Mister Muffin Blog

When you have data that is not instantly generated but trickles in one by one (either because calculation is difficult or because the data source produces new datapoints over time) but you still want to visualize it with gnuplot the easiest option is to hit the refresh button in the wx interface everytime you want to refresh the data.

A more cool way, is to make gnuplot "autoupdate" the data it displays whenever new data is available automagically.

Lets simulate a data source by the output of this python script that will just output coordinates on a circle with radius one around the coordinate center and wait for 0.5 seconds after every step.

from math import pi, sin, cos
from time import sleep

for alpha in [2*pi*i/16.0 for i in range(16)]:
    print sin(alpha), cos(alpha)
    sleep(0.5)

The magic will be done by this shell script which does nothing else than reading lines from stdin, appending them to a buffer and printing the buffer with gnuplot syntax around it every time new input is received.

while read line; do
        lines="$lines$line\n"
        echo "plot \"-\""
        echo -n $lines
        echo "e"
done

These two scripts can now be used like this:

python -u circle.py | sh live-gnuplot.sh | gnuplot -p

The -u option has to be passed to python to enable unbuffered output. The -p option for gnuplot makes the plot window remain even after the main gnuplot exited.

Since you with this setup gnuplot would auto-scale the coordinate axes according to the current input, a more beatiful output with a fixed size coordinate frame would be produced by:

{ echo "set xrange [-2:2]"; echo "set yrange [-2:2]"; python -u circle.py | sh live-gnuplot.sh } | gnuplot -p

This way you can also pass more options like title, style, terminal or others.