ProcessPoolExecutor与Popen配合使用

 一直以为concurrent.futures.ThreadPoolExecutor()可以配合os.execlp()，然后并不行，os.execlp()是装载一个新的程序运行，而保持原来的pid不变，原来的stdout和stderr没有被保留到新程序，会出现如下错误：

concurrent.futures.process.BrokenProcessPool: A process in the process pool was terminated abruptly while the future was running or pending.

据说可以通过修改stdout和stderr属性达到获取输出。不过不管怎么样，更好的配合应该是concurrent.futures.ThreadPoolExecutor()和subprocess.Popen()，它新启一个进程，然后用io.open建立好pipe，特别是有cwd参数，也是非常有用。

举例：

from subprocess import Popen, PIPE, STDOUT

def jobs():
    with Popen(["ifconfig", '-badargs'], stdout=PIPE, stderr=STDOUT) as proc:
        print(proc.stdout.read().decode())

import asyncio
from concurrent.futures import ThreadPoolExecutor, ProcessPoolExecutor

loop = asyncio.get_event_loop()

async def run():
    with ProcessPoolExecutor() as executor:
        await loop.run_in_executor(executor, jobs)

loop.run_until_complete(run())

这里将stderr的输出导入到stdout的pipe里，也可以各自建立一个pipe：

def jobs():
    with Popen(["ifconfig", '-badargs'], stdout=PIPE, stderr=PIPE) as proc:
        print(proc.stdout.read().decode())
        print(proc.stderr.read().decode())

with语法让关系文件，回收资源变成自动。