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
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
use bv::BitVec;
use fnv::FnvHasher;
use rand::{self, Rng};
use serde::{Deserialize, Serialize};
use std::fmt;
use std::{cmp, hash::Hasher, marker::PhantomData};
pub trait BloomHashIndex {
fn hash_at_index(&self, hash_index: u64) -> u64;
}
#[derive(Serialize, Deserialize, Default, Clone, PartialEq, AbiExample)]
pub struct Bloom<T: BloomHashIndex> {
pub keys: Vec<u64>,
pub bits: BitVec<u64>,
num_bits_set: u64,
_phantom: PhantomData<T>,
}
impl<T: BloomHashIndex> fmt::Debug for Bloom<T> {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
write!(
f,
"Bloom {{ keys.len: {} bits.len: {} num_set: {} bits: ",
self.keys.len(),
self.bits.len(),
self.num_bits_set
)?;
const MAX_PRINT_BITS: u64 = 10;
for i in 0..std::cmp::min(MAX_PRINT_BITS, self.bits.len()) {
if self.bits.get(i) {
write!(f, "1")?;
} else {
write!(f, "0")?;
}
}
if self.bits.len() > MAX_PRINT_BITS {
write!(f, "..")?;
}
write!(f, " }}")
}
}
impl<T: BloomHashIndex> solana_sdk::sanitize::Sanitize for Bloom<T> {}
impl<T: BloomHashIndex> Bloom<T> {
pub fn new(num_bits: usize, keys: Vec<u64>) -> Self {
let bits = BitVec::new_fill(false, num_bits as u64);
Bloom {
keys,
bits,
num_bits_set: 0,
_phantom: PhantomData::default(),
}
}
pub fn random(num_items: usize, false_rate: f64, max_bits: usize) -> Self {
let m = Self::num_bits(num_items as f64, false_rate);
let num_bits = cmp::max(1, cmp::min(m as usize, max_bits));
let num_keys = Self::num_keys(num_bits as f64, num_items as f64) as usize;
let keys: Vec<u64> = (0..num_keys).map(|_| rand::thread_rng().gen()).collect();
Self::new(num_bits, keys)
}
fn num_bits(num_items: f64, false_rate: f64) -> f64 {
let n = num_items;
let p = false_rate;
((n * p.ln()) / (1f64 / 2f64.powf(2f64.ln())).ln()).ceil()
}
fn num_keys(num_bits: f64, num_items: f64) -> f64 {
let n = num_items;
let m = num_bits;
if n == 0.0 {
0.0
} else {
1f64.max(((m / n) * 2f64.ln()).round())
}
}
fn pos(&self, key: &T, k: u64) -> u64 {
key.hash_at_index(k) % self.bits.len()
}
pub fn clear(&mut self) {
self.bits = BitVec::new_fill(false, self.bits.len());
self.num_bits_set = 0;
}
pub fn add(&mut self, key: &T) {
for k in &self.keys {
let pos = self.pos(key, *k);
if !self.bits.get(pos) {
self.num_bits_set += 1;
self.bits.set(pos, true);
}
}
}
pub fn contains(&self, key: &T) -> bool {
for k in &self.keys {
let pos = self.pos(key, *k);
if !self.bits.get(pos) {
return false;
}
}
true
}
}
fn slice_hash(slice: &[u8], hash_index: u64) -> u64 {
let mut hasher = FnvHasher::with_key(hash_index);
hasher.write(slice);
hasher.finish()
}
impl<T: AsRef<[u8]>> BloomHashIndex for T {
fn hash_at_index(&self, hash_index: u64) -> u64 {
slice_hash(self.as_ref(), hash_index)
}
}
#[cfg(test)]
mod test {
use super::*;
use solana_sdk::hash::{hash, Hash};
#[test]
fn test_bloom_filter() {
let bloom: Bloom<Hash> = Bloom::random(0, 0.1, 100);
assert_eq!(bloom.keys.len(), 0);
assert_eq!(bloom.bits.len(), 1);
let bloom: Bloom<Hash> = Bloom::random(10, 0.1, 100);
assert_eq!(bloom.keys.len(), 3);
assert_eq!(bloom.bits.len(), 48);
let bloom: Bloom<Hash> = Bloom::random(100, 0.1, 100);
assert_eq!(bloom.keys.len(), 1);
assert_eq!(bloom.bits.len(), 100);
}
#[test]
fn test_add_contains() {
let mut bloom: Bloom<Hash> = Bloom::random(100, 0.1, 100);
bloom.keys = vec![0, 1, 2, 3];
let key = hash(b"hello");
assert!(!bloom.contains(&key));
bloom.add(&key);
assert!(bloom.contains(&key));
let key = hash(b"world");
assert!(!bloom.contains(&key));
bloom.add(&key);
assert!(bloom.contains(&key));
}
#[test]
fn test_random() {
let mut b1: Bloom<Hash> = Bloom::random(10, 0.1, 100);
let mut b2: Bloom<Hash> = Bloom::random(10, 0.1, 100);
b1.keys.sort();
b2.keys.sort();
assert_ne!(b1.keys, b2.keys);
}
#[test]
fn test_filter_math() {
assert_eq!(Bloom::<Hash>::num_bits(100f64, 0.1f64) as u64, 480u64);
assert_eq!(Bloom::<Hash>::num_bits(100f64, 0.01f64) as u64, 959u64);
assert_eq!(Bloom::<Hash>::num_keys(1000f64, 50f64) as u64, 14u64);
assert_eq!(Bloom::<Hash>::num_keys(2000f64, 50f64) as u64, 28u64);
assert_eq!(Bloom::<Hash>::num_keys(2000f64, 25f64) as u64, 55u64);
assert_eq!(Bloom::<Hash>::num_keys(20f64, 1000f64) as u64, 1u64);
}
#[test]
fn test_debug() {
let mut b: Bloom<Hash> = Bloom::new(3, vec![100]);
b.add(&Hash::default());
assert_eq!(
format!("{:?}", b),
"Bloom { keys.len: 1 bits.len: 3 num_set: 1 bits: 001 }"
);
let mut b: Bloom<Hash> = Bloom::new(1000, vec![100]);
b.add(&Hash::default());
b.add(&hash(&[1, 2]));
assert_eq!(
format!("{:?}", b),
"Bloom { keys.len: 1 bits.len: 1000 num_set: 2 bits: 0000000000.. }"
);
}
}