結果
問題 | No.2428 Returning Shuffle |
ユーザー | Gandalfr |
提出日時 | 2023-08-18 22:37:18 |
言語 | C++17 (gcc 12.3.0 + boost 1.83.0) |
結果 |
WA
|
実行時間 | - |
コード長 | 36,771 bytes |
コンパイル時間 | 2,681 ms |
コンパイル使用メモリ | 222,428 KB |
実行使用メモリ | 141,988 KB |
最終ジャッジ日時 | 2024-05-06 04:58:18 |
合計ジャッジ時間 | 10,720 ms |
ジャッジサーバーID (参考情報) |
judge1 / judge3 |
(要ログイン)
テストケース
テストケース表示入力 | 結果 | 実行時間 実行使用メモリ |
---|---|---|
testcase_00 | WA | - |
testcase_01 | WA | - |
testcase_02 | WA | - |
testcase_03 | AC | 2 ms
5,376 KB |
testcase_04 | AC | 2 ms
5,376 KB |
testcase_05 | AC | 2 ms
5,376 KB |
testcase_06 | WA | - |
testcase_07 | AC | 1 ms
5,376 KB |
testcase_08 | AC | 2 ms
5,376 KB |
testcase_09 | WA | - |
testcase_10 | AC | 2 ms
5,376 KB |
testcase_11 | WA | - |
testcase_12 | WA | - |
testcase_13 | WA | - |
testcase_14 | AC | 2 ms
5,376 KB |
testcase_15 | AC | 2 ms
5,376 KB |
testcase_16 | AC | 2 ms
5,376 KB |
testcase_17 | AC | 1 ms
5,376 KB |
testcase_18 | AC | 2 ms
5,376 KB |
testcase_19 | WA | - |
testcase_20 | WA | - |
testcase_21 | AC | 2 ms
5,376 KB |
testcase_22 | AC | 2 ms
5,376 KB |
testcase_23 | AC | 2 ms
5,376 KB |
testcase_24 | AC | 1,021 ms
135,504 KB |
testcase_25 | AC | 1,009 ms
135,488 KB |
ソースコード
#line 1 "playspace/main.cpp" #include <bits/stdc++.h> #line 8 "library/gandalfr/other/io_supporter.hpp" template <typename T> std::ostream &operator<<(std::ostream &os, const std::vector<T> &v) { for (int i = 0; i < (int)v.size(); i++) os << v[i] << (i + 1 != (int)v.size() ? " " : ""); return os; } template <typename T> std::ostream &operator<<(std::ostream &os, const std::set<T> &st) { for (const T &x : st) { std::cout << x << " "; } return os; } template <typename T> std::ostream &operator<<(std::ostream &os, const std::multiset<T> &st) { for (const T &x : st) { std::cout << x << " "; } return os; } template <typename T> std::ostream &operator<<(std::ostream &os, const std::deque<T> &dq) { for (const T &x : dq) { std::cout << x << " "; } return os; } template <typename T1, typename T2> std::ostream &operator<<(std::ostream &os, const std::pair<T1, T2> &p) { os << p.first << ' ' << p.second; return os; } template <typename T> std::ostream &operator<<(std::ostream &os, std::queue<T> &q) { int sz = q.size(); while (--sz) { os << q.front() << ' '; q.push(q.front()); q.pop(); } os << q.front(); q.push(q.front()); q.pop(); return os; } template <typename T> std::istream &operator>>(std::istream &is, std::vector<T> &v) { for (T &in : v) is >> in; return is; } template <typename T1, typename T2> std::istream &operator>>(std::istream &is, std::pair<T1, T2> &p) { is >> p.first >> p.second; return is; } #line 8 "library/gandalfr/graph/graph.hpp" #line 3 "library/gandalfr/data_structure/union_find.hpp" #line 6 "library/gandalfr/data_structure/union_find.hpp" class union_find { private: int N; mutable std::vector<int> par; std::vector<int> nxt; int group_num; // 集合の数 public: union_find() : N(0), group_num(0) {} union_find(int n) : N(n), par(n, -1), nxt(n), group_num(n) { std::iota(nxt.begin(), nxt.end(), 0); } /** * @brief 頂点を n 個に増やす * @attention 小さくはできない */ void expand(int n) { if (n <= N) return; par.resize(n, -1); nxt.resize(n); for (int i = N; i < n; ++i) nxt[i] = i; group_num += n - N; N = n; } int leader(int x) const { return (par[x] < 0 ? x : par[x] = leader(par[x])); } bool same(int x, int y) const { return leader(x) == leader(y); } bool merge(int x, int y) { if ((x = leader(x)) == (y = leader(y))) return false; if (-par[x] > -par[y]) std::swap(x, y); par[x] += par[y]; par[y] = x; std::swap(nxt[x], nxt[y]); group_num--; return true; } /** * @brief x の属するグループのサイズを返す */ int size(int x) const { return -par[leader(x)]; } /** * @brief すべてのノードの数 */ int size() const { return N; } std::vector<int> contained_group(int x) const { std::vector<int> ret{x}; for (int cu = nxt[x]; cu != ret[0]; cu = nxt[cu]) ret.push_back(cu); return ret; } int count_groups() const { return group_num; } std::vector<std::vector<int>> all_groups() const { std::vector<std::vector<int>> result; result.reserve(group_num); std::vector<bool> used(N, false); for (int i = 0; i < N; ++i) { if (!used[i]) { result.emplace_back(contained_group(i)); for (int x : result.back()) { used[x] = true; } } } return result; } }; #line 3 "library/gandalfr/math/matrix.hpp" #line 8 "library/gandalfr/math/matrix.hpp" template <class T> class matrix { private: int H, W; std::valarray<std::valarray<T>> table; enum rowtrans_operation_name { SCALE, SWAP, ADD }; struct rowtrans_operation { int op, tar, res; T scl; }; using operations_history = std::vector<rowtrans_operation>; public: matrix() = default; matrix(int _H, int _W, T val = 0) : H(_H), W(_W), table(std::valarray<T>(val, _W), _H) {} matrix(const std::vector<std::vector<T>> &vv) : H(vv.size()), W(vv[0].size()), table(std::valarray<T>(W), H) { for (int i = 0; i < H; i++) for (int j = 0; j < W; j++) table[i][j] = vv[i][j]; } matrix(const std::valarray<std::valarray<T>> &vv) : H(vv.size()), W(vv[0].size()), table(vv) {} /** * @brief 行列をリサイズする。 * @param val 拡張部分の値 */ void resize(int _H, int _W, T val = 0) { H = _H, W = _W; table.resize(_H, std::valarray<T>(val, _H)); } int size_H() const { return H; } int size_W() const { return W; } void transpose() { matrix<T> ret(W, H); for (int i = 0; i < H; i++) for (int j = 0; j < W; j++) ret.table[j][i] = table[i][j]; *this = std::move(ret); } void row_assign(int i, const std::valarray<T> &row) { assert(W == (int)row.size()); table[i] = std::move(row); } void row_swap(int i, int j) { assert(0 <= i && i < H); assert(0 <= j && j < H); table[i].swap(table[j]); } /** * @attention O(n^3) * @attention 整数型では正しく計算できない。double や fraction を使うこと。 * @attention 枢軸選びをしていないので double では誤差が出るかも。 */ operations_history sweep_method() { operations_history hist; T ret = 1; for (int h = 0, w = 0; h < H && w < W; w++) { if (table[h][w] == 0) { for (int piv = h + 1; piv < H; piv++) { if (table[piv][w] != 0) { hist.push_back({SWAP, h, piv, 0}); row_swap(h, piv); break; } } if (table[h][w] == 0) { continue; } } T inv = 1 / table[h][w]; hist.push_back({SCALE, -1, w, inv}); table[h] *= inv; for (int j = h + 1; j < H; j++) { hist.push_back({ADD, h, j, -table[j][w]}); table[j] -= table[h] * table[j][w]; } h++; } return hist; } int rank() { auto U(*this); U.sweep_method(); int r = 0; for (int i = 0; i < H; ++i) { for (int j = i; j < W; ++j) { if (U.table[i][j] != 0) { r++; break; } } } return r; } T determinant() const { assert(H == W); matrix<T> U(*this); T det = 1; auto hist = U.sweep_method(); if (U.table[H - 1][H - 1] == 0) return 0; for (auto &[op, tar, res, scl] : hist) { switch (op) { case SCALE: det /= scl; break; case SWAP: det *= -1; break; } } return det; } std::vector<T> solve_system_of_equations(const std::vector<T> &y) { assert(H == W); std::vector<T> x(y); matrix<T> U(*this); auto hist = U.sweep_method(); if (U.table[H - 1][H - 1] == 0) return {}; for (auto &[op, tar, res, scl] : hist) { switch (op) { case SCALE: x[res] *= scl; break; case