well as immutable cases. I've created a small gist that shows what I mean.

[vorot]

Thanks for using this library.

There are applications that change convergence criteria depending on circumstances. Those applications need to maintain a (mutable) state machine inside the implementation of the trait. DebugConvergency is an oversimplified example of such applications.

An interface that uses associated types is acceptable if it allows both mutable and immutable implementation of the trait. Would you make a patch/pull request with the interface you want?

[JP-Ellis]

I'll have a go at implementing this and see how it goes.

[KarelPeeters]

As a related issue, why dyn? Why not another generic type parameter? Typically this stuff will be called in performance-intensive code so why waste time with dynamic dispatch?