rand

Random number generation toolkit for Ada

Design principles mostly inspired by the rand Rust crate.

The project is split into several subcrates. rand_core, rand_chacha, rand_xoshiro256 and rand_distributions can be used on embedded; However, rand_sys (and its dependent rand, the main crate) pull entropy from system sources, using the system_random Alire crate.

Usage examples

Get a thread local instance of a secure Random Number Generator (RNG), seeded with system entropy:

with Rand;
R : Rand.Rng := Rand.Thread_Rng;
--  alternatively:
R : Rand.Rng := Rand.Small_Rng;
   --  a fast, unsecure RNG seeded with system entropy
R : Rand.Rng := Rand.Sys.Get;
   --  RNG based on system randomness sources
   --  (OS-dependent)

Use convenience methods on the RNG to generate basic types:

V1 : Float := R.Gen;        --  a float in the range [0, 1)
V2 : Long_Integer := R.Gen; --  a long integer over the whole range

You can also define your own random number generators by implementing the Rand.Core_Rng interface (alias for Rand_Core.Rng).

Use Next and Next_Bytes to get the raw output of any RNG:

Buf : Rand.Core.Bytes (1 .. 256);
R.Next_Bytes (Buf);
X : Rand.Core.U64 := R.Next;

Use a predefined random distribution to get finer random value selection:

use Rand.Distributions;
D1 : Uniform_Nat.Distribution := Uniform_Nat.Create (8, 27);
S : Natural := D1.Sample (R);
   --  sample in the inclusive range [8, 27]

D2 : Bernoulli := Bernoulli.Create (0.25);
S : Boolean := D2.Sample (R);
   --  a boolean that is True 25% of the time

Or define your own distributions:

use Rand.Distributions;

type Gaussian is new Long_Float_Distr.Distribution with record
   Sigma : Long_Float;
end record;

overriding
function Sample (D : Gaussian; R : in out Rand.Rng) return Long_Float
is (...); --  sampling the custom distribution

Even on your own types:

with Rand_Distributions; use Rand_Distributions;

type My_Rec is record
   A : Integer;
   B : Float;
end record;

package I is new Generic_Distribution (My_Rec);
--  define the interface distributions over your type must implement

type My_Distr is new I.Distribution with null record;

overriding
function Sample (D : My_Distr; R : in out Rand.Rng) return My_Rec
is (A => 4, --  chosen by fair dice roll
    B => R.Gen);