SWAP: std::swap(x[tar], x[res]); break; case ADD: x[res] += x[tar] * scl; break; } } for (int i = H - 1; i >= 0; --i) { for (int j = 0; j < i; ++j) { x[j] -= U.table[j][i] * x[i]; } } return x; } matrix<T> inverse() { assert(H == W); matrix<T> INV(matrix<T>::E(H)), U(*this); auto hist = U.sweep_method(); if (U.table[H - 1][H - 1] == 0) return matrix<T>(0, 0); for (auto &[op, tar, res, scl] : hist) { switch (op) { case SCALE: INV.table[res] *= scl; break; case SWAP: std::swap(INV.table[tar], INV.table[res]); break; case ADD: INV.table[res] += INV.table[tar] * scl; break; } } for (int i = H - 1; i >= 0; --i) { for (int j = 0; j < i; ++j) { INV.table[j] -= INV.table[i] * U.table[j][i]; } } return INV; } void print() const { for (int i = 0; i < H; i++) { for (int j = 0; j < W; j++) { std::cout << table[i][j] << (j == W - 1 ? "" : " "); } std::cout << std::endl; } } matrix<T> &operator+=(const matrix<T> &a) { this->table += a.table; return *this; } matrix<T> &operator-=(const matrix<T> &a) { this->table -= a.table; return *this; } matrix<T> &operator*=(const T &a) { this->table *= a; return *this; } matrix<T> &operator*=(const matrix<T> &a) { assert(W == a.H); matrix<T> a_t(a), ret(H, a.W); a_t.transpose(); for (int i = 0; i < H; i++) { for (int j = 0; j < a_t.H; j++) { ret.table[i][j] = (table[i] * a_t.table[j]).sum(); } } *this = std::move(ret); return *this; } matrix<T> &operator/=(const T &a) { this->table /= a; return *this; } /** * @brief 行列の冪乗。 * @param n 指数 * @attention n が 0 なら単位行列。 * @attention 演算子の優先度に注意。 */ matrix<T> operator^=(long long n) { assert(H == W); if (n == 0) return *this = E(H); n--; matrix<T> x(*this); while (n) { if (n & 1) *this *= x; x *= x; n >>= 1; } return *this; } matrix<T> operator+() { return *this; } matrix<T> operator-() { return matrix<T>(*this) *= -1; } matrix<T> operator+(const matrix<T> &a) { return matrix<T>(*this) += a; } matrix<T> operator-(const matrix<T> &a) { return matrix<T>(*this) -= a; } template <typename S> matrix<T> operator*(const S &a) { return matrix<T>(*this) *= a; } matrix<T> operator/(const T &a) { return matrix<T>(*this) /= a; } matrix<T> operator^(long long n) { return matrix<T>(*this) ^= n; } friend std::istream &operator>>(std::istream &is, matrix<T> &mt) { for (auto &arr : mt.table) for (auto &x : arr) is >> x; return is; } T &operator()(int h, int w) { assert(0 <= h && h < H && 0 <= w && w <= W); return table[h][w]; } /** * @brief サイズ n の単位行列。 */ static matrix<T> E(int N) { matrix<T> ret(N, N); for (int i = 0; i < N; i++) ret.table[i][i] = 1; return ret; } }; #line 3 "library/gandalfr/graph/edge.hpp" namespace internal { template <class DERIVED, class WEIGHT> struct _base_edge { int from; int to; WEIGHT cost; int id; _base_edge() {} _base_edge(int _from, int _to, WEIGHT _cost, int _id) : from(_from), to(_to), cost(_cost), id(_id) {} friend bool operator>(const _base_edge &e1, const _base_edge &e) { return e1.compare(e) > 0; } friend bool operator>=(const _base_edge &e1, const _base_edge &e) { return e1.compare(e) >= 0; } friend bool operator<(const _base_edge &e1, const _base_edge &e) { return e1.compare(e) < 0; } friend bool operator<=(const _base_edge &e1, const _base_edge &e) { return e1.compare(e) <= 0; } friend std::ostream &operator<<(std::ostream &os, const _base_edge<DERIVED, WEIGHT> &e) { e.print(os); return os; } _base_edge &operator=(const _base_edge &e) = default; virtual ~_base_edge() = default; DERIVED minmax() const { auto [f, t] = std::minmax(from, to); return {f, t, cost, id}; } DERIVED reverse() const { return {to, from, cost, id}; } operator int() const { return to; } protected: virtual void print(std::ostream &os) const = 0; virtual int compare(const _base_edge &e) const = 0; }; } // namespace internal template <class WEIGHT> struct edge : public internal::_base_edge<edge<WEIGHT>, WEIGHT> { using internal::_base_edge<edge<WEIGHT>, WEIGHT>::_base_edge; protected: void print(std::ostream &os) const override { os << this->from << " " << this->to << " " << this->cost; } int compare( const internal::_base_edge<edge<WEIGHT>, WEIGHT> &e) const override { if (this->cost == e.cost) { if (this->from == e.from) return this->to - e.to; return this->from - e.from; } return this->cost - e.cost; } }; template <> struct edge<int> : public internal::_base_edge<edge<int>, int> { static inline const int cost = 1; using internal::_base_edge<edge<int>, int>::_base_edge; edge(int _from, int _to, int _id) : _base_edge<edge<int>, int>(_from, _to, 0, _id) {} protected: void print(std::ostream &os) const override { os << this->from << " " << this->to; } int compare(const internal::_base_edge<edge<int>, int> &e) const override { if (this->from == e.from) return this->to - e.to; return this->from - e.from; } }; #line 12 "library/gandalfr/graph/graph.hpp" /** * @brief グラフを管理するクラス。 * @tparam WEIGHT int なら重みなし、そうでないなら重みつきグラフ * @tparam is_directed 有向グラフかとうか */ template <typename WEIGHT, bool is_directed> class graph { private: int N; std::vector<std::vector<edge<WEIGHT>>> G; std::vector<edge<WEIGHT>> E; union_find uf; WEIGHT W = 0; mutable std::vector<bool> visited; // dfs / bfs のための領域 bool forest_flag = true; const WEIGHT WEIGHT_MAX = std::numeric_limits<WEIGHT>::max(); void reset_visited_flag(int node) const { for (int x : uf.contained_group(node)) visited[x] = false; } void reset_visited_flag() const { visited.assign(N, false); } public: graph() : N(0){}; graph(int n) : N(n), G(n), uf(n), visited(n){}; /** * @brief ノードの数をn個まで増やす * @param n サイズ * @attention 今のノード数より小さい数を渡したとき、変化なし */ void expand(int n) { if (n <= N) return; N = n; G.resize(n); visited.resize(n); uf.expand(n); } /** * @return ノードの数 */ int count_nodes() const { return N; } /** * @return 辺の数 */ int count_edges() const { return E.size(); } /** * @param n ノード番号 * @return ノード n からの隣接頂点のリストの const 参照 */ const std::vector<edge<WEIGHT>> &operator[](int n) const { return G[n]; } /** * @return グラフ全体の辺のリストの const 参照 */ const std::vector<edge<WEIGHT>> &edges() const { return E; } /** * @param x ノード番号 * @param y ノード番号 * @return x, y が連結かどうか */ bool are_connected(int x, int y) const { return uf.same(x, y); } /** * @return 連結成分の数 */ int count_connected_components() const { return uf.count_groups(); } /** * @return 連結成分のリストのリスト */ std::vector<std::vector<int>> weakly_connected_components() const { return uf.all_groups(); } /** * @return 木か */ bool is_tree() const { return forest_flag && uf.count_groups() == 1; } /** * @return 森か */ bool is_forest() const { return forest_flag; } /** * @return グラフの重み */ WEIGHT weight() const { return W; } /** * @param e 辺 * @attention 渡した辺の id は保持される */ void add_edge(const edge<WEIGHT> &e) { forest_flag &= uf.merge(e.from, e.to); G[e.from].emplace_back(e); if (!is_directed && e.from != e.to) G[e.to].emplace_back(e.reverse()); if constexpr (is_directed) { E.emplace_back(e); } else { E.emplace_back(e.minmax()); } W += e.cost; } /** * @attention 辺の id は、(現在の辺の本数)番目 が振られる * @attention WEIGHT が int だとエラー */ void add_edge(int from, int to, WEIGHT cost) { static_assert(!std::is_same<WEIGHT, int>::value); add_edge({from, to, cost, (int)E.size()}); } /** * @attention 辺の id は、(現在の辺の本数)番目 が振られる * @attention WEIGHT が int 以外だとエラー */ void add_edge(int from, int to) { static_assert(std::is_same<WEIGHT, int>::value); add_edge({from, to, (int)E.size()}); } /** * @brief グラフを連結なグラフに分けてリストにして返す * @example auto[Gs, gr, nd] = G.decompose(); * @returns * 1.