Hanno Braun

Due to #193, a lot has changed since I opened this issue. Any investigation should include whether using the `math` module from the host application instead of nalgebra makes sense....

> Shouldn't the intrinsic types from `math` be exposed for the modeling as well? At the moment a lot of functions are taking raw `f64` or `Vec` which is problematic...

Updated issue title and description to reflect the current situation.

I've experimented with this. The compile times of the star model go up from ~0.5s to ~1.5-1.6s when integrating the math types into `fj`. This is not acceptable. I consider...

Thank you for your comment, @LoganDark. I didn't know about `ultraviolet`. It looks really interesting! Although it's probably not the best basis to implement our current `fj-math` API, seeing how...

> Is there any particular reason why you couldn't use a sealed trait for this? AFAIK, it's definitely possible to smooth over the different structs using a trait. That should...

> Would certainly be less of an impact than `nalgebra`. The Rust compiler is designed to handle hundreds of different traits per compilation, but `nalgebra` incurs deep monomorphization costs (tons...

I hadn't heard of [`cint`](https://crates.io/crates/cint), so thanks for letting me know! Might be a good dependency to use to represent colors, although it doesn't seem to be as accepted in...

I don't think that's right? According to crates.io, `ultraviolet` and `cint` are by Gray Olson, while `mint` is by Dzmitry Malyshau.

Hey @Michael-F-Bryan, thank you for doing this! This looks good, in principle, but that segmentation fault... I didn't notice anything suspicious. I wonder, if we could be doing something to...