How to close a file in a subprocess before terminating the subprocess?

Question

I want to make a subprocess(writer.py) that keeps adding "test" to a file('20180309.txt') every second and maintain the subprocess with another python process(starter.py).

It works fine, but one thing that bothers me is that the file object is not closed before terminating the subprocess. I'd like to know if there's a way to close the file object when the subprocess recieves terminate() signal, or if there are other ways to handle this problem.

#starter.py
import os
import sys
import subprocess

p1 = subprocess.Popen( ["d:\python36-32\python.exe", r"""writer.py"""], shell=False )
print(p1.pid)

try:
    p1.communicate(timeout=5)
except subprocess.TimeoutExpired:
    p1.terminate()




#writer.py
import time
from datetime import datetime

file = open( datetime.now().strftime('%Y-%m-%d') + '.txt', 'a')

while True:
    file.write('test\n')
    file.flush()
    time.sleep(1)

Answer 1

First, any modern OS will close any (non-shared) files that a process has open when it goes away. You do not have to worry about this.

In a more complex application, you might want to worry about partial writes (or parts of any other "transactions" that can't be done atomically), but for simple cases, there's nothing to worry about at all.

Meanwhile, the terminate method in subprocess is not meant for polite shutdown requests. If you want to shut down the child, there are better ways to send it a message. In fact, since the child and parent are both Python, you may be happier using multiprocessing instead of subprocess in the first place.

That being said, on *nix, terminate happens to send SIGTERM, and SIGTERM is designed for polite shutdown. If you really want to handle this—and don't care about Windows—read the signal library for how to write a signal handler for SIGTERM. From the signal handler, you can set a global "quit next time through the loop" variable. or call sys.exit (which I think will raise an exception into the top-level code, but don't quote me on that), or whatever else you want.