22241361593

Committed 20 Feb 2026 09:11PM UTC coverage: 53.913% (-0.06%) from 53.968%

Build # 22241361593

Build Type

push

github

Committed by

vigna

Commit Message

Removed anyhow, rand is optional (implied module gated by feature)

Coverage Stats

0 of 18 new or added lines in 4 files covered. (0.0%)

2 existing lines in 1 file now uncovered.

2108 of 3910 relevant lines covered (53.91%)

3096788.1 hits per line

Source File
Press 'n' to go to next uncovered line, 'b' for previous

0.0

/src/utils/implied.rs

/*
 * SPDX-FileCopyrightText: 2025 Tommaso Fontana
 *
 * SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
 */

use crate::utils::FindChangePoints;
use alloc::vec::Vec;
use rand::distr::weighted::WeightedIndex;
use rand::prelude::*;

/// Given the len function of a code, generates data that allows to sample
/// its implied distribution, i.e. a code-word with length l has
/// probability 2^(-l).
///
/// This code works only with monotonic non-decreasing len functions.
///
/// Returns two vectors, the first one contains the input values where the
/// function changes value and the code length at that point. The second
/// vector contains the probability of each code length.
///
/// Since we cannot write more than 64 bits at once, the codes are limited to
/// 128 bits.
pub fn get_implied_distribution(f: impl Fn(u64) -> usize) -> (Vec<(u64, usize)>, Vec<f64>) {
    let change_points = FindChangePoints::new(f)
        .take_while(|(_input, len)| *len <= 128)
        .collect::<Vec<_>>();

    // convert to len probabilities
    let probabilities = change_points
        .windows(2)
        .map(|window| {
            let (input, len) = window[0];
            let (next_input, _next_len) = window[1];
            let prob = 2.0_f64.powi(-(len as i32));
            prob * (next_input - input) as f64
        })
        .collect::<Vec<_>>();
    // TODO: this ignores the last change point

    (change_points, probabilities)
}

#[derive(Clone, Copy, Debug)]
/// An infinite iterator that always returns ().
pub struct InfiniteIterator;

impl Iterator for InfiniteIterator {
    type Item = ();

    #[inline(always)]
    fn next(&mut self) -> Option<Self::Item> {
        Some(())
    }
}

/// Returns an **infinite iterator** of samples from the implied distribution of
/// the given code length function.
/// The function f should be the len function of the code.
///
/// This code works only with monotonic non-decreasing len functions and
/// the codes are limited to 128 bits as we cannot write more than 64 bits at once.
///
/// # Example
///
/// ```rust
/// use dsi_bitstream::utils::sample_implied_distribution;
/// use dsi_bitstream::codes::len_gamma;
/// use rand::SeedableRng;
/// use rand::rngs::SmallRng;
///
/// let mut rng = SmallRng::seed_from_u64(42);
/// let vals: Vec<u64> = sample_implied_distribution(len_gamma, &mut rng)
///     .take(1000).collect::<Vec<_>>();
///
/// assert_eq!(vals.len(), 1000);
/// ```
pub fn sample_implied_distribution(
    f: impl Fn(u64) -> usize,
    rng: &mut impl Rng,
) -> impl Iterator<Item = u64> + '_ {
    let (change_points, probabilities) = get_implied_distribution(f);
    let dist = WeightedIndex::new(probabilities).unwrap();

    InfiniteIterator.map(move |_| {
        // sample a len with the correct probability
        let idx = dist.sample(rng);
        // now we sample a random value with the sampled len
        let (start_input, _len) = change_points[idx];
        let (end_input, _len) = change_points[idx + 1];
        rng.random_range(start_input..end_input)
    })
}

1	/*
2	* SPDX-FileCopyrightText: 2025 Tommaso Fontana
3	*
4	* SPDX-License-Identifier: Apache-2.0 OR LGPL-2.1-or-later
5	*/
6
7	use crate::utils::FindChangePoints;
8	use alloc::vec::Vec;
9	use rand::distr::weighted::WeightedIndex;
10	use rand::prelude::*;
11
12	/// Given the len function of a code, generates data that allows to sample
13	/// its implied distribution, i.e. a code-word with length l has
14	/// probability 2^(-l).
15	///
16	/// This code works only with monotonic non-decreasing len functions.
17	///
18	/// Returns two vectors, the first one contains the input values where the
19	/// function changes value and the code length at that point. The second
20	/// vector contains the probability of each code length.
21	///
22	/// Since we cannot write more than 64 bits at once, the codes are limited to
23	/// 128 bits.
UNCOV 24	pub fn get_implied_distribution(f: impl Fn(u64) -> usize) -> (Vec<(u64, usize)>, Vec<f64>) {	×
25	let change_points = FindChangePoints::new(f)	×
26	.take_while(\|(_input, len)\| *len <= 128)	×
27	.collect::<Vec<_>>();
28
29	// convert to len probabilities
30	let probabilities = change_points	×
31	.windows(2)
32	.map(\|window\| {	×
33	let (input, len) = window[0];	×
34	let (next_input, _next_len) = window[1];	×
35	let prob = 2.0_f64.powi(-(len as i32));	×
36	prob * (next_input - input) as f64	×
37	})
38	.collect::<Vec<_>>();
39	// TODO: this ignores the last change point
40
41	(change_points, probabilities)	×
42	}
43
44	#[derive(Clone, Copy, Debug)]
45	/// An infinite iterator that always returns ().
46	pub struct InfiniteIterator;
47
48	impl Iterator for InfiniteIterator {
49	type Item = ();
50
51	#[inline(always)]
52	fn next(&mut self) -> Option<Self::Item> {	×
53	Some(())	×
54	}
55	}
56
57	/// Returns an infinite iterator of samples from the implied distribution of
58	/// the given code length function.
59	/// The function f should be the len function of the code.
60	///
61	/// This code works only with monotonic non-decreasing len functions and
62	/// the codes are limited to 128 bits as we cannot write more than 64 bits at once.
63	///
64	/// # Example
65	///
66	/// ```rust
67	/// use dsi_bitstream::utils::sample_implied_distribution;
68	/// use dsi_bitstream::codes::len_gamma;
69	/// use rand::SeedableRng;
70	/// use rand::rngs::SmallRng;
71	///
72	/// let mut rng = SmallRng::seed_from_u64(42);
73	/// let vals: Vec<u64> = sample_implied_distribution(len_gamma, &mut rng)
74	/// .take(1000).collect::<Vec<_>>();
75	///
76	/// assert_eq!(vals.len(), 1000);
77	/// ```
UNCOV 78	pub fn sample_implied_distribution(	×
79	f: impl Fn(u64) -> usize,
80	rng: &mut impl Rng,
81	) -> impl Iterator<Item = u64> + '_ {
82	let (change_points, probabilities) = get_implied_distribution(f);	×
83	let dist = WeightedIndex::new(probabilities).unwrap();	×
84
85	InfiniteIterator.map(move \|_\| {	×
86	// sample a len with the correct probability
87	let idx = dist.sample(rng);	×
88	// now we sample a random value with the sampled len
89	let (start_input, _len) = change_points[idx];	×
90	let (end_input, _len) = change_points[idx + 1];	×
91	rng.random_range(start_input..end_input)	×
92	})
93	}

vigna / dsi-bitstream-rs / 22241361593

Source File Press 'n' to go to next uncovered line, 'b' for previous

Source File
Press 'n' to go to next uncovered line, 'b' for previous