Josh Stone

That ratio is not the input we have here, nor the 17-digit decimal approximation, but rather a pair of `f64` -- which are the “binary64” type defined in IEEE 754-2008....

> Just for my understanding: My reference value for `Log[z]` was not correct, because the rational number that I used to approximate `z` did not quite match the decimal `f64`...

> I suspect that removing the relationship between `CheckedAdd` and `Add` might be a breaking change, Yes, that is a breaking change, sorry. > I wonder if the real "right"...

I found the commit history, but not much discussion. For this particular trait: * It was removed with its methods transferred to `trait Int` in https://github.com/rust-lang/rust/commit/e51cc089da7f5a067d348ee48f494c5bca662f95 * `trait Int` was...

> * detect major version change in this crate, so breaking change will be enabled automatically on toml update. Major version of what? I don't know what you mean here....

It **is** a breaking change, because users can no longer rely on that equivalence. ```rust use num_traits::WrappingMul; pub trait WrappingMul2: Sized + std::ops::Mul { fn wrapping_mul(&self, v: &Self) -> Self;...

It's an interesting question! > I am probably oversimplyfing things, but I imagine currently the IndexMap is backed by "something like a Vec", The core *is* literally a `Vec`, and...

I think you've nerd-sniped me, because I'm still thinking about this, and I may go ahead and try a fork based on `VecDeque` myself. Let me know if you're already...

No worries! I think I will explore this, and I'll share it here if I do.

I did make a local prototype -- I'll see about cleaning that up and publishing a new crate.