Live graph for a C application

Question

I have an application which logs periodically on to a host system it could be on a file or just a console. I would like to use this data to plot a statistical graph for me. I am not sure if I can use the live graph for my application.

If this tool is the right one, may I have an example on integrating the external application with the live graph?

this is livegraph link --> http://www.live-graph.org/download.html

Answer 1

Well, you only need to write your data in the given format of livegraph and set livegraph up to plot what you want. If wrote small C example which generates random numbers and dumps them together with the time every second. Next, you just attach the livegraph program to the file. That's it.

Playing around with LiveGraph I must say that its use is rather limited. I still would stick to a python script with matplotlib, since you have much more control over how and what is plotted.

#include <stdio.h>
#include <time.h>
#include <unistd.h>
#include <gsl/gsl_rng.h>
#include <gsl/gsl_randist.h>

int main(int argc, char** argv)
{
        FILE *f; 
        gsl_rng *r = NULL;
        const gsl_rng_type *T; 
        int seed = 31456;   
        double rndnum;
        T = gsl_rng_ranlxs2;
        r = gsl_rng_alloc(T);
        gsl_rng_set(r, seed);

        time_t t;
        t = time(NULL);



        f = fopen("test.lgdat", "a");
        fprintf(f, "##;##\n");
        fprintf(f,"@LiveGraph test file.\n");
        fprintf(f,"Time;Dataset number\n");

        for(;;){
                rndnum = gsl_ran_gaussian(r, 1); 
                fprintf(f,"%f;%f\n", (double)t, rndnum);
                sleep(1);
                fflush(f);
                t = time(NULL);
        }   

        gsl_rng_free(r);
        return 0;
}

compile with

gcc -Wall main.c  `gsl-config --cflags --libs`

Answer 2

I think this can be achieved easiest using Python plus matplotlib. To achieve this there are actually multiple ways: a) integrating the Python Interpreter directly in your C application, b) printing the data to stdout and piping this to a simple python script that does the actual plotting. In the following I will describe both approaches.

We have the following C application (e.g. plot.c). It uses the Python interpreter to interface with matplotlib's plotting functionality. The application is able to plot the data directly (when called like ./plot --plot-data) and to print the data to stdout (when called with any other argument set).

#include <Python.h>
#include <stdlib.h>
#include <stdio.h>
#include <stdbool.h>
#include <string.h>

#define CMD_BUF_SIZE 256

void initializePlotting() {
  Py_Initialize();
  // load matplotlib for plotting
  PyRun_SimpleString("from matplotlib import pyplot as pp");
  PyRun_SimpleString("pp.ion()"); // use pp.draw() instead of pp.show()
}

void uninitializePlotting() {
  Py_Finalize();
}

void plotPoint2d(double x, double y) {
  // this buffer will be used later to handle the commands to python
  static char command[CMD_BUF_SIZE];
  snprintf(command, CMD_BUF_SIZE, "pp.plot([%f],[%f],'r.')\npp.draw()", x, y);
  PyRun_SimpleString(command);
}

double myRandom() {
  double sum = .0;
  int count = 1e4;
  int i;
  for (i = 0; i < count; i++)
    sum = sum + rand()/(double)RAND_MAX;
  sum = sum/count;
  return sum;
}

int main (int argc, const char** argv) {
  bool plot = false;
  if (argc == 2 && strcmp(argv[1], "--plot-data") == 0)
    plot = true;

  if (plot) initializePlotting();

  // generate and plot the data
  int i = 0;
  for (i = 0; i < 1000; i++) {
    double x = myRandom(), y = myRandom();
    if (plot) plotPoint2d(x,y);
    else printf("%f %f\n", x, y);
  }

  if (plot) uninitializePlotting();
  return 0;
}

You can build it like this:

$ gcc plot.c -I /usr/include/python2.7 -l python2.7 -o plot

And run it like:

$ ./plot --plot-data

Then it will run for some time plotting red dots onto an axis.

When you choose not to plot the data directly but to print it to the stdout you may do the plotting by an external program (e.g. a Python script named plot.py) that takes input from stdin, i.e. a pipe, and plots the data it gets. To achieve this call the program like ./plot | python plot.py, with plot.py being similar to:

from matplotlib import pyplot as pp
pp.ion()

while True:
  # read 2d data point from stdin
  data = [float(x) for x in raw_input().split()]
  assert len(data) == 2, "can only plot 2d data!"
  x,y = data
  # plot the data
  pp.plot([x],[y],'r.')
  pp.draw()

I have tested both approaches on my debian machine. It requires the packages python2.7 and python-matplotlib to be installed.

EDIT

I have just seen, that you wanted to plot a bar plot or such thing, this of course is also possible using matplotlib, e.g. a histogram:

from matplotlib import pyplot as pp
pp.ion()

values = list()
while True:
  data = [float(x) for x in raw_input().split()]
  values.append(data[0])
  pp.clf()
  pp.hist([values])
  pp.draw()