I need to generate a sequence of random (pseudorandom do suit) strings in such a way that the probability of two strings in the sequence being the same is negligible.
So, how to obtain a sequence of pseudorandoms?
Internet Computer Loading tells something sophisticated about entropy. But I don’t need entropy, because pseudorandomness is enough for my purpose.
Check out the UUID and ULID libraries: GitHub - aviate-labs/ulid.mo: ULID Generator for Motoko
Perhaps they meet your requirements? I think they have a pseudorandom path.
It is unclear what ULID.MonotonicEntropy does with entropy and how to run it to obtain a pseudo-random sequence.
We have a Prng library here: motoko-lib/Prng.mo at main · research-ag/motoko-lib · GitHub
It provides SFC64 which is a known generator that is also part of numpy. The output is compatible with numpy.
It produces a sequence of Nat64 and you have to convert that to a string yourself. If it is a common use case then we could add that to the library (output as Text or Blob) for convenience.
We published the Prng library on Mops now: https://mops.one/prng
It would be nice to add convenience functions to generate random values of other types than only Nat64. Hence the question what was the original requirement for the pseudo-random strings? Where they Blob or Text and what was the required character set?
Your Prng type is not usable for my purposes, because it is not a stable type.
It is Text in any character set.
You wish to persist the Prng’s internal state across updates? Or why does the generator itself have to be stable?
Does that mean you don’t care what the character set is or that you would like to be able to supply an arbitrary character set to the generator?
Yes. I want to persist Prng’s internal state across updates.
I want to store results of the generator in Text DB keys in such a way that lexicographical order of keys matches order of the random numbers. The charset does not matter (except of, maybe, for debugging is better to use UTF-8 or even ASCII).