Genericity over f32 and f64

[guillaumegomez]

Genericity over f32 and f64

[GuillaumeGomez]

I think it's a good idea. However I'm quite busy right now with gtk so, if you want it quickly, you'll have to do it yourself. Otherwise, I'll try to take a took in one or two week.

[golddranks]

I can try and look into it. However, I just realised that that specialised versions of the functions aren't (of course) going to work without the trait specialisation support which is slowly coming along, but still quite far off. Dang! Well, I'll try and think about this.

[GuillaumeGomez]

I was thinking to create a Float type alias (like you proposed) on f32 or f64, depending of the option. However, replacing all f32 will take a while (there are... a lot ^^).

[iduartgomez]

Somehow related to this, I was working with matrix operations and unfortunately the linear algebra ops use the double precision matrix while the BLAS ops use the float matrix.

I get this is a limitation or design with GSL itself but I wonder if there is someway to get around it doing some magic in rust with custom #[repr(C)] structs casting between f64 and f32 types to avoid unnecessary copying.

[GuillaumeGomez]

I don't have much time to work on this crate lately unfortunately... My primary focus for the moment is to make the binding to the 2.0 version. But if you have an idea on how to do this, I'd gladly accept a PR.

[GuillaumeGomez]

There are now MatrixF32 and MatrixF64 types as well as equivalent for other types. I can't make the GSL functions generic over them unfortunately though. So I think everything that could be done was already done.