グラフのリスト * 2.各ノードがグラフのリストの何番目に属するか * 3.各ノードがグラフのどのノードになっているか */ std::tuple<std::vector<graph>, std::vector<int>, std::vector<int>> decompose() const { std::vector<graph> Gs(uf.count_groups()); std::vector<std::vector<int>> groups(uf.all_groups()); std::vector<int> group_id(N), node_id(N); for (int i = 0; i < (int)groups.size(); i++) { Gs[i].expand(groups[i].size()); for (int j = 0; j < (int)groups[i].size(); j++) { group_id[groups[i][j]] = i; node_id[groups[i][j]] = j; } } for (auto e : E) { int id = group_id[e.from]; e.from = node_id[e.from]; e.to = node_id[e.to]; Gs[id].add_edge(e); } return std::make_tuple(std::move(Gs), std::move(group_id), std::move(node_id)); } /** * @brief グラフを隣接行列に変換 * @param invalid 辺のないときの値 * @attention 自己ループが含まれていない限り、対角成分は 0 * @attention 多重辺を持たないと仮定 */ matrix<WEIGHT> to_adjajency(WEIGHT invalid = 0) const { matrix<WEIGHT> ret(N, N, invalid); for (int i = 0; i < N; i++) ret(i, i) = 0; for (int i = 0; i < N; i++) for (auto &e : G[i]) ret(i, e.to) = e.cost; return ret; } /** * @brief 行きがけ順に bfs */ std::vector<int> preorder(int start) const { std::vector<int> result; std::stack<std::pair<int, int>> stk; reset_visited_flag(start); visited[start] = true; stk.push({start, 0}); while (!stk.empty()) { auto &[cu, idx] = stk.top(); if (idx == 0) result.push_back(cu); if (idx == G[cu].size()) { stk.pop(); } else { int to = G[cu][idx++]; if (!visited[to]) { visited[to] = true; stk.push({to, 0}); } } } return result; } /** * @brief 通りがけ順に bfs */ std::vector<int> inorder(int start) const { std::vector<int> result; std::stack<std::pair<int, int>> stk; reset_visited_flag(start); visited[start] = true; stk.push({start, 0}); while (!stk.empty()) { auto &[cu, idx] = stk.top(); if (idx == G[cu].size()) { stk.pop(); result.push_back(cu); } else { int to = G[cu][idx++]; if (!visited[to]) { visited[to] = true; stk.push({to, 0}); result.push_back(cu); } } } return result; } /** * @brief 帰りがけ順に bfs */ std::vector<int> postorder(int start) const { std::vector<int> result; std::stack<std::pair<int, int>> stk; reset_visited_flag(start); visited[start] = true; stk.push({start, 0}); while (!stk.empty()) { auto &[cu, idx] = stk.top(); if (idx == G[cu].size()) { stk.pop(); result.push_back(cu); } else { int to = G[cu][idx++]; if (!visited[to]) { visited[to] = true; stk.push({to, 0}); } } } return result; } private: using PAIR = std::pair<WEIGHT, int>; using Dijkstra_queue = std::priority_queue<PAIR, std::vector<PAIR>, std::greater<PAIR>>; void run_bfs(std::vector<int> &dist, std::queue<int> &q) const { while (!q.empty()) { int cu = q.front(); q.pop(); for (auto &e : G[cu]) { if (dist[e.to] != WEIGHT_MAX) continue; dist[e.to] = dist[cu] + 1; q.push(e.to); } } } void run_Dijkstra(std::vector<WEIGHT> &dist, Dijkstra_queue &q) const { while (!q.empty()) { WEIGHT cur_dist = q.top().first; int cu = q.top().second; q.pop(); if (visited[cu]) continue; visited[cu] = true; for (auto &e : G[cu]) { WEIGHT alt = cur_dist + e.cost; if (dist[e.to] <= alt) continue; dist[e.to] = alt; q.push({alt, e.to}); } } } public: /** * @brief 最短距離を計算する * @param start_node 始点 * @param invalid 到達不能な頂点に格納される値 * @return 各ノードまでの最短距離のリスト */ std::vector<WEIGHT> distances(int start_node, WEIGHT invalid) const { std::vector<WEIGHT> dist(N, WEIGHT_MAX); dist[start_node] = 0; if constexpr (std::is_same<WEIGHT, int>::value) { // BFS algorithm std::queue<int> q; q.push(start_node); run_bfs(dist, q); } else { // Dijkstra's algorithm Dijkstra_queue q; q.push({0, start_node}); reset_visited_flag(start_node); run_Dijkstra(dist, q); } for (auto &x : dist) if (x == WEIGHT_MAX) x = invalid; return dist; } public: /** * @brief 最短距離を計算する * @param start_nodes 始点のリスト * @param invalid 到達不能な頂点に格納される値 * @return 各ノードまでの最短距離のリスト */ std::vector<WEIGHT> distances(const std::vector<int> &start_nodes, WEIGHT invalid) const { std::vector<WEIGHT> dist(N, WEIGHT_MAX); for (auto &x : start_nodes) dist[x] = 0; if constexpr (std::is_same<WEIGHT, int>::value) { // BFS algorithm std::queue<int> q; for (auto &x : start_nodes) q.push(x); run_bfs(dist, q); } else { // Dijkstra's algorithm Dijkstra_queue q; std::set<int> st; for (auto &x : start_nodes) { q.push({0, x}); st.insert(uf.leader(x)); } for (auto &x : st) { reset_visited_flag(x); } run_Dijkstra(dist, q); } for (auto &x : dist) if (x == WEIGHT_MAX) x = invalid; return dist; } matrix<WEIGHT> distances_from_all_nodes(WEIGHT invalid = -1) { auto mt(to_adjajency(WEIGHT_MAX)); int N = mt.size_H(); for (int k = 0; k < N; k++) // 経由する頂点 for (int i = 0; i < N; i++) // 始点 for (int j = 0; j < N; j++) // 終点 if (mt(i, k) != WEIGHT_MAX && mt(k, j) != WEIGHT_MAX) mt(i, j) = std::min(mt(i, j), mt(i, k) + mt(k, j)); for (int i = 0; i < N; ++i) for (int j = 0; j < N; ++j) if (mt(i, j) == WEIGHT_MAX) mt(i, j) = invalid; return mt; } /** * @brief 復元付き最短経路 * @attention 到達可能でないとき、空の配列で返る */ std::vector<edge<WEIGHT>> shortest_path(int start_node, int end_node) { if (start_node == end_node) return {}; auto dist = distances(start_node, WEIGHT_MAX); if (dist[end_node] == WEIGHT_MAX) return {}; auto R(this->reverse()); reset_visited_flag(end_node); visited[end_node] = true; int cu = end_node; std::vector<edge<WEIGHT>> route; while (cu != start_node) { for (auto e : R[cu]) { if (visited[e.to]) continue; if (dist[cu] - e.cost == dist[e.to]) { visited[cu = e.to] = true; route.push_back(e.reverse()); break; } } } std::reverse(route.begin(), route.end()); return route; } WEIGHT diameter() const { static_assert(!is_directed); assert(is_tree()); std::vector<WEIGHT> dist(distances(0, -1)); dist = distances( std::max_element(dist.begin(), dist.end()) - dist.begin(), -1); return *std::max_element(dist.begin(), dist.end()); } graph reverse() const { if constexpr (!is_directed) { return *this; } else { graph ret(N); for (auto &e : E) { ret.add_edge(e.reverse()); } return ret; } } std::vector<int> topological_sort() { static_assert(is_directed); std::vector<int> indeg(N, 0), sorted; for (int to : E) indeg[to]++; std::queue<int> q; for (int i = 0; i < N; i++) if (!indeg[i]) q.push(i); while (!q.empty()) { int cu = q.front(); q.pop(); for (int to : G[cu]) { if (!