1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193

use glam::Vec2;
use rand::distributions::Uniform;
use rand::Rng;
use sphere::sphere_volume;

use crate::algorithm::{Algorithm, Creator};
use crate::utils::*;
use crate::Builder;

/// Generates uniform maximal Poisson disk samplings with O(n2<sup>d</sup>) time and O(n2<sup>d</sup>) space complexity relative to the number of samples generated and the dimensionality of the sampling volume.
/// Based on Ebeida, Mohamed S., et al. "A Simple Algorithm for Maximal Poisson‐Disk Sampling in High Dimensions." Computer Graphics Forum. Vol. 31. No. 2pt4. Blackwell Publishing Ltd, 2012.
#[derive(Debug, Clone, Copy)]
pub struct Ebeida;

impl Creator for Ebeida {
    type Algo = Algo;

    fn create(poisson: &Builder) -> Self::Algo {
        let grid = Grid::new(poisson.radius, poisson.poisson_type);
        let mut indices = Vec::with_capacity(grid.cells() * 2);
        let choices = (0..grid.side()).map(|i| i as f32).collect::<Vec<_>>();
        indices.extend(each_combination(&choices));
        let a = 0.3;
        Algo {
            a,
            grid,
            throws: (a * indices.len() as f64).ceil() as usize,
            range: Uniform::new(0, indices.len()),
            indices,
            level: 0,
            success: 0,
            outside: vec![],
            mantissa_digits: f32::MANTISSA_DIGITS as usize,
        }
    }
}

/// Implementation for the Ebeida algorithm
pub struct Algo {
    grid: Grid,
    indices: Vec<Vec2>,
    level: usize,
    range: Uniform<usize>,
    throws: usize,
    success: usize,
    outside: Vec<Vec2>,
    mantissa_digits: usize,
    a: f64,
}

impl Algorithm for Algo {
    fn next<R>(&mut self, poisson: &mut Builder, rng: &mut R) -> Option<mint::Vector2<f32>>
    where
        R: Rng,
    {
        if self.indices.is_empty() {
            return None;
        }
        while self.level < self.mantissa_digits {
            while self.throws > 0 {
                self.throws -= 1;
                let index = rng.sample(self.range);
                let cur = self.indices[index].clone();
                let parent = get_parent(cur.clone(), self.level);
                if !self
                    .grid
                    .get(parent.clone())
                    .expect("Indexing base grid by valid parent failed.")
                    .is_empty()
                {
                    self.indices.swap_remove(index);
                    if self.indices.is_empty() {
                        return None;
                    }
                    self.range = Uniform::new(0, self.indices.len());
                } else {
                    let sample = choose_random_sample(rng, &self.grid, cur.clone(), self.level);
                    if is_disk_free(
                        &self.grid,
                        poisson,
                        cur.clone(),
                        self.level,
                        sample.clone(),
                        &self.outside,
                    ) {
                        self.grid
                            .get_mut(parent)
                            .expect("Indexing base grid by already indexed valid parent failed.")
                            .push(sample.clone());
                        self.indices.swap_remove(index);
                        if !self.indices.is_empty() {
                            self.range = Uniform::new(0, self.indices.len());
                        }
                        self.success += 1;
                        return Some(sample.into());
                    }
                }
            }
            self.subdivide(&poisson);
            if self.indices.is_empty() {
                return None;
            }
            self.range = Uniform::new(0, self.indices.len());
            self.throws = (self.a * self.indices.len() as f64).ceil() as usize;
            self.level += 1;
        }
        let index = rng.sample(self.range);
        let cur = self.indices.swap_remove(index);
        let side = 2usize.pow(self.level as u32);
        let sample = index_to_sample(&cur, side);
        if is_disk_free(
            &self.grid,
            poisson,
            cur.clone(),
            self.level,
            sample.clone(),
            &self.outside,
        ) {
            Some(sample.into())
        } else {
            None
        }
    }

    fn size_hint(&self, poisson: &Builder) -> (usize, Option<usize>) {
        // Calculating lower bound should work because we calculate how much volume is left to be filled at worst case and
        // how much sphere can fill it at best case and just figure out how many fills are still needed.
        let side = 2usize.pow(self.level as u32);
        let spacing = self.grid.cell() / (side as f32);
        let grid_volume = (self.indices.len() as f32) * spacing.powi(2);
        let sphere_volume = sphere_volume(2.0 * poisson.radius, 2);
        let lower = grid_volume / sphere_volume;
        let mut lower = lower.floor() as usize;
        if lower > 0 {
            lower -= 1;
        }
        // Calculating upper bound should work because there is this many places left in the grid and no more can fit into it.
        let upper = self.grid.cells() - self.success;
        (lower, Some(upper))
    }

    fn restrict(&mut self, sample: mint::Vector2<f32>) {
        let sample: Vec2 = sample.into();
        self.success += 1;
        let index = sample_to_index(&sample, self.grid.side());
        if let Some(g) = self.grid.get_mut(index) {
            g.push(sample);
        } else {
            self.outside.push(sample);
        }
    }

    fn stays_legal(&self, poisson: &Builder, sample: mint::Vector2<f32>) -> bool {
        let sample: Vec2 = sample.into();
        let index = sample_to_index(&sample, self.grid.side());
        is_disk_free(&self.grid, poisson, index, 0, sample.clone(), &self.outside)
    }
}

impl Algo {
    fn subdivide(&mut self, poisson: &Builder) {
        let choices = &[0.0, 1.0];
        let (grid, outside, level) = (&self.grid, &self.outside, self.level);
        self.indices.flat_map_inplace(|i| {
            each_combination(choices)
                .map(move |n: Vec2| n + i.clone() * 2.0)
                .filter(|c| !covered(grid, poisson, outside, c.clone(), level + 1))
        });
    }
}

fn covered(
    grid: &Grid,
    poisson: &Builder,
    outside: &[Vec2],
    index: Vec2,
    level: usize,
) -> bool {
    // TODO: This does 4^d checking of points even though it could be done 3^d
    let side = 2usize.pow(level as u32);
    let spacing = grid.cell() / (side as f32);
    let sqradius = (2.0 * poisson.radius).powi(2);
    let parent = get_parent(index.clone(), level);
    each_combination(&[0.0, 1.0])
        .map(|t| (index.clone() + t) * spacing)
        .all(|t| {
            each_combination(&[-2.0, -1.0, 0.0, 1.0, 2.0])
                .filter_map(|t| grid.get(parent.clone() + t))
                .flat_map(|t| t)
                .any(|v| sqdist(v.clone(), t.clone(), poisson.poisson_type) < sqradius)
                || !is_valid(poisson, &outside, t)
        })
}