Pseudo-random number generator for EOS.IO via a single C++14 header file.

Since <random> is not available in EOS.IO, and they don't have good randomization primitives, we propose a simple pseudo-random number generator based on xoroshiro128+ for generation and splitmix64 for seed expansion.

• Usage
  • Seeding the generator
  • Generating values
  • Shuffling containers
  • Sampling from a population
• Testing

To access the generator, all that is needed is to copy "Random.h" into the source folder and include it in the smart contract.

The pseudo-random number generator operates on a 64-bit seed, which must be supplied manually for best randomness. It could be via a random number or cryptographic hash value retrieved via smart contract parameters.

If the constructor Random::Random() is used then the seed will be accumulated from the EOS.IO Transactions as Proof-of-Stake (TAPOS) block number and prefix in order to provide randomness of some kind. However, it is not recommended to rely solely on this!

Examples:

Random gen; // Seeded with TAPOS block number and prefix.
gen.accumSeed(42); // Provide extra seed info via number 42.

// Seed using EOS.IO types.
checksum256 hash; // Assume already constructed.
gem.accumSeed(hash);

// Seed using initializer lists and general containers.
gem.accumSeedRange({1, 2, 3});
gem.accumSeedRange(std::string("Hello, World!"));
gem.accumSeedRange(std::vector<int>{1, 2, 3});

// Seed using arrays.
const int arr[3]{1, 2, 3};
gen.accumSeedArray(arr);

The following EOS.IO types are directly supported via Random::accumSeed():

• checksum160
• checksum256
• checksum512
• signature
• public_key

They are defined in <eosiolib/types.h>.

Once the generator has been seeded, there are four ways of generating random values:

• uint64_t Random::next()
• double Random::nextDouble()
• uint64_t Random::nextInRange(const uint64_t min, const uint64_t max)
• T Random::nextSample(const Container &population)

Examples:

Random gen;
gen.next(); // Value in [0, 18446744073709551615]
gen.nextDouble(); // Value in [0.0, 1.0[
gen.nextInRange(20, 50); // Value in [20, 50[
gen.nextSample(std::string("wasd")); // Value in ['w', 'a', 's', 'd']

General containers can be shuffled by using Random::shuffle():

Random gen;
std::string str("Hello, World!");
gen.shuffle(str);
// str has been reordered at this point.

Every element of the container is shuffled exactly once.

General containers can be used for sampling a number of values from:

Random gen;
const std::string pop("abcdef");
const sample = gen.sample(10, pop);
// sample contains 10 randomly selected values from pop at this point.

A simple test suite can be compiled and executed like this:

% make tests && ./tests

If a test fails it might look like the following:

% ./tests
Assertion failed in file tests.cc, line 120, func testNext():
  expected 849456367364724273
  got      7364724273
make: *** [test] Abort trap: 6