136 lines
3.3 KiB
C
136 lines
3.3 KiB
C
|
/*
|
||
|
* Copyright (c) Meta Platforms, Inc. and affiliates.
|
||
|
*
|
||
|
* This source code is licensed under the MIT license found in the
|
||
|
* LICENSE file in the root directory of this source tree.
|
||
|
*/
|
||
|
|
||
|
#pragma once
|
||
|
|
||
|
#include <algorithm>
|
||
|
#include <random>
|
||
|
|
||
|
namespace facebook {
|
||
|
namespace react {
|
||
|
|
||
|
/*
|
||
|
* The source of pseudo-random numbers and some problem-oriented tools built on
|
||
|
* top of that. We need this class to maintain a reproducible stream of random
|
||
|
* numbers and abstract away complex math of and C++ STL API behind that.
|
||
|
*/
|
||
|
class Entropy final {
|
||
|
public:
|
||
|
using Generator = std::mt19937;
|
||
|
|
||
|
/*
|
||
|
* Creates an instance seeded with a real, not pseudo-random, number.
|
||
|
*/
|
||
|
Entropy() {
|
||
|
std::random_device device;
|
||
|
seed_ = device();
|
||
|
generator_ = std::mt19937(seed_);
|
||
|
}
|
||
|
|
||
|
/*
|
||
|
* Creates an instance seeded with a given number.
|
||
|
*/
|
||
|
Entropy(uint_fast32_t seed) {
|
||
|
seed_ = seed;
|
||
|
generator_ = std::mt19937(seed_);
|
||
|
}
|
||
|
|
||
|
uint_fast32_t getSeed() const {
|
||
|
return seed_;
|
||
|
}
|
||
|
|
||
|
/*
|
||
|
* Family of methods that return uniformly distributed instances of a type
|
||
|
* within a specified range.
|
||
|
*/
|
||
|
template <typename T>
|
||
|
bool random() const {
|
||
|
T result;
|
||
|
generateRandomValue(generator_, result);
|
||
|
return result;
|
||
|
}
|
||
|
|
||
|
template <typename T, typename Arg1>
|
||
|
T random(Arg1 arg1) const {
|
||
|
T result;
|
||
|
generateRandomValue(generator_, result, arg1);
|
||
|
return result;
|
||
|
}
|
||
|
|
||
|
template <typename T, typename Arg1, typename Arg2>
|
||
|
T random(Arg1 arg1, Arg2 arg2) const {
|
||
|
T result;
|
||
|
generateRandomValue(generator_, result, arg1, arg2);
|
||
|
return result;
|
||
|
}
|
||
|
|
||
|
void generateRandomValue(
|
||
|
Generator &generator,
|
||
|
bool &result,
|
||
|
double ratio = 0.5) const {
|
||
|
result = generator() % 10000 < 10000 * ratio;
|
||
|
}
|
||
|
|
||
|
void generateRandomValue(Generator &generator, int &result) const {
|
||
|
result = generator();
|
||
|
}
|
||
|
|
||
|
void generateRandomValue(Generator &generator, int &result, int min, int max)
|
||
|
const {
|
||
|
std::uniform_int_distribution<int> distribution(min, max);
|
||
|
result = distribution(generator);
|
||
|
}
|
||
|
|
||
|
/*
|
||
|
* Shuffles `std::vector` in place.
|
||
|
*/
|
||
|
template <typename T>
|
||
|
void shuffle(T array) const {
|
||
|
std::shuffle(array.begin(), array.end(), generator_);
|
||
|
}
|
||
|
|
||
|
/*
|
||
|
* Distribute items from a given array into buckets using a normal
|
||
|
* distribution and given `deviation`.
|
||
|
*/
|
||
|
template <typename T>
|
||
|
std::vector<std::vector<T>> distribute(std::vector<T> items, double deviation)
|
||
|
const {
|
||
|
std::normal_distribution<> distribution{0, deviation};
|
||
|
|
||
|
auto deviationLimit = int(deviation * 10);
|
||
|
auto spreadResult = std::vector<std::vector<T>>(deviationLimit * 2);
|
||
|
std::fill(spreadResult.begin(), spreadResult.end(), std::vector<T>{});
|
||
|
|
||
|
for (auto const &item : items) {
|
||
|
auto position = int(distribution(generator_) + deviationLimit);
|
||
|
position = std::max(0, std::min(position, deviationLimit * 2));
|
||
|
|
||
|
if (position < spreadResult.size()) {
|
||
|
spreadResult[position].push_back(item);
|
||
|
}
|
||
|
}
|
||
|
|
||
|
auto result = std::vector<std::vector<T>>{};
|
||
|
for (auto const &chunk : spreadResult) {
|
||
|
if (chunk.size() == 0) {
|
||
|
continue;
|
||
|
}
|
||
|
result.push_back(chunk);
|
||
|
}
|
||
|
|
||
|
return result;
|
||
|
}
|
||
|
|
||
|
private:
|
||
|
mutable std::mt19937 generator_;
|
||
|
uint_fast32_t seed_;
|
||
|
};
|
||
|
|
||
|
} // namespace react
|
||
|
} // namespace facebook
|