Add and Substract - clamp after operations?

[sidred]

For colors with known ranges like rgb, hsl etc., should we clamp the values? A possible issue is with the f32 -> u8 conversion panicking if not clamped.

[Ogeon]

That may mess with the math. I think it's better to clamp just before converting. Isn't that already implemented?

[sidred]

Currently not implemented even in the current master.

impl<T: Float> RgbPixel<T> for (u8, u8, u8, u8) {
    fn from_rgba(red: T, green: T, blue: T, alpha: T) -> (u8, u8, u8, u8) {
        ((red * T::from(255.0).unwrap()).to_u8().unwrap(),
         (green * T::from(255.0).unwrap()).to_u8().unwrap(),
         (blue * T::from(255.0).unwrap()).to_u8().unwrap(),
         (alpha * T::from(255.0).unwrap()).to_u8().unwrap())
    }

    fn to_rgba(&self) -> (T, T, T, T) {
        let (r, g, b, a) = *self;
        (T::from(r).unwrap() / T::from(255.0).unwrap(),
         T::from(g).unwrap() / T::from(255.0).unwrap(),
         T::from(b).unwrap() / T::from(255.0).unwrap(),
         T::from(a).unwrap() / T::from(255.0).unwrap())
    }
}

[Ogeon]

Yes, it is in each conversion function.

[sidred]

I've at this looked at this function link

I am not sure I understand it correctly. Lets say I have 2 Rgb and I add them and convert to [u8;3]. These is no clamping in this process, is there?

[Ogeon]

No, it happens in the functions I linked to, just before those are called. It may still be a good idea to clamp after multiplying with 255, because of how floating point math works.