See the Make_random functor in the janestreet Bigint library, which is built on top of zarith. (It's a functor to produce random distributions based on both Random.State.t and Base_quickcheck.Generator.t; the functor itself is not exposed.)

https://github.com/janestreet/bignum/blob/master/bigint/src/bigint.ml

In short, we generate 30-bit chunks of randomness until we have at least enough bits for our range. We combine those into a number. Usually, we just modulo that by the range and return it. But to preserve fairness, we first have to check if the number is in the last fraction-of-range part of the N bits, and if so retry from scratch. The odds of retry are always less than 50%, so retrying is never too bad.

This is the same trick that Random.int does, but with an unbounded number of bits instead of a fixed number of bits.

On Wed, Nov 27, 2019 at 4:31 PM François Pottier <francois.pottier@inria.fr> wrote:

Hello,

I am using zarith and would like to pick a random integer
comprised between 0 and some bound. I would like a function
Z.random of type Z.t -> Z.t, but this function seems to be
missing, and I am not sure how to program it in an efficient
and correct way. Any suggestions would be welcome. Thanks!

--
François Pottier
francois.pottier@inria.fr
http://gallium.inria.fr/~fpottier/


--
Carl Eastlund