The documentation stated that Conversion is handled like this: (value * coefficient()) + constant()

But using this code snipped for a manual conversion like Fahrenheit to Kelvin this breaks:

fn test() {
    let f: f32 = 2.0;
    let k = convert::<uom::si::thermodynamic_temperature::degree_fahrenheit>(f);
    println!("{k:?}");
}

fn convert<N>(value: f32) -> f32
where
    N: uom::Conversion<f32, T = f32>,
{
    (value * N::coefficient()) + N::constant(uom::ConstantOp::Add)
}

The result is 460.78113, but 2 degrees Fahrenheit is 256.483

Rewriting the equation to (value + constant()) * coefficient() results in 256.48337

So there is an inconsistency between code and documentation.