--indeg[to]) q.push(to); } sorted.push_back(cu); } return sorted; } /** * @return 最小全域森 */ graph minimum_spanning_forest() const { static_assert(!is_directed); graph ret(N); std::vector<edge<WEIGHT>> tmp(edges()); std::sort(tmp.begin(), tmp.end()); for (auto &e : tmp) if (!ret.are_connected(e.from, e.to)) ret.add_edge(e); return ret; } private: /** * @see https://ei1333.github.io/luzhiled/snippets/graph/lowlink.html * @attention 非連結でも動作 */ int run_lowlink(int idx, int k, int par, std::vector<int> &ord, std::vector<int> &low, std::vector<edge<WEIGHT>> &brds, std::vector<int> &apts) { visited[idx] = true; ord[idx] = k++; low[idx] = ord[idx]; bool is_apt = false; int cnt = 0; for (auto &e : G[idx]) { if (!visited[e.to]) { ++cnt; k = run_lowlink(e.to, k, idx, ord, low, brds, apts); low[idx] = std::min(low[idx], low[e.to]); is_apt |= ~par && low[e.to] >= ord[idx]; if (ord[idx] < low[e.to]) { brds.emplace_back(e.minmax()); } } else if (e.to != par) { low[idx] = std::min(low[idx], ord[e.to]); } } is_apt |= par == -1 && cnt > 1; if (is_apt) apts.push_back(idx); return k; } public: std::pair<std::vector<edge<WEIGHT>>, std::vector<int>> lowlink() { static_assert(!is_directed); std::vector<edge<WEIGHT>> brds; std::vector<int> apts, ord(N, 0), low(N, 0); reset_visited_flag(); int k = 0; for (int i = 0; i < N; i++) { if (!visited[i]) k = run_lowlink(i, k, -1, ord, low, brds, apts); } return {brds, apts}; } void print() const { std::cout << this->N << " " << this->E.size() << std::endl; for (const edge<WEIGHT> &e : this->E) std::cout << e << std::endl; } }; #line 5 "library/gandalfr/standard/mod_integer.hpp" inline long long mod_inverse(long long a, int mod) { assert(mod > 0); long long b = mod, u = 1, v = 0; while (b) { long long t = a / b; a -= t * b, std::swap(a, b); u -= t * v, std::swap(u, v); } u %= mod; if (u < 0) u += mod; return u; } // mod_integer<P> a := Pを法とするときの整数型; template <int mod> class mod_integer { private: long long val; // 値は必ず 0 <= val < mod に保たれる friend mod_integer operator+(const mod_integer &a) { return a; } friend mod_integer operator-(const mod_integer &a) { return -a.val; } friend mod_integer operator+(const mod_integer &a, const mod_integer &b) { return mod_integer(a.val + b.val); } friend mod_integer operator-(const mod_integer &a, const mod_integer &b) { return mod_integer(a.val - b.val); } friend mod_integer operator*(const mod_integer &a, const mod_integer &b) { return mod_integer(a.val * b.val); } friend mod_integer operator/(const mod_integer &a, const mod_integer &b) { return mod_integer((a.val * mod_inverse(b.val, mod)) % mod); } friend bool operator==(const mod_integer &a, const mod_integer &b) { return a.val == b.val; } friend bool operator!=(const mod_integer &a, const mod_integer &b) { return a.val != b.val; } // map とかに乗せたいので、便宜的に定義 friend bool operator<(const mod_integer &a, const mod_integer &b) { return a.val < b.val; } public: mod_integer(long long n) : val(n) { val %= mod; if (val < 0) val += mod; } mod_integer() : val(0) {} int value() const { return (int)val; } mod_integer &operator=(const mod_integer &a) = default; mod_integer &operator+=(const mod_integer &a) { val += a.val; if (val >= mod) val -= mod; return *this; } mod_integer &operator-=(const mod_integer &a) { val -= a.val; if (val < 0) val += mod; return *this; } mod_integer &operator*=(const mod_integer &a) { (val *= a.val) %= mod; return *this; } mod_integer &operator/=(const mod_integer &a) { (val *= mod_inverse(a.val, mod)) %= mod; return *this; } friend std::istream &operator>>(std::istream &is, mod_integer &a) { is >> a.val; a.val %= mod; if (a.val < 0) a.val += mod; return is; } friend std::ostream &operator<<(std::ostream &os, const mod_integer &a) { os << a.val; return os; } }; // d_mod_integer<P> a := Pを法とするときの整数型; template <int id> class dynamic_mod_integer { private: using d_mod_integer = dynamic_mod_integer<id>; static