是否可以将 subprocess.Popen 的 stdout 重新连接到 sys.stdout？ （蟒蛇3）

Question

I have a python program that runs another program using popen. Is it possible to read from the programme's output for a while, and the connect the subprogram's output to sys.stdout , and allow it to run in parallel as if I had started it without the stdout=subprocess.PIPE option?

wrapper:

#! /usr/bin/env python3

import subprocess
from time import sleep
import sys

p = subprocess.Popen(
  ["./count.py"],
  stdout=subprocess.PIPE,
)

for line in iter(p.stdout.readline,''):
  l = line.decode('utf-8')
  sys.stdout.write(">>" + l)
  if "8" in l:
    # somehow reconnect p.stdout to sys.stdout
    break

sleep(1)

print("now do some stuff in parallel")

sleep(1)

p.terminate()

test program:

#! /usr/bin/env python

from time import sleep
import sys

for x in range(1,100):
  sleep(.1)
  print x
  sys.stdout.flush()

output:

$ ./wrapper.py
>>1
>>2
>>3
>>4
>>5
>>6
>>7
>>8
now do some stuff in parallel

is it possible to do this?

$ ./wrapper.py
>>1
>>2
...
>>7
>>8
9
10
...
18
19
now do some stuff in parallel
20
21
...
28
29

Answer 1

You cannot "reconnect" the file descriptors of a subprocess you've already launched (except perhaps with very dirty and OS-specific tricks). The only file descriptor you have control over, is the one that's still in your process, and it's only useful for reading. So if you want things to go to stdout, then everything that comes into the read end would have to be copied to stdout somehow.

Fortunately, you can always spawn another subprocess to do that. This is just the kind of thing that the handy Unix utility cat does well.

subprocess.Popen(
  ["/bin/cat"],
  stdin=p.stdout,
)

This will basically create a pipeline, where you pipe the output of one process into another process. It can be very useful. In this case the other process ( cat ) just outputs to stdout.