QOJ.ac
QOJ
ID | Problem | Submitter | Result | Time | Memory | Language | File size | Submit time | Judge time |
---|---|---|---|---|---|---|---|---|---|
#327670 | #3307. Query on a Tree 17 | yz_ly | WA | 6ms | 9784kb | C++14 | 2.9kb | 2024-02-15 11:52:01 | 2024-02-15 11:52:02 |
Judging History
answer
#include<bits/stdc++.h>
#define ll long long
using namespace std;
inline int read(){
char ch=getchar();
int f=1,x=0;
while(ch<'0'||ch>'9'){
if(ch=='-')
f=-f;
ch=getchar();
}
while(ch>='0'&&ch<='9'){
x=x*10+ch-'0';
ch=getchar();
}
return f*x;
}
inline void work(int k){
if(k<0){
putchar('-');
k=-k;
}
if(k>9)
work(k/10);
putchar(k%10+'0');
}
/*
现将dfs序搞下来变成一个序列
容易发现第一个满足前缀和>=sum/2的点一定在带权重心
直接倍增跳就行了
*/
int n,q,cnt,first[100005],siz[100005],son[100005],dep[100005],fa[100005],dp[100005],dfn[100005],rev[100005],num,dp1[100005][25];
struct q1{
int u,w,nex;
}a[200005];
void add(int u1,int w1){
a[++cnt]={u1,w1,first[u1]};
first[u1]=cnt;
}
void dfs(int u,int dad){
dp1[u][0]=dad;
for(int i=1;i<=20;i++){
dp1[u][i]=dp1[dp1[u][i-1]][i-1];
}
dep[u]=dep[dad]+1;
fa[u]=dad;
siz[u]=1;
for(int i=first[u];i;i=a[i].nex){
if(a[i].w==dad)
continue;
dfs(a[i].w,u);
siz[u]+=siz[a[i].w];
if(siz[son[u]]<siz[a[i].w])
son[u]=a[i].w;
}
}
void dfs1(int u,int t){
dp[u]=t;
dfn[u]=++num;
rev[num]=u;
if(son[u])
dfs1(son[u],t);
for(int i=first[u];i;i=a[i].nex){
if(a[i].w==fa[u]||a[i].w==son[u])
continue;
dfs1(a[i].w,a[i].w);
}
}
struct node{
ll val[400005],lazy[400005];
void pushdown(int k,int l,int r){
int mid=(l+r)>>1;
val[2*k]+=lazy[k]*(mid-l+1);
val[2*k+1]+=lazy[k]*(r-mid);
lazy[2*k]+=lazy[k];
lazy[2*k+1]+=lazy[k];
lazy[k]=0;
}
void change(int k,int l,int r,int x,int y,ll v){
if(l>y||r<x)
return ;
if(l>=x&&r<=y){
val[k]+=(r-l+1)*v;
lazy[k]+=v;
return ;
}
pushdown(k,l,r);
int mid=(l+r)>>1;
change(2*k,l,mid,x,y,v);
change(2*k+1,mid+1,r,x,y,v);
val[k]=val[2*k]+val[2*k+1];
}
ll query(int k,int l,int r,int x,int y){
if(l>y||r<x)
return 0;
if(l>=x&&r<=y)
return val[k];
pushdown(k,l,r);
int mid=(l+r)>>1;
return query(2*k,l,mid,x,y)+query(2*k+1,mid+1,r,x,y);
}
}tree;
int main(){
n=read();
for(int i=1,x,y;i<n;i++){
x=read();
y=read();
add(x,y);
add(y,x);
}
dfs(1,0);
dfs1(1,1);
q=read();
while(q--){
int f=read(),u=read(),v;
if(f==1)
tree.change(1,1,n,dfn[u],dfn[u]+siz[u]-1,1);
else{
v=read();
int x=u,y=v;
while(dp[x]!=dp[y]){
if(dep[dp[x]]>dep[dp[y]])
swap(x,y);
tree.change(1,1,n,dfn[dp[y]],dfn[y],1);
y=fa[dp[y]];
}
if(dep[x]>dep[y])
swap(x,y);
tree.change(1,1,n,dfn[x],dfn[y],1);
}
int l=1,r=n+1;
ll sum=tree.query(1,1,n,1,n);
while(l<r){
int mid=(l+r)>>1;
if(tree.query(1,1,n,1,mid)>=sum/2.0)
r=mid;
else
l=mid+1;
}
l=rev[l];
for(int i=20;i>=0;i--){
int now=dp1[l][i];
if(now&&tree.query(1,1,n,dfn[now],dfn[now]+siz[now]-1)<=sum/2.0)
l=now;
}
if(sum==1||!dp1[l][0])
work(l);
else
work(dp1[l][0]);
putchar('\n');
}
return 0;
}
Details
Tip: Click on the bar to expand more detailed information
Test #1:
score: 100
Accepted
time: 1ms
memory: 9784kb
input:
7 1 6 1 7 7 3 3 2 7 5 5 4 4 1 2 1 4 1 6 2 6 7
output:
2 7 7 1
result:
ok 4 lines
Test #2:
score: 0
Accepted
time: 1ms
memory: 7676kb
input:
15 1 2 1 3 2 4 3 7 6 2 3 8 4 9 2 5 5 11 7 13 10 4 11 15 12 5 14 7 11 2 6 11 1 6 2 10 9 1 9 2 2 6 1 4 2 7 6 1 2 2 8 13 1 10 2 8 15
output:
2 2 2 2 2 2 2 2 2 2 2
result:
ok 11 lines
Test #3:
score: 0
Accepted
time: 1ms
memory: 7812kb
input:
2 1 2 1 1 1
output:
1
result:
ok single line: '1'
Test #4:
score: 0
Accepted
time: 1ms
memory: 7740kb
input:
2 1 2 1 2 1 2
output:
1
result:
ok single line: '1'
Test #5:
score: -100
Wrong Answer
time: 6ms
memory: 7632kb
input:
100 1 2 58 2 2 87 63 87 65 63 65 19 6 87 58 21 87 8 87 74 91 6 95 65 2 61 87 59 3 61 37 87 67 87 23 2 63 9 87 46 40 67 70 2 12 58 46 75 75 36 28 3 12 33 72 87 39 6 75 52 85 70 1 76 26 40 47 28 2 49 41 65 66 28 51 37 15 47 86 51 8 60 97 19 48 58 72 90 33 83 97 54 36 5 23 14 78 52 90 7 99 2 48 82 48 6...
output:
87 3 3 87 87 2 2 2 2 87 87 87 87 87 2 87 87 87 87 87 87 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 2 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87 87...
result:
wrong answer 1st lines differ - expected: '72', found: '87'