big_brain/
evaluators.rs

1//! Utilities for turning values within a certain range into different curves.
2
3use bevy::prelude::*;
4
5/**
6Trait that any evaluators must implement. Must return an `f32` value between `0.0..=100.0`.
7 */
8#[reflect_trait]
9pub trait Evaluator: std::fmt::Debug + Sync + Send {
10    fn evaluate(&self, value: f32) -> f32;
11}
12
13/**
14[`Evaluator`] for linear values. That is, there's no curve to the value mapping.
15 */
16#[derive(Debug, Clone, Reflect)]
17pub struct LinearEvaluator {
18    xa: f32,
19    ya: f32,
20    yb: f32,
21    dy_over_dx: f32,
22}
23
24impl LinearEvaluator {
25    pub fn new() -> Self {
26        Self::new_full(0.0, 0.0, 1.0, 1.0)
27    }
28    pub fn new_inversed() -> Self {
29        Self::new_ranged(1.0, 0.0)
30    }
31    pub fn new_ranged(min: f32, max: f32) -> Self {
32        Self::new_full(min, 0.0, max, 1.0)
33    }
34    fn new_full(xa: f32, ya: f32, xb: f32, yb: f32) -> Self {
35        Self {
36            xa,
37            ya,
38            yb,
39            dy_over_dx: (yb - ya) / (xb - xa),
40        }
41    }
42}
43
44impl Default for LinearEvaluator {
45    fn default() -> Self {
46        Self::new()
47    }
48}
49
50impl Evaluator for LinearEvaluator {
51    fn evaluate(&self, value: f32) -> f32 {
52        clamp(
53            self.ya + self.dy_over_dx * (value - self.xa),
54            self.ya,
55            self.yb,
56        )
57    }
58}
59
60/**
61[`Evaluator`] with an exponent curve. The value will grow according to its `power` parameter.
62 */
63#[derive(Debug, Clone, Reflect)]
64pub struct PowerEvaluator {
65    xa: f32,
66    ya: f32,
67    xb: f32,
68    power: f32,
69    dy: f32,
70}
71
72impl PowerEvaluator {
73    pub fn new(power: f32) -> Self {
74        Self::new_full(power, 0.0, 0.0, 1.0, 1.0)
75    }
76    pub fn new_ranged(power: f32, min: f32, max: f32) -> Self {
77        Self::new_full(power, min, 0.0, max, 1.0)
78    }
79    fn new_full(power: f32, xa: f32, ya: f32, xb: f32, yb: f32) -> Self {
80        Self {
81            power: clamp(power, 0.0, 10000.0),
82            dy: yb - ya,
83            xa,
84            ya,
85            xb,
86        }
87    }
88}
89
90impl Default for PowerEvaluator {
91    fn default() -> Self {
92        Self::new(2.0)
93    }
94}
95
96impl Evaluator for PowerEvaluator {
97    fn evaluate(&self, value: f32) -> f32 {
98        let cx = clamp(value, self.xa, self.xb);
99        self.dy * ((cx - self.xa) / (self.xb - self.xa)).powf(self.power) + self.ya
100    }
101}
102
103/**
104[`Evaluator`] with a "Sigmoid", or "S-like" curve.
105 */
106#[derive(Debug, Clone, Reflect)]
107pub struct SigmoidEvaluator {
108    xa: f32,
109    xb: f32,
110    ya: f32,
111    yb: f32,
112    k: f32,
113    two_over_dx: f32,
114    x_mean: f32,
115    y_mean: f32,
116    dy_over_two: f32,
117    one_minus_k: f32,
118}
119
120impl SigmoidEvaluator {
121    pub fn new(k: f32) -> Self {
122        Self::new_full(k, 0.0, 0.0, 1.0, 1.0)
123    }
124
125    pub fn new_ranged(k: f32, min: f32, max: f32) -> Self {
126        Self::new_full(k, min, 0.0, max, 1.0)
127    }
128
129    fn new_full(k: f32, xa: f32, ya: f32, xb: f32, yb: f32) -> Self {
130        let k = clamp(k, -0.99999, 0.99999);
131        Self {
132            xa,
133            xb,
134            ya,
135            yb,
136            two_over_dx: (2.0 / (xb - ya)).abs(),
137            x_mean: (xa + xb) / 2.0,
138            y_mean: (ya + yb) / 2.0,
139            dy_over_two: (yb - ya) / 2.0,
140            one_minus_k: 1.0 - k,
141            k,
142        }
143    }
144}
145
146impl Evaluator for SigmoidEvaluator {
147    fn evaluate(&self, x: f32) -> f32 {
148        let cx_minus_x_mean = clamp(x, self.xa, self.xb) - self.x_mean;
149        let numerator = self.two_over_dx * cx_minus_x_mean * self.one_minus_k;
150        let denominator = self.k * (1.0 - 2.0 * (self.two_over_dx * cx_minus_x_mean)).abs() + 1.0;
151        clamp(
152            self.dy_over_two * (numerator / denominator) + self.y_mean,
153            self.ya,
154            self.yb,
155        )
156    }
157}
158
159impl Default for SigmoidEvaluator {
160    fn default() -> Self {
161        Self::new(-0.5)
162    }
163}
164
165pub(crate) fn clamp<T: PartialOrd>(val: T, min: T, max: T) -> T {
166    let val = if val > max { max } else { val };
167    if val < min {
168        min
169    } else {
170        val
171    }
172}
big_brain/evaluators.rs

big_brain/
evaluators.rs
