$\newcommand{\O}{\mathrm{O}}$
時間計算量: 構築にかかる時間計算量, 空間計算量は O(σS) (S は文字列の合計の長さとする.) (以下の実装では実際 σ 倍かかるケースは珍しく, 感覚的には O(S) と考えてよい.)
// 使う文字は 31 文字以下の文字コード上で連続する文字の列とする. template<char START_CHARACTER, unsigned int CHARACTER_SIZE> class CompressedTrie { private: struct node { string *s; node **to; // sub: 部分木に含まれる要素の数, adj: 子の bit 表現, cnt: ここで終わる頂点の数 uint32_t sub, adj, cnt; node() : s(nullptr), to(nullptr), sub(1u), adj(0u), cnt(1u){} node(string&& _s, node *v, unsigned int index, uint32_t _sub, uint32_t _cnt) : s(new string[CHARACTER_SIZE]()), to(new node*[CHARACTER_SIZE]()), sub(_sub), adj(1u << index), cnt(_cnt){ s[index] = move(_s), to[index] = v; } // ~node(){ delete[] s, delete[] to; } #define lsb(v) (__builtin_ctz(v)) inline unsigned int begin() const { return adj ? lsb(adj) : CHARACTER_SIZE; } inline unsigned int next(unsigned int cur) const { cur = adj & ~((1u << (cur + 1u)) - 1u); return cur ? lsb(cur) : CHARACTER_SIZE; } inline static unsigned int end(){ return CHARACTER_SIZE; } inline bool isExist(const unsigned int v) const { return adj >> v & 1u; } inline bool isFinal() const { return !s; } void direct_push(string&& _s, unsigned int index){ if(!s) s = new string[CHARACTER_SIZE](), to = new node*[CHARACTER_SIZE](); s[index] = move(_s), to[index] = new node(), ++sub, adj |= (1u << index); } }; void make_node(string& orgs, unsigned int start, node*& to, bool is_end){ string tmp = orgs.substr(0, start); orgs.erase(orgs.begin(), orgs.begin() + start); to = new node(move(orgs), to, orgs[0] - START_CHARACTER, to->sub + is_end, is_end); orgs = move(tmp); } void new_push(const string& s, unsigned int index, node *to){ string _s(s.substr(index, s.size() - index)); to->direct_push(move(_s), s[index] - START_CHARACTER); } void new_push(string&& s, unsigned int index, node *to){ s.erase(s.begin(), s.begin() + index); to->direct_push(move(s), s[0] - START_CHARACTER); } template<typename String> void push(node *cur, String&& news){ if(news.size() == 0u){ ++cur->sub, ++cur->cnt; return; } const unsigned int _ls = news.size(); unsigned int index = 0u, prefix; while(true){ const unsigned int num = news[index] - START_CHARACTER; if(cur->isExist(num)){ ++cur->sub; string& orgs = cur->s[num]; const unsigned int ls = orgs.size(); for(prefix = 0u; prefix < ls && index < _ls; ++prefix, ++index){ if(orgs[prefix] == news[index]) continue; make_node(orgs, prefix, cur->to[num], false); new_push(forward<String>(news), index, cur->to[num]); return; } if(index == _ls){ if(prefix == ls){ ++cur->to[num]->sub, ++cur->to[num]->cnt; return; } make_node(orgs, prefix, cur->to[num], true); return; }else{ cur = cur->to[num]; } }else{ new_push(forward<String>(news), index, cur); return; } } } // void clear_dfs(node *cur){ // if(!cur) return; // for(unsigned int i = 0u; i < CHARACTER_SIZE; ++i) if(cur->to) clear_dfs(cur->to[i]); // delete cur; // } public: node* root; CompressedTrie() : root(new node()){ --root->cnt; } // ~CompressedTrie(){ clear_dfs(root); } // CompressedTrie 木に s を加える void add(const string& s){ push(root, s); } void add(string&& s){ push(root, move(s)); } // 文字列 s がいくつ含まれるか int query1(const string& s){ if(s.size() == 0u) return root->cnt; node *cur = root; int i, d = 0; while(true){ if(d == (int)s.size()) return cur->cnt; if(cur->isFinal()) break; const unsigned int next = s[d] - START_CHARACTER; if(!cur->isExist(next)) return 0; for(i = 0; i < (int)cur->s[next].size(); ++i){ if(d + i >= (int)s.size() || s[d + i] != cur->s[next][i]) return 0; } d += (int)cur->s[next].size(); cur = cur->to[next]; } return 0; } // 文字列 s を prefix とする文字列はいくつか int query2(const string& s){ node *cur = root; int d = 0; while(true){ if(d >= (int)s.size()) return cur->sub; if(cur->isFinal()) break; const unsigned int next = s[d] - START_CHARACTER; if(!cur->isExist(next)) return 0; for(int i = 0; i < min((int)cur->s[next].size(), (int)s.size() - d); ++i){ if(s[d + i] != cur->s[next][i]) return 0; } d += (int)cur->s[next].size(); cur = cur->to[next]; } return 0; } // 辞書順で文字列 s 以下の文字列はいくつか int query3(const string& s){ node *cur = root; int d = 0, res = 0; while(true){ if(d <= (int)s.size()) res += cur->cnt; if(d >= (int)s.size() || cur->isFinal()) break; const unsigned int next = s[d] - START_CHARACTER; for(unsigned int i = cur->begin(); i < next; i = cur->next(i)){ res += cur->to[i]->sub; } if(!cur->isExist(next)) break; for(int i = 0; i < min((int)cur->s[next].size(), (int)s.size() - d); ++i){ if(s[d + i] < cur->s[next][i]) return res; else if(s[d + i] > cur->s[next][i]) return res + cur->to[next]->sub; } d += (int)cur->s[next].size(); cur = cur->to[next]; } return res; } };
Atcoder : Lexicographical disorder 提出コード(非想定解)