Xoshiro256++ by default
Random::new() seeds a 32-byte Xoshiro256++ generator from the operating system. Bounded integers are unbiased and floats use every mantissa bit.
How the default worksOpen-source Rust library
Random numbers for Rust, from microcontrollers to Monte Carlo
vrd generates random numbers with Xoshiro256++ by default, and adds Mersenne Twister, PCG, ChaCha20 and quasi-random sequences when you need them.
Platforms and targets
- Linux
- macOS
- Cortex-M (no_std)
- WebAssembly
- NEON and AVX2
What it does
One crate for the random numbers a program needs
Each card opens its section on the features page.
Five backends, one API
Mersenne Twister, PCG32, PCG64 and a ChaCha20 generator for secrets sit behind the same Random type.
Choose a backendReproducible sequences
The same seed gives the same sequence. Snapshot a generator with serde, or split off a stream 2128 draws ahead.
Reproduce a runDistributions
Uniform, normal (Ziggurat), exponential and Poisson built in, and a Distribution trait for your own.
See the distributionsQuasi-random sequences
Halton, Sobol and Van der Corput low-discrepancy sequences for Monte Carlo integration.
Sample evenlyno_std and WebAssembly
With default features off, the core builds for Cortex-M and wasm32 without an allocator, and CI checks both.
Where it runsBackends
Fast by default, secure when you ask
The default and the alternatives are the same Random type. Only ChaCha20 is suitable for secrets.
Default
Xoshiro256++
32 bytes of state. The default for simulations, games, sampling and tests. Not for secrets.
Compare the backendspcg feature
PCG32 and PCG64
16 and 32 bytes of state. PCG32 is the fastest backend in the published benchmarks.
See the timingscrypto feature
ChaCha20
A cryptographically secure generator from rand_chacha, for tokens, keys and anything an attacker would want to predict.
When you need itInstall
Random numbers in a few lines of Rust
Add the crate, create a generator and draw. vrd 0.0.12 needs Rust 1.85 or newer.
- cargo add vrd
- cargo install vrd
use vrd::Random;
fn main() {
// Xoshiro256++, seeded by the operating system
let mut rng = Random::new();
let roll = rng.int(1, 6); // i32 in [1, 6]
let x = rng.double(); // f64 in [0, 1)
let pick = rng.choose(&["a", "b", "c"]);
println!("{roll} {x} {pick:?}");
}Security
Predictable by design, unless you ask otherwise
The default generator is fast and statistically good, and anyone who sees enough of its output can predict the rest.
For session tokens, API keys, password-reset links and nonces, turn on the crypto feature and use Random::new_secure(). The security page sets out what vrd does and does not promise.
- Unsafe code
- SIMD module only
- Fuzz targets
- 6
- Secure backend
- ChaCha20
- Constant time
- Not promised
In numbers
What the crate ships
Counted from the vrd repository at version 0.0.14.
- Generator backends
- 5
- Built-in distributions
- 4
- Quasi-random sequences
- 3
- Runnable examples
- 43
- std (default)
- alloc
- serde
- simd
- pcg
- crypto
- quasirandom
- crush
What is vrd?
vrd is a free, open-source random number generator library for Rust. Its Random type gives you integers in a range, floats, booleans, choices from a slice, shuffles, tokens and samples from four distributions. The default generator is Xoshiro256++, seeded from the operating system; Cargo features add a Mersenne Twister, PCG32 and PCG64, a ChaCha20 generator for secrets, SIMD byte fills and quasi-random sequences. It is dual-licensed under Apache-2.0 or MIT.
Can I use it for passwords and tokens?
Only with the crypto feature and Random::new_secure(). The default generator, like the Mersenne Twister and PCG backends, is not cryptographically secure: anyone who sees enough of its output can predict the rest. The security page explains where the line is.
Does it need the standard library?
No. With default features off, the core builds without the standard library or an allocator, and CI checks it for a Cortex-M target and for WebAssembly. The features page shows how.
Who makes it?
vrd is written by Sebastien Rousseau and its contributors. The source, the issue tracker and every release are on GitHub; the crate is on crates.io. Read more on the about page.