commit 7bcb018244cffc2ea4e7abde3447b578908facfb
Author: Maximilian Keßler <git@maximilian-kessler.de>
Date:   Fri Aug 4 16:28:41 2023 +0200

    initial commit of game state

diff --git a/game_state.h b/game_state.h
new file mode 100644
index 0000000..a306624
--- /dev/null
+++ b/game_state.h
@@ -0,0 +1,198 @@
+//
+// Created by maximilian on 7/13/23.
+//
+
+#ifndef DYNAMIC_PROGRAM_GAME_STATE_H
+#define DYNAMIC_PROGRAM_GAME_STATE_H
+
+#include <array>
+#include <cstdint>
+#include <algorithm>
+#include <cstddef>
+#include <bitset>
+#include <limits>
+#include <optional>
+#include <boost/container/static_vector.hpp>
+#include <list>
+#include <ostream>
+
+
+using rank_t = std::uint8_t;
+using suit_t = std::uint8_t;
+using clue_t = std::uint8_t;
+using player_t = std::int8_t;
+
+using state_t = std::uint32_t;
+
+
+/**
+ * We will generally assume that stacks are played from n to 0
+ * Playing a 0 will yield a clue
+ * Therefore, for the default hanabi, we will play 4,3,2,1,0 in that order
+ * on each stack. A stack with no cards played implicitly has value 5 on it
+ * This is just easier to implement, since then the remaining number of cards
+ * to be played is always the current number of the stack
+ */
+constexpr rank_t starting_card_rank = 5;
+constexpr suit_t max_suit_index = 5;
+constexpr size_t max_card_duplicity = 3;
+constexpr player_t draw_pile = -1;
+constexpr player_t trash_or_play_stack = -2;
+constexpr clue_t max_num_clues = 8;
+
+constexpr std::array<std::string, 5> suit_initials {"r", "y", "g", "b", "p"};
+
+struct Card {
+  suit_t suit;
+  rank_t rank;
+  uint8_t copy;
+
+  Card& operator++() {
+    rank++;
+    return *this;
+  }
+
+  Card successor() {
+    return {suit, static_cast<rank_t>(rank + 1)};
+  }
+
+  const Card operator++(int) {
+    Card ret = *this;
+    rank++;
+    return ret;
+  }
+
+  auto operator<=>(const Card&) const = default;
+};
+
+std::ostream& operator<<(std::ostream& os, const Card& card) {
+  os << suit_initials[card.suit] << +card.rank;
+  return os;
+}
+
+constexpr Card r0 = {0, 0, 0};
+constexpr Card r1 = {0, 1, 0};
+constexpr Card r2 = {0, 2, 0};
+constexpr Card r3 = {0, 3, 0};
+constexpr Card r4 = {0, 4, 0};
+constexpr Card y0 = {1, 0, 0};
+constexpr Card y1 = {1, 1, 0};
+constexpr Card y2 = {1, 2, 0};
+constexpr Card y3 = {1, 3, 0};
+constexpr Card y4 = {1, 4, 0};
+
+/**
+ * To store:
+ * - Draw pile size
+ * - Distribution of cards
+ *      - Which cards exist?
