本文版权归ljh2000和博客园共有,欢迎转载,但须保留此声明,并给出原文链接,谢谢合作。

本文作者:ljh2000
作者博客:http://www.cnblogs.com/ljh2000-jump/
转载请注明出处,侵权必究,保留最终解释权!

Description

All our characters have hobbies. The same is true for Fedor. He enjoys shopping in the neighboring supermarket.

The goods in the supermarket have unique integer ids. Also, for every integer there is a product with id equal to this integer. Fedor has ndiscount coupons, the i-th of them can be used with products with ids ranging from li to ri, inclusive. Today Fedor wants to take exactly kcoupons with him.

Fedor wants to choose the k coupons in such a way that the number of such products x that all coupons can be used with this product x is as large as possible (for better understanding, see examples). Fedor wants to save his time as well, so he asks you to choose coupons for him. Help Fedor!

Input

The first line contains two integers n and k (1 ≤ k ≤ n ≤ 3·105) — the number of coupons Fedor has, and the number of coupons he wants to choose.

Each of the next n lines contains two integers li and ri ( - 109 ≤ li ≤ ri ≤ 109) — the description of the i-th coupon. The coupons can be equal.

Output

In the first line print single integer — the maximum number of products with which all the chosen coupons can be used. The products with which at least one coupon cannot be used shouldn't be counted.

In the second line print k distinct integers p1, p2, ..., pk (1 ≤ pi ≤ n) — the ids of the coupons which Fedor should choose.

If there are multiple answers, print any of them.

Examples
input
4 2
1 100
40 70
120 130
125 180
output
31
1 2
input
3 2
1 12
15 20
25 30
output
0
1 2
input
5 2
1 10
5 15
14 50
30 70
99 100
output
21
3 4
Note

In the first example if we take the first two coupons then all the products with ids in range [40, 70] can be bought with both coupons. There are 31 products in total.

In the second example, no product can be bought with two coupons, that is why the answer is 0. Fedor can choose any two coupons in this example.

正解:堆+贪心

解题报告:

  这道题概括出来的模型十分简洁经典:从n条线段中取出恰好k条使得交集长度尽可能长,输出最优值和方案。

  我开始想了很久的单调性,但是并不能实现单调决策,更不能还原历史版本。所以我就想了想,似乎带个log就很可做了?

  考虑先按左端点排序,维护一个右端点坐标的小根堆,那么很容易发现我只需要保证堆的大小始终小于等于k即可。当我每次扫到一个左端点时,将其右端点与堆顶作比较,如果比堆顶小则不作考虑,否则,删除堆顶,把这个新的右端点坐标加入堆中。每次只需用堆顶减去当前处理的线段的左端点来更新答案(当且仅当堆中恰好有k个元素)。输出方案的话,用同样方法再做一次即可。

//It is made by ljh2000
#include <iostream>
#include <cstdlib>
#include <cstring>
#include <cstdio>
#include <cmath>
#include <algorithm>
#include <ctime>
#include <vector>
#include <queue>
#include <map>
#include <set>
#include <string>
using namespace std;
typedef long long LL;
const int MAXN = 300011;
int n,k,ans,dui[MAXN];
struct node{int pos,id; inline bool operator < (const node &a)const{ return a.pos<pos; } }tmp;
priority_queue<node>Q;
struct seq{int l,r,id;}a[MAXN];
inline bool cmp(seq q,seq qq){ return q.l<qq.l; }
inline int getint(){
int w=0,q=0; char c=getchar(); while((c<'0'||c>'9') && c!='-') c=getchar();
if(c=='-') q=1,c=getchar(); while (c>='0'&&c<='9') w=w*10+c-'0',c=getchar(); return q?-w:w;
} inline void work(){
n=getint(); k=getint(); for(int i=1;i<=n;i++) a[i].l=getint(),a[i].r=getint(),a[i].id=i;
sort(a+1,a+n+1,cmp); ans=-1;//!!!
for(int i=1;i<=n;i++) {
if(!Q.empty())tmp=Q.top();
if((int)Q.size()<k) {
tmp.pos=a[i].r;
tmp.id=i;
Q.push(tmp);
}
else {
if(a[i].r>tmp.pos) {
Q.pop();
tmp.pos=a[i].r;
tmp.id=i;
Q.push(tmp);
}
}
if((int)Q.size()>=k) ans=max(Q.top().pos-a[i].l,ans);
}
printf("%d\n",ans+1);
if(ans==-1) { for(int i=1;i<=k;i++) printf("%d ",i); return ; } int lans=ans; ans=-1;
while(!Q.empty()) Q.pop();
for(int i=1;i<=n;i++) {
if(!Q.empty()) tmp=Q.top();
if((int)Q.size()<k) {
tmp.pos=a[i].r;
tmp.id=a[i].id;//!!!
Q.push(tmp);
}
else {
if(a[i].r>tmp.pos) {
Q.pop();
tmp.pos=a[i].r;
tmp.id=a[i].id;//!!!
Q.push(tmp);
}
}
if((int)Q.size()>=k) {
ans=max(Q.top().pos-a[i].l,ans);
if(ans==lans) {
int cnt=0;
while(!Q.empty()) {
tmp=Q.top();
dui[++cnt]=tmp.id;
Q.pop();
}
sort(dui+1,dui+k+1);
for(int i=1;i<=k;i++) printf("%d ",dui[i]);
return ;
}
}
}
} int main()
{
work();
return 0;
}

