Codeforces Beta Round #1 C. Ancient Berland Circus

C. Ancient Berland Circus

time limit per test
2 seconds
memory limit per test
64 megabytes
input
standard input
output
standard output

Nowadays all circuses in Berland have a round arena with diameter 13 meters, but in the past things were different.

In Ancient Berland arenas in circuses were shaped as a regular (equiangular) polygon, the size and the number of angles could vary from one circus to another. In each corner of the arena there was a special pillar, and the rope strung between the pillars marked the arena edges.

Recently the scientists from Berland have discovered the remains of the ancient circus arena. They found only three pillars, the others were destroyed by the time.

You are given the coordinates of these three pillars. Find out what is the smallest area that the arena could have.
Input

The input file consists of three lines, each of them contains a pair of numbers –– coordinates of the pillar. Any coordinate doesn’t exceed 1000 by absolute value, and is given with at most six digits after decimal point.
Output

Output the smallest possible area of the ancient arena. This number should be accurate to at least 6 digits after the decimal point. It’s guaranteed that the number of angles in the optimal polygon is not larger than 100.
Examples
Input

0.000000 0.000000
1.000000 1.000000
0.000000 1.000000

链接http://codeforces.com/problemset/problem/1/C

题意

给出三个点，问这三个点所确定的最小的正多边形的面积

思路

多边形的顶点一定在三角形的外切圆上，要使面积最小则应边数最小，找出三条边所对圆心角的最大公约数a，2π/a即为边数

代码

#include <cstdio>
#include <cmath>
#include <algorithm>
using namespace std;
double eqs=1e-2;
double fgcd(double a,double b)
{
    return a <= eqs ? b : fgcd(fmod(b,a),a);
}
double area(double x1,double y1,double x2,double y2)
{
    double ans;
    ans=abs(x1*y2-x2*y1);
    ans/=2;
    return ans;
}
double pi=asin(1.0)*2;
int main()
{
    double x1,y1,x2,y2,x3,y3,s,ans,l1,l2,l3,angle1,angle2,angle3,ansa;
    scanf("%lf%lf%lf%lf%lf%lf",&x1,&y1,&x2,&y2,&x3,&y3);
    l1=sqrt((x1-x2)*(x1-x2)+(y1-y2)*(y1-y2));
    l2=sqrt((x1-x3)*(x1-x3)+(y1-y3)*(y1-y3));
    l3=sqrt((x3-x2)*(x3-x2)+(y3-y2)*(y3-y2));
    double t;
    if (l1<l2)
    {
        t=l1;
        l1=l2;
        l2=t;
    }
    if (l1<l3)
    {
        t=l1;
        l1=l3;
        l3=t;
    }
    s=area(x2-x1,y2-y1,x3-x1,y3-y1);
    angle1=asin(s*2.0/l1/l2)*2.0;
    angle2=asin(s*2.0/l1/l3)*2.0;
    angle3=pi*2-angle1-angle2;
    ansa=fgcd(angle1,fgcd(angle2,angle3));
    ans=sin(ansa)*l1*l1*l2*l2*l3*l3/s/s/16.0*pi/ansa;
    //printf("%lf\n",pi*2.0/ansa);
    //printf("%lf %lf %lf %lf %lf\n",l1,l2,l3,s,l1*l2*l3/s/4.0);
    //printf("%lf %lf %lf %lf\n",angle1,angle2,angle3,ansa);
    printf("%lf",ans);
}

PS

学到了一个算double型gcd的方法。。

double eqs=1e-2;
double fgcd(double a,double b)
{
    return a <= eqs ? b : fgcd(fmod(b,a),a);
}

fgcd()函数在cmath库中，使用时注意题目要求的精度。
比如这道题告诉了正多边形边数最大为100，所以eqs取到0.01即可。精度过高会使边数超过100，精度过低会使面积误差较大