It would be nice if MonadRandom did not depend on System.Random

[jeffreyrosenbluth]

In particular if the IO instance was not tied to randomIO and stdGen.
So that we could plug in other prng's like tf-random for things like evalRandIO.

[tom-bop]

Does this mean there's currently no way to use MonadRandom and avoid the wonky behavior of http://kenta.blogspot.co.nz/2014/02/rzzllapd-nonrandomness-of-first-random.html ?

[tom-bop]

Hm, appears we can at least work around it(?). Sorry for the noise if so.

>  :m + Control.Monad.Random
>  :m + System.Random.TF
>  (fst . randomR (0::Int, 7)) <$> newTFGen
5

[lehins]

All these issues has been fixed in random-1.2.

IO instance for MonadRandom is a limitation of design IMHO, which has also been solved with StatefulGen in random-1.2