Aho-Corasick automation

多串匹配，类似于KMP，可以看成是KMP的推广。

首先建立所有匹配串的trie，然后对该trie建立fail指针。

匹配文本时，只要让文本在改自动机上跑一遍就可以得到所有匹配成功的位置。

具体实现可以用指针，也可以用数组，但是数组要求的空间很大。

#include <iostream> #include <cstdio> #include <cstring> #include <climits> using namespace std; const int N = 1000000; const int M = 50; const int KEYSIZE = 26; struct node { node *fail; node *next[KEYSIZE]; int count; }; const int KEYWORD_NUM = 10000; node *q[KEYWORD_NUM * M]; node table[KEYWORD_NUM * M], *last; void insert(char *str, node *root) { node *p = root; for (int i = 0; str[i]; ++i) { int j = str[i] - 'a'; if (p->next[j] == NULL) { last->fail = NULL; memset(last->next, 0, sizeof(node *) * KEYSIZE); last->count = 0; p->next[j] = last++; } p = p->next[j]; } p->count++; } void build_ac_automation(node *root) { int head, tail; q[0] = root; for (head = 0, tail = 1; head < tail; ++head) { node *p = q[head]; for (int i = 0; i < KEYSIZE; ++i) { if (p->next[i] == NULL) continue; node *r; for (r = p->fail; r; r = r->fail) if (r->next[i]) { p->next[i]->fail = r->next[i]; break; } if (r == NULL) p->next[i]->fail = root; q[tail++] = p->next[i]; } } } int query(char *str, node *root) { int ret = 0; node *p = root; for (int i = 0; str[i]; ++i) { int j = str[i] - 'a'; while (p->next[j] == NULL && p != root) p = p->fail; p = p->next[j]; if (p == NULL) p = root; for (node *r = p; r != root && r->count != -1; r = r->fail) { ret += r->count; r->count = -1; } } return ret; } char str[N + 1]; char keyword[M + 1]; int main() { int n, cas; node *root; scanf("%d", &cas); while (cas--) { last = table; scanf("%d", &n); last->fail = NULL; memset(last->next, 0, sizeof(node *) * KEYSIZE); root = last++; while (n--) { scanf("%s", keyword); insert(keyword, root); } build_ac_automation(root); scanf("%s", str); printf("%d/n", query(str, root)); } return 0; }

有一个优化，就是fail指针的传递可以用类似disjoint-set路径压缩的技巧把上面代码的第51行的for去掉。

只需要修改build_ac_automation部分代码：

#include <iostream> #include <cstdio> #include <cstring> #include <climits> using namespace std; const int N = 1000000; const int M = 50; const int KEYSIZE = 26; struct node { node *fail; node *next[KEYSIZE]; int count; }; const int KEYWORD_NUM = 10000; node *q[KEYWORD_NUM * M]; node table[KEYWORD_NUM * M], *last; void insert(char *str, node *root) { node *p = root; for (int i = 0; str[i]; ++i) { int j = str[i] - 'a'; if (p->next[j] == NULL) { last->fail = root; memset(last->next, 0, sizeof(node *) * KEYSIZE); last->count = 0; p->next[j] = last++; } p = p->next[j]; } p->count++; } void build_ac_automation(node *root) { int head, tail; head = tail = 0; for (int i = 0; i < KEYSIZE; ++i) if (root->next[i]) q[tail++] = root->next[i]; for (; head < tail; ++head) { node *p = q[head]; for (int i = 0; i < KEYSIZE; ++i) { if (p->next[i] == NULL) { p->next[i] = p->fail->next[i]; } else { p->next[i]->fail = p->fail->next[i]; if (p->next[i]->fail == NULL) p->next[i]->fail = root; q[tail++] = p->next[i]; } } } } int query(char *str, node *root) { int ret = 0; node *p = root; for (int i = 0; str[i]; ++i) { int j = str[i] - 'a'; p = p->next[j]; if (p == NULL) p = root; for (node *r = p; r != root && r->count != -1; r = r->fail) { ret += r->count; r->count = -1; } } return ret; } char str[N + 1]; char keyword[M + 1]; int main() { int n, cas; node *root; scanf("%d", &cas); while (cas--) { last = table; scanf("%d", &n); last->fail = NULL; memset(last->next, 0, sizeof(node *) * KEYSIZE); root = last++; while (n--) { scanf("%s", keyword); insert(keyword, root); } build_ac_automation(root); scanf("%s", str); printf("%d/n", query(str, root)); } return 0; }