Struct parameterized with colorspace newtype not accepted by compiler

[mechazoidal]

Hi,
I'm working with texture generation using this crate and would like to have the pixel operations use saturating_* math functions instead of the default.

Per the docs, I implemented a colorspace newtype and a different "add" implementation for it. However, the compiler insists I must provide Add for the parametrized type itself (rgb::RGBA<openktg::LinearLight, u16>): which I can't do as this is an external crate.

I am relatively new to Rust, so is this something I'm missing, or a legitimate bug with parameterized RGB structs?

Here's a minimal gist to illustrate: https://gist.github.com/rust-play/40a8fc9422f88c5af8bcd5900fe9230a

[kornelski]

Rust has "orphan rules" which guarantee that globally for every possible program, there can't be more than one implementation of any trait for any type.

Simplifying, you can implement a trait for a type only if either:

• it's your type, or
• it's your trait.

You can't implement someone else's trait (like std::ops::Add) for someone else's type (like openktg::LinearLight, since there's no guarantee that nobody else could create a conflicting implementation.

However, I may be able to add a more general Add implementation. I'll check that.

[kornelski]

I've released v0.8.9 which allows px+px for arbitrary newtypes.

Note that px.map(||) can't work if types of rgb and alpha are different, because Rust is statically typed, and the closure can't accept more than one type (it's still statically typed, even though it doesn't look like it).
There's map_rgb and new_alpha for this reason.

[mechazoidal]

Thank you for the prompt turnaround! The rules make sense, I have no problem using map_rgb/new_alpha, and this should be what I need!

[ripytide]

This issue looks to have been resolved in v0.8.9, I'd recommend closing it.