poj_1204 Trie图

题目大意

    给出一个RxC的字符组成的puzzle，中间可以从左向右，从右到左，从上到下，从下到上，从左上到右下，从右下到左上，从左下到右上，从右上到左下，八个方向进行查找字符串。 
    给出M个字符串，找出他们在puzzle中的位置，返回该字符串在puzzle中的起点横纵坐标以及方向。

字符串长度L <=1000, R,C <= 1000, W <= 1000

题目分析

    多模式串的字符串匹配问题，考虑使用Trie图。将M个待查的字符串作为模式串插入Trie图中，然后设置前缀指针，构造DFA。 
    查找的时候，在puzzle的四个边上每个点，沿8个方向分别确定最长的串作为母串，用母串在Trie图上进行行走，进行匹配。

题目比较坑的是，special judge没能对有些正确的结果给AC，只能按照“正常”的顺序来查找。

实现(c++)

#define _CRT_SECURE_NO_WARNINGS
#include<stdio.h>
#include<string.h>
#include<vector>
#include<deque>
using namespace std;
#define MAX_SIZE 1005
#define LETTERS 26
#define MAX_NODES 150000
#define MIN(a, b) a < b? a :b
char gPizza[MAX_SIZE][MAX_SIZE];
struct PosInfo{
	int row;
	int col;
	int dir;
	void SetInfo(int r, int c, int d){
		row = r;
		col = c;
		dir = d;
	}
};
PosInfo gPosInfo[MAX_SIZE];



struct Node{
	Node* childs[LETTERS];
	bool danger_node;
	Node* prev;
	int pattern_index;
	Node(){
		memset(childs, 0, sizeof(childs));
		prev = NULL;
		danger_node = false;
		pattern_index = 0;		//可以指示某个终止节点确定的字符串是第几个pattern
	}
};
Node gNodes[MAX_NODES];
int gNodeCount = 2;

void Insert(Node* root, char* str, int pattern_index){
	Node* node = root;
	char*p = str;
	while (*p != '\0'){
		int index = *p - 'A';
		if (! node->childs[index]){
			node->childs[index] = gNodes + gNodeCount++;
		}
		node = node->childs[index];
		p++;
	}
	node->danger_node = true;
	node->pattern_index = pattern_index;
}

void BuildDfa(){
	Node* root = gNodes + 1;
	for (int i = 0; i < LETTERS; i++){
		gNodes[0].childs[i] = root;
	}
	root->prev = gNodes;
	gNodes[0].prev = NULL;

	deque<Node*> Q;
	Q.push_back(root);
	while (!Q.empty()){
		Node* node = Q.front();
		Node* prev = node->prev, *p;
		Q.pop_front();
		for (int i = 0; i < LETTERS; i++){
			if (node->childs[i]){
				p = prev;
				while (p && !p->childs[i]){
					p = p->prev;
				}
				node->childs[i]->prev = p->childs[i];
				//这个地方注意，不能写成 p->childs[i]->danger_node = node->childs[i]->danger_node
				if (p->childs[i]->danger_node)				
					node->childs[i]->danger_node = true;
				Q.push_back(node->childs[i]);
			}
		}
	}
}

bool gPatterFind[MAX_SIZE];
int gPatternFoundNum = 0;
int gMinPatternLen = 0;
int gPatternLen[MAX_SIZE];
int gMoveStep[8][2] = { { -1, 0 }, { -1, 1 }, { 0, 1 }, { 1, 1 }, { 1, 0 }, { 1, -1 }, {0, -1 }, { -1, -1 } };

//在Trie图上达到一个“危险”节点，则该节点的各个前缀指针，仍然可能为“终止”节点，沿前缀指针找出所有的终止节点，以防止遗漏
//比如 ABCDFF 中查找 ABCD CD 若到达D，确定为一个危险节点，可以找到ABCD，若不沿着前缀指针找出所有的终止节点，则会遗漏CD
void FindPatternFromEndPoint(Node* node, int r, int c, int dir){
	do{
		if (node->pattern_index == 0){
			node = node->prev;
			continue;
		}

		int pattern_index = node->pattern_index;
		if (gPatterFind[pattern_index]){		//此时找到的串，有可能是别的串的前缀，因此继续向后找
			node = node->prev;
			continue;
		}
		gPatterFind[pattern_index] = true;
		gPatternFoundNum++;

		int beg_r = r - gPatternLen[pattern_index] * gMoveStep[dir][0];
		int beg_c = c - gPatternLen[pattern_index] * gMoveStep[dir][1];
		if (gMoveStep[dir][0] == 0)
			beg_r--;

		if (gMoveStep[dir][1] == 0)
			beg_c--;

		if (dir == 1 || dir == 7 || dir == 0){
			beg_r -= 2;
		}
		if (dir == 5 || dir == 7 || dir == 6){
			beg_c -= 2;
		}
		gPosInfo[pattern_index].SetInfo(beg_r, beg_c, dir);
		
		node = node->prev;
	} while (node);