+ * - Number of clues
+ */
+
+template<std::size_t num_suits>
+using Stacks = std::array<rank_t, num_suits>;
+
+template<std::size_t num_suits>
+std::ostream& operator<<(std::ostream& os, const Stacks<num_suits>& stacks) {
+  for (size_t i = 0; i < stacks.size() - 1; i++) {
+    os << +stacks[i] << ", ";
+  }
+  os << +stacks.back();
+  return os;
+}
+
+template <std::size_t num_suits>
+struct CardSetSpecification {
+  using AdditionalCardsFlags = std::bitset<num_suits * (starting_card_rank - 1)>;
+
+  Stacks<num_suits> stacks {};
+  AdditionalCardsFlags additional_cards_flags {};
+};
+
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size>
+struct PartialHanabiInstance {
+  std::array<boost::container::static_vector<Card, hand_size>, num_players> initial_cards;
+  CardSetSpecification<num_suits> initial_state;
+};
+
+struct CardMultiplicity {
+  Card card;
+  std::uint8_t multiplicity;
+
+  auto operator<=>(const CardMultiplicity&) const = default;
+};
+
+template<std::size_t num_suits>
+struct CardPositions {
+
+  const player_t & operator[](const Card& card) const {
+      return _card_positions[card.suit][card.rank][card.copy];
+  };
+
+  player_t & operator[](const Card& card) {
+    return _card_positions[card.suit][card.rank][card.copy];
+  };
+
+  auto operator<=>(const CardPositions&) const = default;
+
+private:
+  std::array<std::array<std::array<player_t, max_card_duplicity>, starting_card_rank>, num_suits> _card_positions;
+};
+
+enum class ActionType {
+  discard,
+  clue,
+  play
+};
+
+struct Action {
+  ActionType type {};
+  Card discarded {};
+  std::uint8_t index {};
+};
+
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, std::uint8_t max_draw_pile_size>
+class HanabiState {
+public:
+
+  Action clue();
+  /**
+   * Plays a card from current hand, drawing top card of draw pile and rotating draw pile
+   * @param index of card in hand to be played
+   */
+  Action play(std::uint8_t index);
+
+  Action discard(std::uint8_t index);
+
+  void revert(const Action& action);
+
+
+  void draw(std::uint8_t index);
+  void revert_draw(std::uint8_t index, Card card);
+  void incr_turn();
+  void decr_turn();
+
+  player_t _turn{};
+  clue_t _num_clues{};
+  std::uint8_t _draw_pile_size{};
+  Stacks<num_suits> _stacks {};
+  std::array<boost::container::static_vector<Card, hand_size>, num_players> _hands {};
+  CardPositions<num_suits> _card_positions {};
+  std::list<CardMultiplicity> _draw_pile {};
+
+  // further statistics that we might want to keep track of
+  uint8_t _pace{};
+
+  auto operator<=>(const HanabiState&) const = default;
+};
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, std::uint8_t max_draw_pile_size>
+std::ostream& operator<<(std::ostream& os, const HanabiState<num_suits, num_players, hand_size, max_draw_pile_size> hanabi_state);
+
+#include "game_state.hpp"
+
+template class HanabiState<5, 3, 4, 20>;
+
+#endif // DYNAMIC_PROGRAM_GAME_STATE_H
diff --git a/game_state.hpp b/game_state.hpp
new file mode 100644
index 0000000..e7b2196
--- /dev/null
+++ b/game_state.hpp
@@ -0,0 +1,143 @@
+#include <cassert>
+#include <algorithm>
+#include <iterator>
+
+
+template <size_t num_suits, size_t num_players, size_t hand_size, uint8_t max_draw_pile_size>
+Action HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::clue() {
+  assert(_num_clues > 0);
+  --_num_clues;
+
+  incr_turn();
+
+  return Action {ActionType::clue, {}, {}};
+}
+
+template <size_t num_suits, size_t num_players, size_t hand_size, uint8_t max_draw_pile_size>
+void HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::incr_turn() {
+  _turn = (_turn + 1) % num_players;
+}
+
+template <size_t num_suits, size_t num_players, size_t hand_size, uint8_t max_draw_pile_size>
+void HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::decr_turn() {
+  _turn = (_turn + num_players - 1) % num_players;
+}
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, uint8_t max_draw_pile_size>
+Action HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::play(std::uint8_t index) {
+  assert(index < _hands[_turn].size());
+  const Card card = _hands[_turn][index];
+  assert(card.rank == _stacks[card.suit] - 1);
+
+  --_stacks[card.suit];
+
+  Action ret {ActionType::play, _hands[_turn][index], index};
+
+  if (card.rank == 0) {
+    // update clues if we played the last card of a stack
+    _num_clues++;
+  }
+
+  draw(index);
+  incr_turn();
+
+  return ret;
+}
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, uint8_t max_draw_pile_size>
