通过fork和pipe调用gnuplot并更新图

Question

I want to do some realtime-plots during a simulation. For this, i would like to use octave or gnuplot. My current approach is to use a frontend to gnuplot, feedgnuplot, which fits actually very well.

The Simulation is written in C++, so i thought about forking (new process for feedgnuplot) and piping the relevant data to feedgnuplot.

The problem i have is that the output is only visible after the simulation. But i want to see the plot updated during simulation.

Here is a MWE:

#define _GNU_SOURCE
#include <stdio.h>
#include <unistd.h>

int main()
{
    FILE* stream = popen("feedgnuplot", "w");

    for(int i = 0; i < 10; ++i)
    {
        fprintf(stream, "%d\n", i * i);
        fflush(stream);
        sleep(1);
    }
}

The program stops after 10 secons, showing the plot.

When using feedgnuplot directly in the shell, everything works as expected. (That is, newly added data is plotted without the need to end the process)

What am i doing wrong? I think i lack some understanding of how popen really works.

Answer 1

First, let's write a fake feedgnuplot.c:

#include <stdio.h>
#include <stdlib.h>

int main()
{
    char *buf = NULL;
    size_t n = 0;

    while (getline (&buf, &n, stdin) != -1) {
        printf ("%s", buf);
    }

    free (buf);
    return 0;
}

With this, your code works, ie the lines are printed as they arrive.

I suspect the problem lies in the way your feedgnuplot program reads incoming data. You should show the relevant part of that code.

If I had to take a guess, you probably need to add

setvbuf (stdin, NULL, _IOLBF, 0);

in feedgnuplot before you start to read from stdin.

That is because by default, when stdin corresponds to a terminal it is line buffered, whereas when it corresponds to a pipe it is fully buffered. The code above makes stdin line buffered no matter what so there should be no difference between reading from a terminal or a pipe.