}
//从某个边界点出发，沿某个方向的最长字符串作为母串，在Trie图上进行查找
void SearchStr(int start_x, int start_y, int dir){
	int r = start_x, c = start_y;
	Node* node = gNodes + 1;
	while (gPizza[r][c] != '\0'){
		int index = gPizza[r][c] - 'A';
		while (node && node->childs[index] == NULL){
			node = node->prev;
		}
		node = node->childs[index];
		if (node->danger_node){
			FindPatternFromEndPoint(node, r, c, dir);
		}
		r += gMoveStep[dir][0];
		c += gMoveStep[dir][1];
	}
}
//确定在边界上的某个点，沿某个方向所构成最长字符串的长度
int MaxLen(int R, int C, int r, int c, int dir){
	if (dir == 0 || dir == 4)
		return R;
	if (dir == 2 || dir == 6)
		return C;
	if (dir == 1){
		if (c == 1)
			return r;
		else if (r == R)
			return C - c + 1;
	}
	if (dir == 5){
		if (r == 1)
			return c;
		else if (c == C)
			return R - r + 1;
	}
	if (dir == 3){
		if (r == 1)
			return C - c + 1;
		if (c == 1)
			return R - r + 1;
	}
	if (dir == 7){
		if (r == R)
			return c;
		if (c == C)
			return r;
	}
	return -1;
}

//对边界上的每个点，在8个方向进行查找
void SearchPuzzle(int R, int C, int total_word_to_find){
	for (int r = 1; r <= R; r++){
		for (int dir = 0; dir < 8; dir++){
			if (gPatternFoundNum == total_word_to_find){
				return;
			}
			if (MaxLen(R, C, r, 1, dir) >= gMinPatternLen){
				SearchStr(r, 1, dir);
			}
		}		
	}
	for (int r = 1; r <= R; r++){
		for (int dir = 0; dir < 8; dir++){
			if (gPatternFoundNum == total_word_to_find){
				return;
			}
			if (MaxLen(R, C, r, C, dir) >= gMinPatternLen){
				SearchStr(r, C, dir);
			}
		}
	}
	for (int c = 1; c <= C; c++){
		for (int dir = 0; dir < 8; dir++){
			if (gPatternFoundNum == total_word_to_find){
				return;
			}
			if (MaxLen(R, C, 1, c, dir) >= gMinPatternLen){
				SearchStr(1, c, dir);
			}
		}
	}
	for (int c = 1; c <= C; c++){
		for (int dir = 0; dir < 8; dir++){
			if (gPatternFoundNum == total_word_to_find){
				return;
			}
			if (MaxLen(R, C, R, c, dir) >= gMinPatternLen){
				SearchStr(R, c, dir);
			}
		}
	}
}


int main(){
	int R, C, M;
	scanf("%d %d %d", &R, &C, &M);	
	memset(gPizza, 0, sizeof(gPizza));
	memset(gPatterFind, false, sizeof(gPatterFind));
	gNodeCount = 2;

	for (int r = 1; r <= R; r++){
		getchar();
		for (int c = 1; c <= C; c++){
			scanf("%c", &gPizza[r][c]);
		}
	}
	getchar();
	char str[MAX_SIZE];
	Node* root = gNodes + 1;
	for (int i = 1; i <= M; i++){
		scanf("%s", str);
		Insert(root, str, i);
		gPatternLen[i] = strlen(str);
		gMinPatternLen = MIN(gMinPatternLen, gPatternLen[i]);		
	}	
	
	BuildDfa();

	SearchPuzzle(R, C, M);
	for (int i = 1; i <= M; i++){
		printf("%d %d %c\n", gPosInfo[i].row, gPosInfo[i].col, gPosInfo[i].dir + 'A');
	}
	return 0;
}