LeetCode765. Couples Holding Hands

N couples sit in 2N seats arranged in a row and want to hold hands. We want to know the minimum number of swaps so that every couple is sitting side by side. A swap consists of choosing any two people, then they stand up and switch seats.

The people and seats are represented by an integer from 0 to 2N-1, the couples are numbered in order, the first couple being (0, 1), the second couple being (2, 3), and so on with the last couple being (2N-2, 2N-1).

The couples' initial seating is given by row[i] being the value of the person who is initially sitting in the i-th seat.

Example 1:

Input: row = [0, 2, 1, 3]
Output: 1
Explanation: We only need to swap the second (row[1]) and third (row[2]) person.

Example 2:

Input: row = [3, 2, 0, 1]
Output: 0
Explanation: All couples are already seated side by side.

Note:

1. len(row) is even and in the range of [4, 60].
2. row is guaranteed to be a permutation of 0...len(row)-1.

思路

本来以为是用dp解题，然而不是，还是好好背题吧。解法有 cyclic swapping，并差集，贪心这三种。

完整解释：https://leetcode.com/problems/couples-holding-hands/discuss/113362/JavaC++-O(N)-solution-using-cyclic-swapping

一大串英文解释，看的不是很懂，百度了下，看到一片篇解释得很易懂的博文：

https://www.cnblogs.com/grandyang/p/8716597.html

首先是贪心的解法：

class Solution {
    public int minSwapsCouples(int[] row) {
        int res=0, n=row.length;
        for(int i=0;i<n;i=i+2){
            if(row[i+1]==(row[i]^1)) continue;
            res++;
            for(int j=i+1;j<n;j++){
                if(row[j]==(row[i]^1)){　　// 这里注意要加括号，因为java中恒等运算符的优先级大于位运算
                    row[j]=row[i+1];
                    row[i+1]=(row[i]^1);
                    break;
                }
            }
        }
        return res;
    }
}

接下来是并查集的解法，关于并查集算法的解释可以见这篇博文：https://blog.csdn.net/dm_vincent/article/details/7655764

LeetCode上的解释：

Think about each couple as a vertex(顶点) in the graph. So if there are N couples, there are N vertices. Now if in position 2i and 2i +1 there are person from couple u and couple v sitting there, that means that the permutations are going to involve u and v. So we add an edge to connect u and v. The min number of swaps = N - number of connected components. This follows directly from the theory of permutations. Any permutation can be decomposed into a composition of cyclic permutations. If the cyclic permutation involve k elements, we need k -1 swaps. You can think about each swap as reducing the size of the cyclic permutation by 1. So in the end, if the graph has k connected components, we need N - k swaps to reduce it back to N disjoint vertices.

class Solution {
    private class UF {
        private int[] parents;
        public int count;
        UF(int n) {　　// 初始化组号
            parents = new int[n];
            for (int i = 0; i < n; i++) {
                parents[i] = i;　　// i-具体节点的值，parents[i]-节点i所对应的组号，放在这题中i就是couple的编号，数组值就是这个couple应该在的组号
            }
            count = n;
        }

        private int find(int i) {
            if (parents[i] == i) {　　// 如果couple的编号和组号对应，所在组号正确，直接返回组号
                return i;
            }
            parents[i] = find(parents[i]);　　// 这种情形时发生了标记1的情况，连接后组号被修改过，不会和原来对应
            return parents[i];
        }

        public void union(int i, int j) {
            int a = find(i);
            int b = find(j);
            if (a != b) {　　// 如果不在一个组，连接之
                parents[a] = b;　　// 将a的组号改成b的，注意原parents数组如果组号是a，那么其数组索引也是a。标记1
                count--;　　
            }
        }
    }
    public int minSwapsCouples(int[] row) {
        int N = row.length/ 2;
        UF uf = new UF(N);　　// 并查集初始化组号
        for (int i = 0; i < N; i++) {
            int a = row[2*i];
            int b = row[2*i + 1];
            uf.union(a/2, b/2);
        }
        return N - uf.count;
    }
}