After the piece of a devilish mirror hit the Kay's eye, he is no longer interested in the beauty of the roses. Now he likes to watch snowflakes.

Once upon a time, he found a huge snowflake that has a form of the tree (connected acyclic graph) consisting of n nodes. The root of tree has index 1. Kay is very interested in the structure of this tree.

After doing some research he formed q queries he is interested in. The i-th query asks to find a centroid of the subtree of the node vi. Your goal is to answer all queries.

Subtree of a node is a part of tree consisting of this node and all it's descendants (direct or not). In other words, subtree of node v is formed by nodes u, such that node v is present on the path from u to root.

Centroid of a tree (or a subtree) is a node, such that if we erase it from the tree, the maximum size of the connected component will be at least two times smaller than the size of the initial tree (or a subtree).

Input

The first line of the input contains two integers n and q (2 ≤ n ≤ 300 000, 1 ≤ q ≤ 300 000) — the size of the initial tree and the number of queries respectively.

The second line contains n - 1 integer p2, p3, ..., pn (1 ≤ pi ≤ n) — the indices of the parents of the nodes from 2 to n. Node 1 is a root of the tree. It's guaranteed that pi define a correct tree.

Each of the following q lines contain a single integer vi (1 ≤ vi ≤ n) — the index of the node, that define the subtree, for which we want to find a centroid.

Output

For each query print the index of a centroid of the corresponding subtree. If there are many suitable nodes, print any of them. It's guaranteed, that each subtree has at least one centroid.

Example

Input

7 4

1 1 3 3 5 3

1

2

3

5

Output

3

2

3

6

Note

The first query asks for a centroid of the whole tree — this is node 3. If we delete node 3 the tree will split in four components, two of size 1 and two of size 2.

The subtree of the second node consists of this node only, so the answer is 2.

Node 3 is centroid of its own subtree.

The centroids of the subtree of the node 5 are nodes 5 and 6 — both answers are considered correct.

题意:

给你一个含有n个节点的树,并且给你q个询问,每一个询问x,问以x为根的子树中重心是哪个节点?

思路:

利用重心的2个性质:

1、对于一棵树来说,删去该树的重心后,所有的子树的大小不会超过原树大小的二分之一

2、两棵树合并后其新的重心在原两个重心的路径中。

那么我们在dfs树的过程中维护cntson[i]代表以i节点为根的子树大小,

然后在利用cntson可以找到一个子树的重心,在一个节点与其儿子节点为根的子树合并时,在当前维护的重心和儿子节点为根的子树的重心的路径中同样通过性质1来更新出当前的重心即可。

细节见代码:

#include <iostream>
#include <cstdio>
#include <cstring>
#include <algorithm>
#include <cmath>
#include <queue>
#include <stack>
#include <map>
#include <set>
#include <vector>
#include <iomanip>
#define ALL(x) (x).begin(), (x).end()
#define sz(a) int(a.size())
#define all(a) a.begin(), a.end()
#define rep(i,x,n) for(int i=x;i<n;i++)
#define repd(i,x,n) for(int i=x;i<=n;i++)
#define pii pair<int,int>
#define pll pair<long long ,long long>
#define gbtb ios::sync_with_stdio(false),cin.tie(0),cout.tie(0)
#define MS0(X) memset((X), 0, sizeof((X)))
#define MSC0(X) memset((X), '\0', sizeof((X)))
#define pb push_back
#define mp make_pair
#define fi first
#define se second
#define eps 1e-6
#define gg(x) getInt(&x)
#define chu(x) cout<<"["<<#x<<" "<<(x)<<"]"<<endl
using namespace std;
typedef long long ll;
ll gcd(ll a, ll b) {return b ? gcd(b, a % b) : a;}
ll lcm(ll a, ll b) {return a / gcd(a, b) * b;}
ll powmod(ll a, ll b, ll MOD) {ll ans = 1; while (b) {if (b % 2)ans = ans * a % MOD; a = a * a % MOD; b /= 2;} return ans;}
inline void getInt(int* p);
const int maxn = 300010;
const int inf = 0x3f3f3f3f;
/*** TEMPLATE CODE * * STARTS HERE ***/
int n;
std::vector<int> son[maxn];
int cntson[maxn];
int q;
int ans[maxn];
int f[maxn];
void dfs(int x, int pre)
{
cntson[x] = 1;
ans[x] = x;
for (auto y : son[x])
{
if (y != pre)
{
dfs(y, x);
cntson[x] += cntson[y];
}
}
for (auto y : son[x])
{
if (y != pre)
{
if (cntson[y] * 2 > cntson[x])
{
ans[x] = ans[y];
}
}
} while ((cntson[x] - cntson[ans[x]]) * 2 > cntson[x])
{
ans[x] = f[ans[x]];
} }
// int num; int main()
{
//freopen("D:\\common_text\\code_stream\\in.txt","r",stdin);
//freopen("D:\\common_text\\code_stream\\out.txt","w",stdout);
gg(n);
gg(q);
repd(i, 2, n)
{
int x;
gg(x);
son[x].push_back(i);
f[i] = x;
}
dfs(1, 1);
int x;
repd(i, 1, q)
{
gg(x);
printf("%d\n", ans[x] );
}
return 0;
} inline void getInt(int* p) {
char ch;
do {
ch = getchar();
} while (ch == ' ' || ch == '\n');
if (ch == '-') {
*p = -(getchar() - '0');
while ((ch = getchar()) >= '0' && ch <= '9') {
*p = *p * 10 - ch + '0';
}
}
else {
*p = ch - '0';
while ((ch = getchar()) >= '0' && ch <= '9') {
*p = *p * 10 + ch - '0';
}
}
}

Kay and Snowflake CodeForces - 686D (树的重心性质)的更多相关文章

  1. Kay and Snowflake CodeForces - 686D

    Kay and Snowflake CodeForces - 686D 题意:给一棵有根树,有很多查询(100000级别的),查询是求以任意一点为根的子树的任意重心. 方法很多,但是我一个都不会 重心 ...

