CodeForces - 811C Vladik and Memorable Trip (普通DP）

Vladik often travels by trains. He remembered some of his trips especially well and I would like to tell you about one of these trips:

Vladik is at initial train station, and now n people (including Vladik) want to get on the train. They are already lined up in some order, and for each of them the city code a_i is known (the code of the city in which they are going to).

Train chief selects some number of disjoint segments of the original sequence of people (covering entire sequence by segments is not necessary). People who are in the same segment will be in the same train carriage. The segments are selected in such way that if at least one person travels to the city x, then all people who are going to city x should be in the same railway carriage. This means that they can’t belong to different segments. Note, that all people who travel to the city x, either go to it and in the same railway carriage, or do not go anywhere at all.

Comfort of a train trip with people on segment from position l to position r is equal to XOR of all distinct codes of cities for people on the segment from position l to position r. XOR operation also known as exclusive OR.

Total comfort of a train trip is equal to sum of comfort for each segment.

Help Vladik to know maximal possible total comfort.

Input

First line contains single integer n (1 ≤ n ≤ 5000) — number of people.

Second line contains n space-separated integers a₁, a₂, ..., a_n (0 ≤ a_i ≤ 5000), where a_i denotes code of the city to which i-th person is going.

Output

The output should contain a single integer — maximal possible total comfort.

Example

Input

6
4 4 2 5 2 3

Output

14

Input

9
5 1 3 1 5 2 4 2 5

Output

9

Note

In the first test case best partition into segments is: [4, 4] [2, 5, 2] [3], answer is calculated as follows: 4 + (2 xor 5) + 3 = 4 + 7 + 3 = 14

In the second test case best partition into segments is: 5 1 [3] 1 5 [2, 4, 2] 5, answer calculated as follows: 3 + (2 xor 4) = 3 + 6 = 9.

题意:给一个序列，求这个序列的不相交的区间亦或最大和，区间首尾必须是相同元素的第一个和最后一个。

思路：首先预处理所有元素的首位置和末位置，定义dp[i]表示以i结尾的子序列最大亦或和，初始时，dp[i]=dp[i-1]，i是a[i]的最后位置，就从后往前找，看是否存在一个新区间，如果存在，dp[i]=max(dp[i],dp[j-1]+ans),j表示新区间的最左位置；

#include<bits/stdc++.h>
typedef long long ll;
using namespace std;
const int N=500000+10;
struct node{
    int x,y;
}a[N];
int v[N];
int dp[N];
int vis[N];
int main(){
    int n;
    //freopen("f://╤тед//data.txt","r",stdin);
    //freopen("f://╤тед//out1.txt","w",stdout);
    while(scanf("%d",&n)!=EOF){
        memset(dp,0,sizeof(dp));
        memset(a,0,sizeof(a));
        memset(vis,0,sizeof(vis));
        for(int i=1;i<=n;i++){
            scanf("%d",&v[i]);
            if(a[v[i]].x==0)a[v[i]].x=i;
            else
                a[v[i]].y=i;
        }
        for(int i=1;i<=n;i++){
            memset(vis,0,sizeof(vis));
            dp[i]=dp[i-1];
            int ans=0;
            int minn=i;
            for(int j=i;j>=1;j--){
                int t=v[j];
                if(vis[t]==0){
                    if(a[t].y>i)break;
                    minn=min(minn,a[t].x);
                    ans=ans^t;
                    vis[t]=1;
                }
                if(j<=minn)dp[i]=max(dp[i],dp[j-1]+ans);
            }
        }
        cout<<dp[n]<<endl;
    }
}