A Fenwick tree or binary indexed tree/bit indexed tree is a data structure
that supports the following two operations efficiently over an array of numbers a[0..n]
:
- Calculate a prefix sum:
a[0] + a[1] + ... + a[i]
- Update one element:
a[i] += delta
With a naïve implementation, only one of the operations can be made to have constant time
complexity while the other one has to be linear. With Fenwick tree, both take only O(log(N))
.
This crate is no_std
and has no (non-dev) dependencies.
Use the array
module for operations on a 1D Fenwick tree:
use fenwick::array::{update, prefix_sum};
let fw = &mut [0i32; 10]; // backing array of Fenwick tree (NOT original array!)
assert_eq!(prefix_sum(fw, 0), 0);
assert_eq!(prefix_sum(fw, 9), 0);
update(fw, 0, 3); // original array: [3, 0, 0, 0, 0, 0, 0, 0, 0, 0]
assert_eq!(prefix_sum(fw, 0), 3);
assert_eq!(prefix_sum(fw, 9), 3);
update(fw, 5, 9); // original array: [3, 0, 0, 0, 0, 9, 0, 0, 0, 0]
assert_eq!(prefix_sum(fw, 4), 3);
assert_eq!(prefix_sum(fw, 5), 12);
assert_eq!(prefix_sum(fw, 6), 12);
update(fw, 4, -5); // original array: [3, 0, 0, 0, -5, 9, 0, 0, 0, 0]
assert_eq!(prefix_sum(fw, 4), -2);
assert_eq!(prefix_sum(fw, 5), 7);
update(fw, 0, -2); // original array: [1, 0, 0, 0, -5, 9, 0, 0, 0, 0]
assert_eq!(prefix_sum(fw, 4), -4);
assert_eq!(prefix_sum(fw, 5), 5);
Use the index
module to implement multidimensional Fenwick trees:
use fenwick::index::zero_based::{down, up};
const MAX: usize = 1000;
fn update(i: usize, j: usize, k: usize, delta: i32) {
for ii in up(i, MAX) {
for jj in up(j, MAX) {
for kk in up(k, MAX) {
/* increment 3D array at [ii, jj, kk] by delta */
}
}
}
}
fn prefix_sum(i: usize, j: usize, k: usize) -> i32 {
let mut sum = 0i32;
for ii in down(i) {
for jj in down(j) {
for kk in down(k) {
/* increment sum by 3D array at [ii, jj, kk] */
}
}
}
sum
}