tuneup sampling in C++(with classic functions)

[azriel1rf]

continue from #56
I cannot reopen and opened another PR.

More fast implementation is available

Unit comparison

result

310 ms
29 20 2 1 25 32 4 3 46 
36 8 24 29 15 7 30 34 28 
0 34 40 23 1 18 5 7 32 

proposal
72 ms
30 15 40 16 18 14 51 36 9 
3 24 14 17 35 42 34 39 41 
9 22 41 5 3 48 17 14 42

roughly implementation

class CardSampler {
  std::array<int, 52> deck;
public:
  CardSampler(void) {
    for (int i = 0; i < deck.size(); i++) {
      deck[i] = i;
    }
  }
  std::vector<int> sample(int size) {
    std::vector<int> ret;
    int residual_cards = deck.size();
    for (int i = 0; i < size; i++) {
      int target_index = generator() % residual_cards;
      int tail_index = residual_cards - 1;
      std::swap(deck[target_index], deck[tail_index]);
      ret.push_back(deck[tail_index]);
      residual_cards--;
    }
    return ret;
  }
};

comparison source code

#include <algorithm>
#include <array>
#include <iostream>
#include <map>
#include <random>
#include <string>
#include <type_traits>
#include <vector>
#include <utility>
#include <ctime>
#include <chrono>
#include <set>
#include <unordered_set>
#include <numeric>

using std::array;
using std::cout;
using std::endl;
using std::map;
using std::string;
using std::unordered_map;
using std::vector;

void print(auto const& a) {
  for (auto const& e : a) cout << e << ' ';
  cout << "\n";
}

static unsigned seed = std::chrono::system_clock::now().time_since_epoch().count();
static std::default_random_engine generator(seed);
static std::uniform_int_distribution<int> distribution(0, 51);

static std::vector<int> RandomCardSample(int size) {
  std::vector<int> ret;
  std::set<int> sample;

  while (sample.size() < size) {
    int number = distribution(generator);
    if (sample.find(number) == sample.end()) {
      ret.push_back(number);
      sample.insert(number);
    }
  }

  return ret;
}

class CardSampler {
  std::array<int, 52> deck;
public:
  CardSampler(void) {
    for (int i = 0; i < deck.size(); i++) {
      deck[i] = i;
    }
  }
  std::vector<int> sample(int size) {
    std::vector<int> ret;
    int residual_cards = deck.size();
    for (int i = 0; i < size; i++) {
      int target_index = generator() % residual_cards;
      int tail_index = residual_cards - 1;
      std::swap(deck[target_index], deck[tail_index]);
      ret.push_back(deck[tail_index]);
      residual_cards--;
    }
    return ret;
  }
};



int main() {
  int repetition = 100000;
  cout << "current" << endl;
  std::clock_t begin = std::clock();
  for (int i = 0; i < repetition; i++) {
    RandomCardSample(9);
  }
  cout << (clock() - begin) * 1000 / CLOCKS_PER_SEC << " ms" << endl;
  print(RandomCardSample(9));
  print(RandomCardSample(9));
  print(RandomCardSample(9));
  cout << endl;

  cout << "proposal" << endl;
  CardSampler cs{};
  begin = std::clock();
  for (int i = 0; i < repetition; i++) {
    cs.sample(9);
  }
  cout << (clock() - begin) * 1000 / CLOCKS_PER_SEC << " ms" << endl;
  print(cs.sample(9));
  print(cs.sample(9));
  print(cs.sample(9));
  cout << endl;
}

I thought there are better way without class just only function but I cannot create 0~51 array in compiling time.

[azriel1rf]

I referenced Fisher–Yates shuffle and its modern algorithm.
https://en.wikipedia.org/wiki/Fisher%E2%80%93Yates_shuffle#The_modern_algorithm

[HenryRLee]

Approved. Just left a comment for the std::iota. You can merge the PR with squash and merge if you feel it's ready. Thanks for your work!

[azriel1rf]

I cannot merge PR after pushing another commit.

[HenryRLee]

Cool. Thank you.