Problem Description
Mery has a beautiful necklace. The necklace is made up of N magic balls. Each ball has a beautiful value. The balls with the same beautiful value look the same, so if two or more balls have the same beautiful value, we just count it once. We define the beautiful value of some interval [x,y] as F(x,y). F(x,y) is calculated as the sum of the beautiful value from the xth ball to the yth ball and the same value is ONLY COUNTED ONCE. For example, if the necklace is 1 1 1 2 3 1, we have F(1,3)=1, F(2,4)=3, F(2,6)=6.

Now Mery thinks the necklace is too long. She plans to take some continuous part of the necklace to build a new one. She wants to know each of the beautiful value of M continuous parts of the necklace. She will give you M intervals [L,R] (1<=L<=R<=N) and you must tell her F(L,R) of them.
 

Input
The first line is T(T<=10), representing the number of test cases.
  For each case, the first line is a number N,1 <=N <=50000, indicating the number of the magic balls. The second line contains N non-negative integer numbers not greater 1000000, representing the beautiful value of the N balls. The third line has a number M, 1 <=M <=200000, meaning the nunber of the queries. Each of the next M lines contains L and R, the query.
 

Output
For each query, output a line contains an integer number, representing the result of the query.
 

Sample Input

261 2 3 4 3 531 23 52 661 1 1 2 3 531 12 43 5
 
Sample Output

3714136


思路:

树状数组对于维护数组有很大优势,复杂度O(M*lgN)。树状数组讲解

我们用树状数组维护题中的数组,但是他要求不能计算重复的点,这里用了一个很巧妙的方法:先将问题按照最右边界从小到大排序,然后对这个数组去重,最后我们记录当前问题的答案,再做下一个问题,这样我们就能保证每个问题都去重并且没有删去元素。

代码:

#include<cstdio>
#include<cstring>
#include<cstdlib>
#include<cctype>
#include<queue>
#include<cmath>
#include<string>
#include<stack>
#include<set>
#include<map>
#include<vector>
#include<iostream>
#include<algorithm>
#include<sstream>
#define ll long long
const int N=50005;
const int MAX=200005;
const int INF=1e9;
using namespace std;
map<int,int> mp;
int num[N],n;
ll ans[MAX],tree[N];
struct node{
int l,r,no;
}que[MAX];
int cmp(node a,node b){
if(a.r<b.r) return 1;
return 0;
} //树状数组操作-start
int lowbit(int x){
return x&(-x);
}
void update(int x,int val){ //更新
for(int i=x;i<=n;i+=lowbit(i)){
tree[i]+=val;
}
}
ll sum(int x){ //计算1~x的和
ll count=0;
for(int i=x;i>=1;i-=lowbit(i)){
count+=tree[i];
}
return count;
}
//树状数组操作-end int main(){
int T,m;
cin>>T;
while(T--){
memset(tree,0,sizeof(tree));
cin>>n;
for(int i=1;i<=n;i++){
scanf("%d",&num[i]);
}
cin>>m;
for(int i=1;i<=m;i++){
scanf("%d%d",&que[i].l,&que[i].r);
que[i].no=i; //表示这是第几个问题
}
sort(que+1,que+m+1,cmp);
mp.clear(); //map存储num[i]的最新位置
int pos=1;
for(int i=1;i<=m;i++){
while(pos<=que[i].r){
if(mp[num[pos]]!=0){ //如果重复出现
update(mp[num[pos]],-num[pos]); //删掉之前那个
}
mp[num[pos]]=pos;
update(pos,num[pos]);
pos++;
}
ans[que[i].no]=sum(que[i].r)-sum(que[i].l-1);
}
for(int i=1;i<=m;i++){
printf("%lld\n",ans[i]);
}
}
return 0;
}

hdu 3874(树状数组)题解的更多相关文章

  1. hdu 3874 树状数组

    思路:和求区间内有多少个不同的数一样,只不过改下权值. #include<iostream> #include<cstdio> #include<algorithm> ...

  2. hdu 4638 树状数组 区间内连续区间的个数(尽可能长)

    Group Time Limit: 4000/2000 MS (Java/Others)    Memory Limit: 32768/32768 K (Java/Others) Total Subm ...

  3. hdu 4777 树状数组+合数分解

    Rabbit Kingdom Time Limit: 6000/3000 MS (Java/Others)    Memory Limit: 32768/32768 K (Java/Others) T ...

