C++ How to write full precision of doubles into a .dat file

Question

I have a program that does an Symplectic ODE integration (physics/mathematics) and I want to export the time series to an .dat file> However the numbers written in the dat file have only 6 digits precision. I write setprecision(15); before writing but it changes nothing. I also post a part of the code, without the actual ODE solver:

#include <iostream>
#include <ctime>         
#include <cstdlib>
#include <string>
#include <sstream>             
#include <iomanip>             
#include <cmath>              
#define pi 3.14159265358979
using namespace std;

int main(int argc, char **argv) {
// many stuff here, probably irrelevant,

ostringstream osE, osb, ospx;
    osb<<b; // using this, I can use some numbers into the file's name
    osE<<E;
    ospx<<px0;
filenamex = "Antidot_v4_x(t)_E=" + osE.str() + "_px0=" + ospx.str() + "_b=" + osb.str() + ".dat";
 ofstream file1( filenamex.c_str() );
file1<<t<<"\t"<<x0<<endl;
while(i<=N){
        i++;
        McLachanAtela(x, y, px, py, h);   
// Does the 4-step ODE solver. x are initial values and after the 
// function call x are final values after h time

        setprecision(15);
        file1<<t<<"\t"<<x<<endl; //I use this to write values in file
    }
 return 0;
}

So when I open the file1 (it's not named file1) the values inside it are 6 digits numbers. How can I write full 16 digit accurate numbers? Thanks. For completeness I also post the void function called:

void McLachanAtela (double& previousx, double& previousy, double& previouspx, double& previouspy, double timestep){
    // Atela Coefficients
    double c[4]={0.134496199277431089, -0.224819803079420805, 0.756320000515668291, 0.334003603286321425};
    double d[4]={0.515352837431122936, -0.085782019412973646, 0.411583023616466525, 0.128846158365384185};
    // Symplectic Algorithm (at dimensionless form)
    for(int j=0; j<4; j++){  
                                                        //this is the derivative of the potential :
        previouspx = previouspx - d[j]*timestep*b*pi*sin(pi*previousx)*cos(pi*previousy)*(pow(cos(pi*previousx)*cos(pi*previousy),b-1)); 
        previouspy = previouspy - d[j]*timestep*b*pi*sin(pi*previousy)*cos(pi*previousx)*(pow(cos(pi*previousx)*cos(pi*previousy),b-1));  
        previousx = previousx + c[j]*previouspx*timestep;
        previousy = previousy + c[j]*previouspy*timestep;
        //cout<<" dpx = "<<d[j]*timestep*b*pi*sin(pi*previousx)*cos(pi*previousy)*(pow(cos(pi*previousx)*cos(pi*previousy),b-1))<<endl;

    }
}

Your definition of pi will cause you trouble if you ever use it with a Solaris long double: it's not precise enough. — P45 Imminent, Jul 12 '15 at 11:50

score 1 · Accepted Answer · answered Jul 12 '15 at 11:42

1

You need to inject set precision into the stream:

file1 << setprecision(15) <<

answered Jul 12 '15 at 11:42

P45 Imminent

7,466
2
30
69

Thank you very much. Another one that works is file1.setprecision(15); before adding values. – George Datseris Jul 12 '15 at 13:21
You can use [`DBL_DIG`](http://stackoverflow.com/a/9999508/3087952), which is 15 from `` – nonsensation Jul 12 '15 at 18:36

C++ How to write full precision of doubles into a .dat file

1 Answers1