Bike loves looking for the second maximum element in the sequence. The second maximum element in the sequence of distinct numbers x1, x2, ..., xk (k > 1) is such maximum element xj, that the following inequality holds: .
The lucky number of the sequence of distinct positive integers x1, x2, ..., xk (k > 1) is the number that is equal to the bitwise excluding OR of the maximum element of the sequence and the second maximum element of the sequence.
You've got a sequence of distinct positive integers s1, s2, ..., sn (n > 1). Let's denote sequence sl, sl + 1, ..., sr as s[l..r] (1 ≤ l < r ≤ n). Your task is to find the maximum number among all lucky numbers of sequences s[l..r].
Note that as all numbers in sequence s are distinct, all the given definitions make sence.
Input
The first line contains integer n (1 < n ≤ 105). The second line contains n distinct integers s1, s2, ..., sn (1 ≤ si ≤ 109).
Output
Print a single integer — the maximum lucky number among all lucky numbers of sequences s[l..r].
Examples
5
5 2 1 4 3
7
5
9 8 3 5 7
15
Note
For the first sample you can choose s[4..5] = {4, 3} and its lucky number is (4 xor 3) = 7. You can also choose s[1..2].
For the second sample you must choose s[2..5] = {8, 3, 5, 7}.
求任意区间最大值，用单调递减栈
要明确 一个元素最多只有两次当次大值的机会（忽略同一个最大值的不同区间 （左边一个最大值 右边一个最大值））
可以很容易的想到当前元素在哪个区间是最大值
设删除元素的下标为id 当前元素为i
那么a[id]即为a[id, i]这个区间的次大值
如果向左拓展时删除的最后一个元素的下标为id
那么a[id - 1]即为 a[i]左边的最大值
也就是a[i]为a[id - 1]右边的次大值
对于a[i]它左边的次大值我们都取了一遍
右边的次大值会乖乖的找到它
就像a[i]找到a[id - 1]一样
想一想 是不是这样
扩展：
那么我们通过这个思路就可以求出每个元素为次大值的区间
那么第三大 第四大·····
#include <iostream> #include <cstdio> #include <sstream> #include <cstring> #include <map> #include <cctype> #include <set> #include <vector> #include <stack> #include <queue> #include <algorithm> #include <list> #include <cmath> #include <bitset> #define rap(i, a, n) for(int i=a; i<=n; i++) #define rep(i, a, n) for(int i=a; i<n; i++) #define lap(i, a, n) for(int i=n; i>=a; i--) #define lep(i, a, n) for(int i=n; i>a; i--) #define rd(a) scanf("%d", &a) #define rlld(a) scanf("%lld", &a) #define rc(a) scanf("%c", &a) #define rs(a) scanf("%s", a) #define rb(a) scanf("%lf", &a) #define rf(a) scanf("%f", &a) #define pd(a) printf("%d ", a) #define plld(a) printf("%lld ", a) #define pc(a) printf("%c ", a) #define ps(a) printf("%s ", a) #define MOD 2018 #define LL long long #define ULL unsigned long long #define Pair pair<int, int> #define mem(a, b) memset(a, b, sizeof(a)) #define _ ios_base::sync_with_stdio(0),cin.tie(0) //freopen("1.txt", "r", stdin); using namespace std; const int maxn = 110000, INF = 0x7fffffff; stack<LL> S; int main() { int n; LL tmp; LL mx = -INF; rd(n); for(int i = 1; i <= n; i++) { rlld(tmp); while(!S.empty() && S.top() < tmp) { mx = max(mx, S.top() ^ tmp); S.pop(); } if(!S.empty()) mx = max(mx, tmp ^ S.top()); S.push(tmp); } cout << mx << endl; return 0; }
Bike loves looking for the second maximum element in the sequence. The second maximum element in the sequence of distinct numbers x1, x2, ..., xk (k > 1) is such maximum element xj, that the following inequality holds: .
The lucky number of the sequence of distinct positive integers x1, x2, ..., xk (k > 1) is the number that is equal to the bitwise excluding OR of the maximum element of the sequence and the second maximum element of the sequence.
You've got a sequence of distinct positive integers s1, s2, ..., sn (n > 1). Let's denote sequence sl, sl + 1, ..., sr as s[l..r] (1 ≤ l < r ≤ n). Your task is to find the maximum number among all lucky numbers of sequences s[l..r].
Note that as all numbers in sequence s are distinct, all the given definitions make sence.
Input
The first line contains integer n (1 < n ≤ 105). The second line contains n distinct integers s1, s2, ..., sn (1 ≤ si ≤ 109).
Output
Print a single integer — the maximum lucky number among all lucky numbers of sequences s[l..r].
Examples
5
5 2 1 4 3
7
5
9 8 3 5 7
15
Note
For the first sample you can choose s[4..5] = {4, 3} and its lucky number is (4 xor 3) = 7. You can also choose s[1..2].
For the second sample you must choose s[2..5] = {8, 3, 5, 7}.