  2. Kay and Snowflake CodeForces - 685B (重心, 好题)

    大意:给定有根树, 求每个子树的重心 我太菜了啊, 只能想到暴力树剖, 然而这就是个B题, 感觉树剖+线段树二分还是挺难写的..... 看了题解发现重心一定在重儿子与根的树链上, 重心最多上跳n-1次 ...

  3. Codeforces 686 D - Kay and Snowflake

    D - Kay and Snowflake 思路: 树的重心 利用重心的一个推论,树的重心必定在子树重心的连线上. 然后利用重心的性质,可知,如果有一颗子树的大小超过整棵树的大小的1/2,那么树的重心 ...

  4. poj1655 Balancing Act 找树的重心

    http://poj.org/problem? id=1655 Balancing Act Time Limit: 1000MS   Memory Limit: 65536K Total Submis ...

  5. 洛谷P4299 首都(BZOJ3510)(LCT,树的重心,二分查找)

    Update:原来的洛谷U21715已成坑qwq 已经被某位管理员巨佬放进公共题库啦!又可以多一个AC记录啦! 洛谷题目传送门 其实也可以到这里交啦 思路分析 动态维护树的重心 题目中说到国家的首都会 ...

  6. Codeforces Round #359 (Div. 2) D. Kay and Snowflake 树的重心

    题目链接: 题目 D. Kay and Snowflake time limit per test 3 seconds memory limit per test 256 megabytes inpu ...

  7. codeforces 685B Kay and Snowflake 树的重心

    分析:就是找到以每个节点为根节点的树的重心 树的重心可以看这三篇文章: 1:http://wenku.baidu.com/link?url=yc-3QD55hbCaRYEGsF2fPpXYg-iO63 ...

  8. codeforces 686D D. Kay and Snowflake(dfs)

    题目链接: D. Kay and Snowflake time limit per test 3 seconds memory limit per test 256 megabytes input s ...

  9. Codeforces Round #359 (Div. 2) D. Kay and Snowflake 树DP

    D. Kay and Snowflake     After the piece of a devilish mirror hit the Kay's eye, he is no longer int ...

随机推荐

  1. Python浮点型数据小数点的取舍

    python默认的是17位小数的精度 1.round()内置方法 π=3.1415926535 new_num=round(π,2)     #四舍五入保留两位小数 print(new_num)    ...

  2. java8:(Lambda 表达式,Supplier,@FunctionalInterface,foreach(),Optional,Stream().collect,双冒号,joining,partitioningBy分区,collectingAndThen,filter())

    1.Lambda 表达式: 引导:http://www.cnblogs.com/yulinfeng/p/8452379.html DEMO1: List<String> names1 = ...

  3. 积累-T

    emmm,各种知识点都有吧,主要方便自己记 随机修改网页图标 <script> var image=new Array(3); image.length=3; image[1]=" ...

  4. 描述下fastJSON,jackson等等的技术

    ①Jackson:依赖的jar包较少,简单易用性能高,更新速度也比较快,但是对于复杂类型的json转换bean会出 现问题,一些集合Map,List的转换出现问题,对于复杂类型的bean转换Json, ...

  5. python(递归实例)

    摘要:在学习python递归知识点时,总是一知半解,似懂非懂的..在反复看视频翻资料同时,也收集案例来分析求证..通过分析下面几个案例希望能有所帮助!!! 1.用递归的方法实现阶乘... def nu ...

  6. 【机器学习】深入理解拉格朗日乘子法(Lagrange Multiplier) 和KKT条件

    在求取有约束条件的优化问题时,拉格朗日乘子法(Lagrange Multiplier) 和KKT条件是非常重要的两个求取方法,对于等式约束的优化问题,可以应用拉格朗日乘子法去求取最优值:如果含有不等式 ...

  7. Flume下载安装

    下载 可以apache官网下载flume的安装包 下载时注意,flume具有两个版本,0.9.x和1.x,两个版本并不兼容,我们用最新的1.x版本,也叫flume-ng版本. 安装 解压到指定目录即可 ...

  8. query函数的可查询数据

    #encoding: utf-8 from sqlalchemy import create_engine,Column,Integer,String,Float,func from sqlalche ...

  9. [转帖]Docker常用命令总结

    Docker常用命令总结 http://www.ha97.com/5546.html 发表于: Linux, 互联网, 虚拟化与云计算 | 作者: 博客教主 标签: docker,常用命令,总结 PS ...

  10. ASP.NET Core中使用Autofac进行属性注入

    一些无关紧要的废话: 作为一名双修程序员(自封的),喜欢那种使用Spring的注解形式进行依赖注入或者Unity的特性形式进行依赖注入,当然,形式大同小异,但结果都是一样的,通过属性进行依赖注入. A ...