【剑指 Offer】矩阵中的路径

题目描述

给定一个 m x n 二维字符网格 board 和一个字符串单词 word 。如果 word 存在于网格中，返回 true ；否则，返回 false 。

单词必须按照字母顺序，通过相邻的单元格内的字母构成，其中“相邻”单元格是那些水平相邻或垂直相邻的单元格。同一个单元格内的字母不允许被重复使用。

题目链接:https://leetcode.cn/problems/ju-zhen-zhong-de-lu-jing-lcof/

解答

递归

自己想的递归方法，想节省空间用字符串记录访问路径，后来发现还不如用数组，不过也算自己独立完成的，笨是笨了些，好在大体思路正确。
时间：2354 ms，内存：41.9 MB。

class Solution {
    public boolean exist(char[][] board, String word) {
        int row = board.length;
        int col = board[0].length;
        for (int i = 0; i < row; i++) {
            for (int j = 0; j < col; j++) {
                if (checkWord(board, word, 0, "", i, j) == true) return true;
            }
        }
        return false;
    }

    public boolean checkWord (char[][] board, String word, int charIndex, String path, int rowIndex, int colIndex) {
        if (charIndex == word.length()) return true;
        if (rowIndex < 0 || rowIndex >= board.length || colIndex < 0 || colIndex >= board[0].length) return false;
        if (path.indexOf(""+rowIndex+colIndex) > -1) return false;

        if (board[rowIndex][colIndex] == word.charAt(charIndex)) {
            path = path + rowIndex + colIndex + " ";
            return checkWord(board, word, charIndex+1, path, rowIndex-1, colIndex)
                || checkWord(board, word, charIndex+1, path, rowIndex+1, colIndex)
                || checkWord(board, word, charIndex+1, path, rowIndex, colIndex-1)
                || checkWord(board, word, charIndex+1, path, rowIndex, colIndex+1);
        } else {
            return false;
        }
    }
}

回溯

参考官方思路修改了代码，重点是使用访问数组记录已经访问的元素，遍历失败时回溯到上个节点寻找其他路径，并恢复访问数组元素。
时间：168 ms，内存：41.7 MB。

class Solution {
    public boolean exist(char[][] board, String word) {
        int row = board.length, col = board[0].length;
        boolean[][] visited = new boolean[row][col];
        for (int i = 0; i < row; i++) {
            for (int j = 0; j < col; j++) {
                if (checkWord(board, word, visited, i, j, 0)) return true;
            }
        }
        return false;
    }

    public boolean checkWord (char[][] board, String word, boolean[][] visited, int rowIndex, int colIndex, int wordIndex) {
        if (board[rowIndex][colIndex] != word.charAt(wordIndex)){
            return false;
        } else if (wordIndex == word.length()-1) {
            return true;
        }

        visited[rowIndex][colIndex] = true;
        int[][] directions = {{-1, 0}, {1, 0}, {0, -1}, {0, 1}};
        for (int[] dir : directions) {
            int newRow = rowIndex + dir[0];
            int newCol = colIndex + dir[1];
            if (newRow >= 0 && newRow < board.length && newCol >= 0 && newCol < board[0].length) {
                if (!visited[newRow][newCol]) {
                    if (checkWord(board, word, visited, newRow, newCol, wordIndex+1)) return true;
                }
            }
        }
        visited[rowIndex][colIndex] = false;
        return false;
    }
}

回溯（优化？）

除了开辟新数组用于存放访问元素，也可以直接在原数组上进行修改，将访问过的元素改为空白字符，不过这种方法感觉不是很规范。
时间：166 ms，内存：41.7 MB。

class Solution {
    public boolean exist(char[][] board, String word) {
        int row = board.length, col = board[0].length;
        for (int i = 0; i < row; i++) {
            for (int j = 0; j < col; j++) {
                if (checkWord(board, word, i, j, 0)) return true;
            }
        }
        return false;
    }

    public boolean checkWord (char[][] board, String word, int rowIndex, int colIndex, int wordIndex) {
        if (board[rowIndex][colIndex] != word.charAt(wordIndex)){
            return false;
        } else if (wordIndex == word.length()-1) {
            return true;
        }

        board[rowIndex][colIndex] = ' ';
        int[][] directions = {{-1, 0}, {1, 0}, {0, -1}, {0, 1}};
        for (int[] dir : directions) {
            int newRow = rowIndex + dir[0];
            int newCol = colIndex + dir[1];
            if (newRow >= 0 && newRow < board.length && newCol >= 0 && newCol < board[0].length) {
                if (checkWord(board, word, newRow, newCol, wordIndex+1)) return true;
            }
        }
        board[rowIndex][colIndex] = word.charAt(wordIndex);
        return false;
    }
}

回溯（时间优化）

其实思路大多数是一致的，代码细节上不一样，结果区别也挺大。
时间：87 ms，内存：39.5 MB。

class Solution {
    public boolean exist(char[][] board, String word) {
        int row = board.length, col = board[0].length;
        for (int i = 0; i < row; i++) {
            for (int j = 0; j < col; j++) {
                if (checkWord(board, word, i, j, 0)) return true;
            }
        }
        return false;
    }

    public boolean checkWord (char[][] board, String word, int rowIndex, int colIndex, int wordIndex) {
        if (wordIndex == word.length()) return true;
        if (rowIndex < 0 || rowIndex == board.length || colIndex < 0 || colIndex == board[0].length || board[rowIndex][colIndex] != word.charAt(wordIndex)) return false;

        board[rowIndex][colIndex] = ' ';
        boolean res = checkWord(board, word, rowIndex+1, colIndex, wordIndex+1)
            || checkWord(board, word, rowIndex-1, colIndex, wordIndex+1)
            || checkWord(board, word, rowIndex, colIndex+1, wordIndex+1)
            || checkWord(board, word, rowIndex, colIndex-1, wordIndex+1);
        board[rowIndex][colIndex] = word.charAt(wordIndex);
        return res;
    }
}