Boring Partition（CF-239D）

Problem Description

This problem is the most boring one you've ever seen.

Given a sequence of integers a1, a2, ..., an and a non-negative integer h, our goal is to partition the sequence into two subsequences (not necessarily consist of continuous elements). Each element of the original sequence should be contained in exactly one of the result subsequences. Note, that one of the result subsequences can be empty.

Let's define function f(ai, aj) on pairs of distinct elements (that is i ≠ j) in the original sequence. If ai and aj are in the same subsequence in the current partition then f(ai, aj) = ai + aj otherwise f(ai, aj) = ai + aj + h.

Consider all possible values of the function f for some partition. We'll call the goodness of this partiotion the difference between the maximum value of function f and the minimum value of function f.

Your task is to find a partition of the given sequence a that have the minimal possible goodness among all possible partitions.

Input

The first line of input contains integers n and h (2 ≤ n ≤ 105, 0 ≤ h ≤ 108). In the second line there is a list of n space-separated integers representing a1, a2, ..., an (0 ≤ ai ≤ 108).

Output

The first line of output should contain the required minimum goodness.

The second line describes the optimal partition. You should print n whitespace-separated integers in the second line. The i-th integer is 1 if ai is in the first subsequence otherwise it should be 2.

If there are several possible correct answers you are allowed to print any of them.

Examples

Input

3 2
1 2 3

Output

1
1 2 2 

Input

5 10
0 1 0 2 1

Output

3
2 2 2 2 2 

Note

In the first sample the values of f are as follows: f(1, 2) = 1 + 2 + 2 = 5, f(1, 3) = 1 + 3 + 2 = 6 and f(2, 3) = 2 + 3 = 5. So the difference between maximum and minimum values of f is 1.

In the second sample the value of h is large, so it's better for one of the sub-sequences to be empty.

题意：给出 n 个数和一个数 h，现在要将 n 个数分成两个序列，定义一个函数 f(ai,aj)，当 ai、aj 处于同一序列时，f(ai,aj)=ai+a[j，当 ai、aj 处于不同序列时，f(ai,aj)=ai+aj+h，现要使得函数 f 的最大值与最小值的差最小，求最小值

思路：

要使得最大值与最小值的差最小，那么就要让最大值尽可能的小，最小值尽可能的大

首先将这 n 个数从小到大进行排序，在排序后可以发现，若将大的数移动到另一序列，只能使得最大值更大，无法使最小值更大，而将最小的数移动到另一序列，不仅使得最大值变小了，也使得最小值变大了

那么问题一共有两种分配方案：

• 当只有一个序列时
• 一个集合序列中为最小的元素 min 时，一个集合序列为剩下的元素时

因此比较两种分配方案哪个的差最小即可

Source Program

#include<iostream>
#include<cstdio>
#include<cstdlib>
#include<string>
#include<cstring>
#include<cmath>
#include<ctime>
#include<algorithm>
#include<utility>
#include<stack>
#include<queue>
#include<vector>
#include<set>
#include<map>
#include<bitset>
#define EPS 1e-9
#define PI acos(-1.0)
#define INF 0x3f3f3f3f
#define LL long long
const int MOD = 1E9+7;
const int N = 1000000+5;
const int dx[] = {-1,1,0,0,-1,-1,1,1};
const int dy[] = {0,0,-1,1,-1,1,-1,1};
using namespace std;

struct Node{
    int val;
    int pos;
    Node(){}
    Node(int val,int pos):val(val),pos(pos){}
    bool operator < (const Node &rhs)const{
        return val<rhs.val;
    }
}node[N];
int main() {
    int n,h;
    scanf("%d%d",&n,&h);
    for(int i=1;i<=n;i++){
        scanf("%d",&node[i].val);
        node[i].pos=i;
    }
    sort(node+1,node+1+n);
    int sameMaxx=node[n].val+node[n-1].val;
    int sameMinn=node[1].val+node[2].val;
    int sameSub=sameMaxx-sameMinn;

    int unsameMaxx=max(node[n].val+node[1].val+h,node[n].val+node[n-1].val);
    int unsameMinn=min(node[1].val+node[2].val+h,node[2].val+node[3].val);
    int unsameSub=unsameMaxx-unsameMinn;


    int minn=min(abs(sameSub),abs(unsameSub));
    printf("%d\n",minn);

    int pos=node[1].pos;
    for(int i=1;i<=n;i++){
        if(i==pos&&abs(unsameSub)<abs(sameSub))
            printf("1 ");
        else
            printf("2 ");
    }
    printf("\n");

    return 0;
}