  

codeforces754D Fedor and coupons的更多相关文章

  1. codeforces 754D. Fedor and coupons

    D. Fedor and coupons time limit per test 4 seconds memory limit per test 256 megabytes input standar ...

  2. Codeforces 390Div2-754D. Fedor and coupons(贪心+优先队列)

    D. Fedor and coupons time limit per test 4 seconds memory limit per test 256 megabytes input standar ...

  3. CodeForces 754D Fedor and coupons&&CodeForces 822C Hacker, pack your bags!

    D. Fedor and coupons time limit per test 4 seconds memory limit per test 256 megabytes input standar ...

  4. 【codeforces 754D】Fedor and coupons

    time limit per test4 seconds memory limit per test256 megabytes inputstandard input outputstandard o ...

  5. CodeForces 754D Fedor and coupons (优先队列)

    题意:给定n个优惠券,每张都有一定的优惠区间,然后要选k张,保证k张共同的优惠区间最大. 析:先把所有的优惠券按左端点排序,然后维护一个容量为k的优先队列,每次更新优先队列中的最小值,和当前的右端点, ...

  6. Codeforces Round #390 (Div. 2) D. Fedor and coupons(区间最大交集+优先队列)

    http://codeforces.com/contest/754/problem/D 题意: 给定几组区间,找k组区间,使得它们的公共交集最大. 思路: 在k组区间中,它们的公共交集=k组区间中右端 ...

  7. D. Fedor and coupons 二分暴力

    http://codeforces.com/contest/754/problem/D 给出n条线段,选出k条,使得他们的公共部分长度最大. 公共部分的长度,可以二分出来,为val.那么怎么判断有k条 ...

  8. Codeforces Round #390 (Div. 2) D. Fedor and coupons

    题意:题目简化了就是要给你n个区间,然后让你选出k个区间  使得这k个区间有公共交集:问这个公共交集最大能是多少,并且输出所选的k个区间.如果有多组答案,则输出任意一种.   这题是用优先队列来处理区 ...

  9. CodeForces 754D Fedor and coupons ——(k段线段最大交集)

    还记得lyf说过k=2的方法,但是推广到k是其他的话有点麻烦.现在这里采取另外一种方法. 先将所有线段按照L进行排序,然后优先队列保存R的值,然后每次用最小的R值,和当前的L来维护答案即可.同时,如果 ...

随机推荐

  1. SpringMVC + Spring + MyBatis 整合 + Spring shrio + easyUI + 权限管理框架,带shrio session和shrio cache集群实现方案

    工作之余先来写了一个不算规范的简单架子 基于spring mvc + spring + mybatis + Spring shrio 基于redis的集群方案 系统权限部分,分成多个机构,其中每个机构 ...

  2. .net FTP上传文件

    FTP上传文件代码实现: private void UploadFileByWebClient() { WebClient webClient = new WebClient(); webClient ...

  3. jQuery为开发插件提拱了两个方法:jQuery.fn.extend(); jQuery.extend();

    jQuery为开发插件提拱了两个方法,分别是: jQuery.fn.extend(); jQuery.extend(); jQuery.fn jQuery.fn = jQuery.prototype ...

  4. Android 绘制view的小知识点

    [onMeasure] 直接继承view或ViewGroup的自定义控件需要重写onMeasure方法并设置wrap_content时的自身大小,否则在布局中使用wrap_content就相当于mat ...

  5. PyCharm断点调试django

    我在用PyCharm开发django程序的时候,对于打印日志调试程序的方式感觉还是有点麻烦和不直观,所以研究了一下断点调试的方法如下: 1.打开你的工程,在菜单栏里找到Run-->Edit Co ...

  6. iOS平台UDID方案比较

    苹果在iOS6中禁用了[UIDevice uniqueIdentifier],在iOS7中又把mac地址的获取给堵上了.没办法,毕竟人家是老大,说不让你用,你也没办法.在这边总结一下现有的一部分UDI ...

  7. swift学习笔记2——函数、闭包

    之前学习swift时的个人笔记,根据github:the-swift-programming-language-in-chinese学习.总结,将重要的内容提取,加以理解后整理为学习笔记,方便以后查询 ...

  8. 7个高性能JavaScript代码高亮插件

    本文由码农网 – 小峰原创,转载请看清文末的转载要求,欢迎参与我们的付费投稿计划! 对于喜欢写技术博客的同学来说,一定对代码高亮组件非常熟悉.一款优秀的JavaScript代码高亮插件,将会帮助你渲染 ...

  9. 办公OA的登陆界面..

    <%@ page language="java" contentType="text/html; charset=UTF-8" pageEncoding= ...

  10. 【msql】关于redo 和 undo log

    InnoDB 有两块非常重要的日志,一个是undo log,另外一个是redo log,前者用来保证事务的原子性以及InnoDB的MVCC,后者用来保证事务的持久性.和大多数关系型数据库一样,Inno ...