Blocking and Non Blocking subprocess calls
Question:
I’m completely confused between subprocess.call() , subprocess.Popen(), subprocess.check_call().
Which is blocking and which is not ?
What I mean to say is if I use subprocess.Popen() whether the parent process waits for the child process to return/exit before it keep on its execution.
How does shell=True affect these calls?
Answers:
Popen is nonblocking. call and check_call are blocking.
You can make the Popen instance block by calling its wait or communicate method.
If you look in the source code, you’ll see call calls Popen(...).wait(), which is why it is blocking.
check_call calls call, which is why it blocks as well.
Strictly speaking, shell=True is orthogonal to the issue of blocking. However, shell=True causes Python to exec a shell and then run the command in the shell. If you use a blocking call, the call will return when the shell finishes. Since the shell may spawn a subprocess to run the command, the shell may finish before the spawned subprocess. For example,
import subprocess
import time
proc = subprocess.Popen('ls -lRa /', shell=True)
time.sleep(3)
proc.terminate()
proc.wait()
Here two processes are spawned: Popen spawns one subprocess running the shell. The shell in turn spawns a subprocess running ls. proc.terminate() kills the shell, but the subprocess running ls remains. (That is manifested by copious output, even after the python script has ended. Be prepared to kill the ls with pkill ls.)
I’m completely confused between subprocess.call() , subprocess.Popen(), subprocess.check_call().
Which is blocking and which is not ?
What I mean to say is if I use subprocess.Popen() whether the parent process waits for the child process to return/exit before it keep on its execution.
How does shell=True affect these calls?
Popen is nonblocking. call and check_call are blocking.
You can make the Popen instance block by calling its wait or communicate method.
If you look in the source code, you’ll see call calls Popen(...).wait(), which is why it is blocking.
check_call calls call, which is why it blocks as well.
Strictly speaking, shell=True is orthogonal to the issue of blocking. However, shell=True causes Python to exec a shell and then run the command in the shell. If you use a blocking call, the call will return when the shell finishes. Since the shell may spawn a subprocess to run the command, the shell may finish before the spawned subprocess. For example,
import subprocess
import time
proc = subprocess.Popen('ls -lRa /', shell=True)
time.sleep(3)
proc.terminate()
proc.wait()
Here two processes are spawned: Popen spawns one subprocess running the shell. The shell in turn spawns a subprocess running ls. proc.terminate() kills the shell, but the subprocess running ls remains. (That is manifested by copious output, even after the python script has ended. Be prepared to kill the ls with pkill ls.)