Python subprocess.call() invoking gnuplot then evi

Question

I've been trying to use Python 3.4 to call a simple gnuplot script and then display the result (a .ps file) using evince. My code looks like this:

plotfile = "plot.p"

with open(plotfile, "w") as plt:
    lines = ['set term postscript',
             'set output "speed.ps"',
             'set xlabel "Time"',
             'set ylabel "Speed"',
             "plot '{}' u 2:4 title 'speed' with lines".format(infile)
            ]
    lines = map(lambda x: x + '\n', lines)
    plt.writelines(lines)

    # Call gnuplot and evince
    subprocess.call(['gnuplot', plotfile])
    subprocess.call(['evince', 'speed.ps'])

However, evince often doesn't display the file. I can create the file correctly using

process = subprocess.Popen(['gnuplot', plotfile])  

but if I try to immediately open the file, by calling

process = subprocess.Popen(['gnuplot', plotfile])  
subprocess.Popen(['evince', plot.ps])

the file often doesn't display correctly, I'm assuming because the first command doesn't finish in time. To fix this issue, I've tried

process = subprocess.call(['gnuplot', plotfile])

or

process = subprocess.Popen(['gnuplot', plotfile])
process.wait() 

but it both cases no .ps file is even created. The only thing that works (and not always) is

process = subprocess.Popen(['gnuplot', plotfile])  
time.sleep(1)
subprocess.Popen(['evince', plot.ps])

but that is really ugly.

So my questions are:

1) Why might waiting for the process to finish (using either subprocess.call() or wait()) prevent the .ps file from even being created?
2) Is there some solution that doesn't involve using "sleep"?

Versions:

• CentOS 6.6
• Python 3.4
• Gnuplot 4.6
• Evince 2.28.2

Answer 1

I believe you've had some misleading test results leading you down the wrong path. Using subprocess.call() or calling .wait() on a Popen object does genuinely wait for the program that was invoked to exit. What's not guaranteed, however, is whether the result of your plt.writelines() calls will be flushed to disk or will still be in an in-memory buffer by the time gnuplot is started.

Given the following code:

plotfile = "plot.p"

with open(plotfile, "w") as plt:
    lines = ['set term postscript',
             'set output "speed.ps"',
             'set xlabel "Time"',
             'set ylabel "Speed"',
             "plot '{}' u 2:4 title 'speed' with lines".format(infile)
            ]
    lines = map(lambda x: x + '\n', lines)
    plt.writelines(lines)

    plt.flush() ### <-- EITHER CALL THIS OR MOVE THE BELOW OUTSIDE THE with BLOCK

    # Call gnuplot and evince
    subprocess.call(['gnuplot', plotfile])
    subprocess.call(['evince', 'speed.ps'])

...absent the flush(), the plt file isn't guaranteed to have all contents flushed before subprocess.call() is invoked; lines can still be buffered and not available to other programs yet.

Closing a file flushes it, so moving the call()s outside the with block will also solve your problem.