| /// A trait for extending [`prim@f32`] and [`prim@f64`] with extra methods. | |
| pub trait FloatExt { | |
| /// Performs a linear interpolation between `self` and `rhs` based on the value `s`. | |
| /// | |
| /// When `s` is `0`, the result will be `self`. When `s` is `1`, the result | |
| /// will be `rhs`. When `s` is outside of the range `[0, 1]`, the result is linearly | |
| /// extrapolated. | |
| #[must_use] | |
| fn lerp(self, rhs: Self, s: Self) -> Self; | |
| /// Returns `v` normalized to the range `[a, b]`. | |
| /// | |
| /// When `v` is equal to `a` the result will be `0`. When `v` is equal to `b` will be `1`. | |
| /// | |
| /// When `v` is outside of the range `[a, b]`, the result is linearly extrapolated. | |
| /// | |
| /// `a` and `b` must not be equal, otherwise the result will be either infinite or `NAN`. | |
| fn inverse_lerp(a: Self, b: Self, v: Self) -> Self; | |
| /// Remap `self` from the input range to the output range. | |
| /// | |
| /// When `self` is equal to `in_start` this returns `out_start`. | |
| /// When `self` is equal to `in_end` this returns `out_end`. | |
| /// | |
| /// When `self` is outside of the range `[in_start, in_end]`, the result is linearly extrapolated. | |
| /// | |
| /// `in_start` and `in_end` must not be equal, otherwise the result will be either infinite or `NAN`. | |
| #[must_use] | |
| fn remap(self, in_start: Self, in_end: Self, out_start: Self, out_end: Self) -> Self; | |
| } |