+Action HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::discard(std::uint8_t index) {
+  assert(index < _hands[_turn].size());
+  assert(_num_clues != max_num_clues);
+
+  _num_clues++;
+
+  Action ret {ActionType::discard, _hands[_turn][index], index};
+
+  draw(index);
+  incr_turn();
+
+  return ret;
+}
+
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, uint8_t max_draw_pile_size>
+std::ostream& operator<<(std::ostream& os, const HanabiState<num_suits, num_players, hand_size, max_draw_pile_size> hanabi_state) {
+  os << "Stacks: " << hanabi_state._stacks << std::endl;
+  os << "Draw pile: ";
+  for (const auto &[card, mul] : hanabi_state._draw_pile) {
+      os << card;
+      if (mul > 1) {
+          os << " (" << +mul << ")";
+      }
+  }
+  os << std::endl;
+  os << "Hands: ";
+  for (const auto& hand: hanabi_state._hands) {
+      for (const auto &card: hand) {
+          os << card << ", ";
+      }
+      os << " | ";
+  }
+  return os;
+}
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, uint8_t max_draw_pile_size>
+void HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::draw(std::uint8_t index) {
+  assert(index < _hands[_turn].size());
+
+  _card_positions[_hands[_turn][index]] = trash_or_play_stack;
+
+  // draw a new card if the draw pile is not empty
+  if (!_draw_pile.empty()) {
+    --_draw_pile_size;
+    CardMultiplicity draw = _draw_pile.front();
+    _draw_pile.pop_front();
+    assert(draw.multiplicity > 0);
+    if (draw.multiplicity > 1) {
+      draw.multiplicity--;
+      _draw_pile.push_back(draw);
+    }
+    _hands[_turn][index] = draw.card;
+    _card_positions[draw.card] = _turn;
+  }
+}
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, uint8_t max_draw_pile_size>
+void HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::revert_draw(std::uint8_t index, Card card) {
+  assert(index < _hands[_turn].size());
+
+  _card_positions[_hands[_turn][index]] = draw_pile;
+
+  // draw a new card if the draw pile is not empty
+  if (_draw_pile.back().card == _hands[_turn][index]) {
+    _draw_pile.back().multiplicity++;
+  } else {
+    _draw_pile.push_back({_hands[_turn][index], 1});
+  }
+
+  _hands[_turn][index] = card;
+  _card_positions[card] = _turn;
+  _draw_pile_size++;
+}
+
+template <std::size_t num_suits, std::size_t num_players, std::size_t hand_size, uint8_t max_draw_pile_size>
+void HanabiState<num_suits, num_players, hand_size, max_draw_pile_size>::revert(const Action &action) {
+  decr_turn();
+  switch (action.type) {
+    case ActionType::clue:
+      assert(_num_clues < max_num_clues);
+      _num_clues++;
+      break;
+    case ActionType::discard:
+      assert(_num_clues > 0);
+      _num_clues--;
+      revert_draw(action.index, action.discarded);
+      break;
+    case ActionType::play:
+      if (action.discarded.rank == 0) {
+        _num_clues--;
+      }
+      revert_draw(action.index, action.discarded);
+      _stacks[action.discarded.suit]++;
+  }
+}
diff --git a/main.cpp b/main.cpp
new file mode 100644
index 0000000..e31ebee
--- /dev/null
+++ b/main.cpp
@@ -0,0 +1,55 @@
+//
+// Created by maximilian on 7/13/23.
+//
+
+#include <boost/rational.hpp>
+#include <cstdint>
+#include <iostream>
+#include <optional>
+#include <vector>
+
+#include "game_state.h"
+
+
+void test_game() {
+  HanabiState<2, 2, 5, 10> state;
+  state._stacks[0] = 2;
+  state._stacks[1] = 3;
+  Card r41 = {0,4,1};
+  state._draw_pile.push_back({r41, 1});
+  state._hands[0] = {y0, y1, y2, r0, r1};
+  state._hands[1] = {r1, r1, y1, r3, r2};
+  state._draw_pile_size = 1;
+  state._card_positions[r41] = draw_pile;
+
+  auto state2 = state;
+
+  auto a = state.play(4);
+  std::cout << state;
+  state.revert(a);
+
+  std::cout << state << std::endl;
+  std::cout << state2 << std::endl;
+  assert(state._hands == state2._hands);
+  assert(state._draw_pile == state2._draw_pile);
+  assert(state._card_positions == state2._card_positions);
+  assert(state == state2);
+}
+
+void print_sizes() {
+  std::cout << "size of card -> hand map: " << sizeof(HanabiState<5,3,4, 5>) << std::endl;
+
+  CardSetSpecification<5> test;
+  std::cout << sizeof(CardSetSpecification<8>::AdditionalCardsFlags) << std::endl;
+
+  unsigned exp = 32;
+  std::cout << "Pair size: " << sizeof(std::pair<std::uint32_t, float>) << std::endl;
+  std::cout << sizeof(boost::rational<int>) << std::endl;
+  std::cout << (1ul << exp) << std::endl;
+}
+
+int main() {
+  test_game();
+
+  return 0;
+}
\ No newline at end of file