Implementing Display for error about sppark

[dot-asm]

If instead ...

Does it have to be "instead"? I mean what prevents us from just adding it? Not that I have some attachment to From, it's just that yanking it is likely to be considered bad style. You kind of add something new and remove old later...

If I should provide a pull request, let me know.

If it's just Display, I can handle it. But maybe there is more to it. Say it might be appropriate to implement Error trait. In other words if you can think of something more nuanced than adding Display, then a pull request would be appreciated. Well, it would be appreciated in either case, but we do need our Rust exercises too, you know :-)

[vmx]

Does it have to be "instead"? I mean what prevents us from just adding it?

You're right. Display could just be added, you can keep the From<Error> for String implementation.

If it's just Display, I can handle it. But maybe there is more to it.

Indeed the "proper" way would be to also implement the Error trait.

Well, it would be appreciated in either case, but we do need our Rust exercises too, you know :-)

Then I'd take a look at https://crates.io/crates/thiserror. This is a simple macro that generates the code you'd otherwise create manually. This would then also implement Dsiplay.

One thing I don't fully understand is, why the CUDA error is wrappen in a macro. Why not just having it as a normal struct and importing it whenever you want to use it?

[dot-asm]

I've added Display, simplified for now as Error needs an overhaul. Because we've discussed using it for conveying errors other than CUDA.

One thing I don't fully understand is, why the CUDA error is wrappen in a macro.

The original rationale was that we don't know if it will be needed (because target application might not be using CUDA code), and providing a macro would give the application freedom to decide. But since it's discussed to be used for other errors, it's arguably appropriate to change that...

[vmx]

Thanks a lot for implementing Display, it's kind of nicer to be able to call .to_string().