inline int mod = 998244353; long long val; // 値は必ず 0 <= val < mod に保たれる friend d_mod_integer operator+(const d_mod_integer &a) { return a; } friend d_mod_integer operator-(const d_mod_integer &a) { return -a.val; } friend d_mod_integer operator+(const d_mod_integer &a, const d_mod_integer &b) { return d_mod_integer(a.val + b.val); } friend d_mod_integer operator-(const d_mod_integer &a, const d_mod_integer &b) { return d_mod_integer(a.val - b.val); } friend d_mod_integer operator*(const d_mod_integer &a, const d_mod_integer &b) { return d_mod_integer(a.val * b.val); } friend d_mod_integer operator/(const d_mod_integer &a, const d_mod_integer &b) { return d_mod_integer((a.val * mod_inverse(b.val, mod)) % mod); } friend bool operator==(const d_mod_integer &a, const d_mod_integer &b) { return a.val == b.val; } friend bool operator!=(const d_mod_integer &a, const d_mod_integer &b) { return a.val != b.val; } // map とかに乗せたいので、便宜的に定義 friend bool operator<(const d_mod_integer &a, const d_mod_integer &b) { return a.val < b.val; } public: dynamic_mod_integer(long long n) : val(n) { val %= mod; if (val < 0) val += mod; } dynamic_mod_integer() : val(0) {} int value() const { return (int)val; } static void set_mod(int _mod) { assert(_mod >= 0); mod = _mod; } d_mod_integer &operator=(const d_mod_integer &a) = default; d_mod_integer &operator+=(const d_mod_integer &a) { val += a.val; if (val >= mod) val -= mod; return *this; } d_mod_integer &operator-=(const d_mod_integer &a) { val -= a.val; if (val < 0) val += mod; return *this; } d_mod_integer &operator*=(const d_mod_integer &a) { (val *= a.val) %= mod; return *this; } d_mod_integer &operator/=(const d_mod_integer &a) { (val *= mod_inverse(a.val, mod)) %= mod; return *this; } friend std::istream &operator>>(std::istream &is, d_mod_integer &a) { is >> a.val; a.val %= mod; if (a.val < 0) a.val += mod; return is; } friend std::ostream &operator<<(std::ostream &os, const d_mod_integer &a) { os << a.val; return os; } }; using mint = mod_integer<1000000007>; using _mint = mod_integer<998244353>; using dmint = dynamic_mod_integer<-1>; #line 5 "playspace/main.cpp" using namespace std; using ll = long long; const int INF = 1001001001; const int MAXINT = std::numeric_limits<int>::max(); const int MININT = std::numeric_limits<int>::min(); const ll INFLL = 1001001001001001001; const ll MAXLL = std::numeric_limits<ll>::max(); const ll MINLL = std::numeric_limits<ll>::min(); const ll MOD = 1000000007; const ll _MOD = 998244353; #define rep(i, j, n) for(ll i = (ll)(j); i < (ll)(n); i++) #define rrep(i, j, n) for(ll i = (ll)(n-1); i >= (ll)(j); i--) #define all(a) (a).begin(),(a).end() #define LF cout << endl #ifdef ENABLE_MULTI_FOR #define mrep(it, mr) for(multi_iter it(mr); !it.fin(); ++it) #endif template<typename T1, typename T2> inline bool chmax(T1 &a, T2 b) { return a < b && (a = b, true); } template<typename T1, typename T2> inline bool chmin(T1 &a, T2 b) { return a > b && (a = b, true); } void Yes(bool ok){ std::cout << (ok ? "Yes" : "No") << std::endl; } int main(void){ int N, M; cin >> N >> M; vector<int> ord(N); iota(all(ord), 0); rep(i,0,M) { int c; cin >> c; vector<int> A(c); cin >> A; rep(j,0,c) A[j]--; int keep = ord[A[0]]; rrep(j,0,c) { ord[A[(j+1)%c]] = ord[A[j]]; } ord[A[1]] = keep; } graph<int, false> G(N); rep(i,0,N) { G.add_edge(i, ord[i]); } auto cc = G.weakly_connected_components(); _mint ans = 1; rep(i,0,cc.size()) { ans *= cc[i].size(); } cout << ans << endl; }