You have been offered a job in a company developing a large social network. Your first task is connected with searching profiles that most probably belong to the same user.

The social network contains n registered profiles, numbered from 1 to n. Some pairs there are friends (the "friendship" relationship is mutual, that is, if i is friends with j, then j is also friends with i). Let's say that profiles i and j (i ≠ j) are doubles, if for any profile k (k ≠ ik ≠ j) one of the two statements is true: either k is friends with i and j, or k isn't friends with either of them. Also, i and j can be friends or not be friends.

Your task is to count the number of different unordered pairs (i, j), such that the profiles i and j are doubles. Note that the pairs are unordered, that is, pairs (a, b) and (b, a) are considered identical.

Input

The first line contains two space-separated integers n and m (1 ≤ n ≤ 106, 0 ≤ m ≤ 106), — the number of profiles and the number of pairs of friends, correspondingly.

Next m lines contains descriptions of pairs of friends in the format "v u", where v and u (1 ≤ v, u ≤ n, v ≠ u) are numbers of profiles that are friends with each other. It is guaranteed that each unordered pair of friends occurs no more than once and no profile is friends with itself.

Output

Print the single integer — the number of unordered pairs of profiles that are doubles.

Please do not use the %lld specificator to read or write 64-bit integers in С++. It is preferred to use the %I64d specificator.

Examples

Input
3 3
1 2
2 3
1 3
Output
3
Input
3 0
Output
3
Input
4 1
1 3
Output
2

Note

In the first and second sample any two profiles are doubles.

In the third sample the doubles are pairs of profiles (1, 3) and (2, 4).

题意:给定无向图,问有多少对点对,在不考虑它们之间的连边情况下,满足它们的直接连边的情况相同。

思路:用hash保存每个点的连边情况,然后排序得到hash值,相同的一并处理。

有两类:第一类是点对间有边,那么我们需要向自己加一条边,那么它们的连边情况是相同的。

第二类是点对间无边,直接处理即可。

#include<bits/stdc++.h>
#define rep(i,a,b) for(int i=a;i<=b;i++)
#define uint unsigned int
using namespace std;
const int maxn=;
const int seed=;
const int seed2=;
uint p[maxn],p2[maxn];long long ans;
pair<uint,uint>h[maxn];
int main()
{
int N,M,u,v;
scanf("%d%d",&N,&M); p[]=; p2[]=;
rep(i,,N) p[i]=p[i-]*seed;
rep(i,,N) p2[i]=p2[i-]*seed2;
rep(i,,N) h[i].first=p[i],h[i].second=p2[i];
rep(i,,M){
scanf("%d%d",&u,&v);
h[u].first+=p[v]; h[v].first+=p[u];
h[u+N].first+=p[v]; h[v+N].first+=p[u];
h[u].second+=p2[v]; h[v].second+=p2[u];
h[u+N].second+=p2[v]; h[v+N].second+=p2[u];
}
sort(h+,h+N+N+);
for(int i=;i<=N+N;i++){
int j=i;
while(j+<=N+N&&h[j+]==h[i]) j++;
ans+=(long long)(j-i+)*(j-i)/;
i=j;
}
printf("%I64d\n",ans);
return ;
}

CodeForces - 154C:Double Profiles (hash+排序)的更多相关文章

  1. Codeforces Round #109 (Div. 2) E. Double Profiles hash

    题目链接: http://codeforces.com/problemset/problem/155/E E. Double Profiles time limit per test 3 second ...

  2. xtu summer individual 2 E - Double Profiles

    Double Profiles Time Limit: 3000ms Memory Limit: 262144KB This problem will be judged on CodeForces. ...

  3. codeforces Round #440 A Search for Pretty Integers【hash/排序】

    A. Search for Pretty Integers [题目链接]:http://codeforces.com/contest/872/problem/A time limit per test ...

  4. sort 树 hash 排序

    STL 中 sort 函数用法简介 做 ACM 题的时候,排序是一种经常要用到的操作.如果每次都自己写个冒泡之类的 O(n^2) 排序,不但程序容易超时,而且浪费宝贵的比赛时间,还很有可能写错. ST ...

