Roadblocks（三分）

Bessie has moved to a small farm and sometimes enjoys returning to visit one of her best friends. She does not want to get to her old home too quickly, because she likes the scenery along the way. She has decided to take the second-shortest rather than the shortest path. She knows there must be some second-shortest path.

The countryside consists of R (1 ≤ R ≤ 100,000) bidirectional roads, each linking two of the N (1 ≤ N ≤ 5000) intersections, conveniently numbered 1…N. Bessie starts at intersection 1, and her friend (the destination) is at intersection N.

The second-shortest path may share roads with any of the shortest paths, and it may backtrack i.e., use the same road or intersection more than once. The second-shortest path is the shortest path whose length is longer than the shortest path(s) (i.e., if two or more shortest paths exist, the second-shortest path is the one whose length is longer than those but no longer than any other path).

Input
Line 1: Two space-separated integers: N and R
Lines 2… R+1: Each line contains three space-separated integers: A, B, and D that describe a road that connects intersections A and B and has length D (1 ≤ D ≤ 5000)
Output
Line 1: The length of the second shortest path between node 1 and node N
Sample Input
4 4
1 2 100
2 4 200
2 3 250
3 4 100
Sample Output
450
Hint
Two routes: 1 -> 2 -> 4 (length 100+200=300) and 1 -> 2 -> 3 -> 4 (length 100+250+100=450)

暑假的时候学的三分，当时也没怎么关注，也不咋会。。
二分主要处理单调的函数，想二次函数这种东西，就得用三分了

如图所示，已知左右端点L、R，要求找到白点的位置。

思路：通过不断缩小 [L,R] 的范围，无限逼近白点。

做法：先取 [L,R] 的中点 mid，再取 [mid,R] 的中点 mmid，通过比较 f(mid) 与 f(mmid) 的大小来缩小范围。

       当最后 L=R-1 时，再比较下这两个点的值，我们就找到了答案。

1、当 f(mid) > f(mmid) 的时候，我们可以断定 mmid 一定在白点的右边。

反证法：假设 mmid 在白点的左边，则 mid 也一定在白点的左边，又由 f(mid) > f(mmid) 可推出 mmid < mid，与已知矛盾，故假设不成立。

所以，此时可以将 R = mmid 来缩小范围。

2、当 f(mid) < f(mmid) 的时候，我们可以断定 mid 一定在白点的左边。

反证法：假设 mid 在白点的右边，则 mmid 也一定在白点的右边，又由 f(mid) < f(mmid) 可推出 mid > mmid，与已知矛盾，故假设不成立。

同理，此时可以将 L = mid 来缩小范围。

代码如下：

#include<iostream>
#include<cstdio>
#include<cstring>
#include<cmath>
#include<algorithm>
#define esp 1e-10
using namespace std;

const int maxx=3e2+10;
struct node{
	double x;
	double y;
	double vx;
	double vy;
}p[maxx];
int n;

double check(double t)
{
	double maxn=0;
	for(int i=0;i<n;i++)
	{
		for(int j=i+1;j<n;j++)
		{
			double tx1=p[i].x+p[i].vx*t;
			double ty1=p[i].y+p[i].vy*t;
			double tx2=p[j].x+p[j].vx*t;
			double ty2=p[j].y+p[j].vy*t;
			maxn=max(maxn,sqrt((tx1-tx2)*(tx1-tx2)+(ty1-ty2)*(ty1-ty2)));
		}	
	}
	return maxn;
}

double cal()
{
	double l=0;
	double r=1e10;
	while(r-l>=esp)
	{
		double mid=(l+r)/2;
		double midd=(mid+r)/2;
		if(check(mid)>check(midd)) l=mid+esp;
		else r=midd-esp;
	}
	return l;
}

int main()
{
	int t;
	scanf("%d",&t);
	int k=0;
	while(t--)
	{
		scanf("%d",&n);
		for(int i=0;i<n;i++) scanf("%lf%lf%lf%lf",&p[i].x,&p[i].y,&p[i].vx,&p[i].vy);
		double x=cal();
		printf("Case #%d: %.2lf %.2lf\n",++k,x,check(x));
	}
}

努力加油a啊，(^o)/~