  4. HDU 1934 树状数组 也可以用线段树

    http://acm.hdu.edu.cn/showproblem.php?pid=1394 或者是我自己挂的专题http://acm.hust.edu.cn/vjudge/contest/view. ...

  5. 【模板】HDU 1541 树状数组

    http://acm.hdu.edu.cn/showproblem.php?pid=1541 题意:给你一堆点,每个点右一个level,为其右下方所有点的数量之和,求各个level包含的点数. 题解: ...

  6. HDU 2852 (树状数组+无序第K小)

    题目链接:http://acm.hdu.edu.cn/showproblem.php?pid=2852 题目大意:操作①:往盒子里放一个数.操作②:从盒子里扔掉一个数.操作③:查询盒子里大于a的第K小 ...

  7. HDU 4911 (树状数组+逆序数)

    题目链接: http://acm.hdu.edu.cn/showproblem.php?pid=4911 题目大意:最多可以交换K次,就最小逆序对数 解题思路: 逆序数定理,当逆序对数大于0时,若ak ...

  8. hdu 5792(树状数组,容斥) World is Exploding

    hdu 5792 要找的无非就是一个上升的仅有两个的序列和一个下降的仅有两个的序列,按照容斥的思想,肯定就是所有的上升的乘以所有的下降的,然后再减去重复的情况. 先用树状数组求出lx[i](在第 i ...

  9. HDU 5775 树状数组

    Bubble Sort Time Limit: 2000/1000 MS (Java/Others)    Memory Limit: 65536/65536 K (Java/Others)Total ...

  10. 2018 CCPC网络赛 1010 hdu 6447 ( 树状数组优化dp)

    链接:http://acm.hdu.edu.cn/showproblem.php?pid=6447 思路:很容易推得dp转移公式:dp[i][j] = max(dp[i][j-1],dp[i-1][j ...

随机推荐

  1. CH0102 64位整数乘法 数论

    正解:数论/一个神仙想法 解题报告: 先放传送门qwq 两种方法,都还挺妙的就都写了qwq 第一种是快速幂 把b用二进制表示成,ck*2k+ck-1*2k-1+...+c0*20 然后就可以表示成,a ...

  2. jquery实现选项卡(两句即可实现)

    <!DOCTYPE HTML><html> <head>        <meta charset="utf-8">         ...

  3. isScroll的滚动组件的用法

    <div class="wrapper">  <ul>     <li>1</li>     <li>2</li& ...

  4. 虚拟机VMware的网络设置出了问题会导致很多莫名的错误

    邪门地CentOS内软件安装失败问题:Xshell与虚拟机的各种连接失败:CentOS下eth0没显示ip地址. 这些原因竟然是一个,虚拟机VMware的网络设置出了问题.     恢复初始设置即可.

  5. [LeetCode] 877. Stone Game == [LintCode] 396. Coins in a Line 3_hard tag: 区间Dynamic Programming, 博弈

    Alex and Lee play a game with piles of stones.  There are an even number of piles arranged in a row, ...

  6. 改善深层神经网络_优化算法_mini-batch梯度下降、指数加权平均、动量梯度下降、RMSprop、Adam优化、学习率衰减

    1.mini-batch梯度下降 在前面学习向量化时,知道了可以将训练样本横向堆叠,形成一个输入矩阵和对应的输出矩阵: 当数据量不是太大时,这样做当然会充分利用向量化的优点,一次训练中就可以将所有训练 ...

  7. keras之save & reload model

    import numpy as np np.random.seed(1337) # for reproducibility from keras.models import Sequential fr ...

  8. jmeter 正则表达式提取器的使用(提取第一个匹配结果)

    原文地址https://www.cnblogs.com/xueli/p/7405258.html?utm_source=itdadao&utm_medium=referral 正则表达式的用处 ...

  9. node初识——node中的require方法与require.js的区别

    出处:http://blog.csdn.net/u013613428/article/details/51966500 作为一个前端的新手,总是诧异于js的模块载入方式,看到了通过requireJs提 ...

  10. unity3d-射线(Ray)

    射线Ray 射线是一个点向另外一个点发生的一条线,一旦与其他模型发生碰撞,他将停止发射.注意这条件是逻辑上的,界面上看不到. 一般使用射线判断是否发射至某个游戏对象上或者获得鼠标点击的游戏对象等. 用 ...