  5. [置顶] Codeforces 70D 动态凸包 (极角排序 or 水平序)

    题目链接:http://codeforces.com/problemset/problem/70/D 本题关键:在log(n)的复杂度内判断点在凸包 或 把点插入凸包 判断:平衡树log(n)内选出点 ...

  6. DataTable 中varchar 转换为 Double 后重新 排序。

    DataTable  查询出某个字段为varchar 类型的.不过里面存的为数字,需要进行排序.可是如果直接排序就会不对.因为为varchar类型的,需要转换一下. 方法一: dt.Columns.A ...

  7. ruby中Hash排序

    当values都是整形时,按照Hash的Values排序: h = {'a'=>1,'b'=>2,'c'=>5,'d'=>4} h.sort {|a,b| a[1]<=& ...

  8. Codeforces 484B Maximum Value(排序+二分)

    题目链接: http://codeforces.com/problemset/problem/484/B 题意: 求a[i]%a[j] (a[i]>a[j])的余数的最大值 分析: 要求余数的最 ...

  9. Codeforces 825E Minimal Labels - 拓扑排序 - 贪心

    You are given a directed acyclic graph with n vertices and m edges. There are no self-loops or multi ...

随机推荐

  1. <script>放在head内和body内有什么区别

    加载的顺序不一样,你可以把HTML看成从上往下加载的. 例如在网速慢的情况下把js代码放在body底部用户会先看到网页结构,等js加载完成后才出现特效 区别简述: 在HTML body部分中的Java ...

  2. JavaScript笔记02——基本语法(包括函数、对象、数组等)

    Doing Math & Logic Conditional & Looping Functions Objects Arrays Doing Math & Logic 1.J ...

  3. Java中系统时间的获取_currentTimeMillis()函数应用解读

    快速解读 System.currentTimeMillis()+time*1000) 的含义 一.时间的单位转换 1秒=1000毫秒(ms) 1毫秒=1/1,000秒(s)1秒=1,000,000 微 ...

  4. Java final static关键字

    Java中的final关键字 使用 final 关键字做标识有 “最终的” 含义. final 可以修饰 类.方法.属性.变量 final 修饰类: 则该类不允许被继承 final 修饰方法:则该方法 ...

  5. vRO 7 添加RestHost证书报错

    报错类似的错误: Cannot execute request: ; java.security.cert.CertificateException: Certificates does not co ...

  6. poj 2762 Going from u to v or from v to u?【强连通分量缩点+拓扑排序】

    Going from u to v or from v to u? Time Limit: 2000MS   Memory Limit: 65536K Total Submissions: 15812 ...

  7. Struts2全局异常处理

    1.在struts.xml中配置全局异常处理 在Action中抛出异常,此异常可以是action自己抛的,也可以是Service抛出来的,都会跳转到全局异常中,只有在当前Action中配置域全局异常返 ...

  8. R语言笔记006——分组获取描述性统计量

    方法一:使用aggregate()分组获取描述性统计量 aggregate(mtcars[vars],by=list(am=mtcars$am),mean) aggregate(mtcars[vars ...

  9. HYSBZ 1500 维修数列(伸展树模板)

    题意: 题解:典型伸展树的题,比较全面. 我理解的伸展树: 1 伸展操作:就是旋转,因为我们只需保证二叉树中序遍历的结果不变,所以我们可以旋转来保持树的平衡,且旋转有左旋与右旋.通过这种方式保证不会让 ...

  10. Codeforces Round #425 (Div. 2) D 树链剖分 + 树状数组维护区间

    一看就知道 可以LCA判断做 也可以树链剖分拿头暴力 然而快速读入和线段树维护区间会T70 于是只能LCA? 线段树的常数不小 于是需要另外一种办法来进行区间加减和查询区间和 就是使用树状数组 这个题 ...