简体   繁体   English

从多个子进程读取非阻塞(Python)

[英]Non-blocking read from multiple subprocesses (Python)

I currently have the following code, inspired by the answer to Non-blocking read on a subprocess.PIPE in python . 我目前有以下代码,灵感来自python中subprocess.PIPE上的非阻塞读取的答案。 It seems to work correctly, outputting the lines to the screen, however it only does so for the first created process, all other processes (which are running) don't get any data printed. 它似乎工作正常,将行输出到屏幕,但它只对第一个创建的进程执行此操作,所有其他进程(正在运行)不会打印任何数据。

How do I make sure I can read data (in a non-blocking way) from multiple subprocesses? 如何确保我可以从多个子进程中读取数据(以非阻塞方式)?

#!/usr/bin/env python
import sys
import os
import subprocess
from threading import Thread
from Queue import Queue, Empty

STREAMER_URL = 'rtmp://127.0.0.1/app'
RTMPDUMP_EXECUTEABLE = 'rtmpdump'

def enqueue_output(out, queue):
    for line in iter(lambda: out.read(16), b''):
        queue.put(line)
    out.close()

def download_rtmp(media, filename):
  # Create parameters
  args=[RTMPDUMP_EXECUTEABLE]
  args.extend(['-r',media[0],'-y',media[1]])

  # Set output file
  OUTPUT_FILE = filename
  args.extend(['-o',OUTPUT_FILE])

  # Send rtmpdump any extra arguments
  if len(sys.argv) > 2:
    args.extend(sys.argv[2:])

  # Execute rtmpdump
  p = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT )
  q = Queue()
  t = Thread(target=enqueue_output, args=(p.stdout, q))
  t.daemon = True # thread dies with the program
  t.start()
  return (p, q, t)

def main():

  # actual data is from somewhere else on the internet
  for (name, playpath, filepath) in data:
    print 'Spawning %s download...' % name
    PROCESSES.append(download_rtmp((STREAMER_URL, playpath), filepath))

  BUFS = dict()

  # infinite loop checking if all processes have finished
  while True:
    done = True
    for (process, queue, thread) in PROCESSES:
      try:
        readdata = queue.get_nowait()
      except Empty:
        pass
      else:
        if process in BUFS:
          readdata = BUFS[process] + readdata
        lines = readdata.split('\n')
        if len(lines) > 1:
          for line in lines[:-1]:
            print 'Line: %s' % line
        if '\r' in lines[-1]:
          lines = readdata.split('\r')
          for line in lines[:-1]:
            print 'Line2: %s' % line
        BUFS[process] = lines[-1]

      process.poll()

      if process.returncode is None:
        done = False
        break
    if done:
      break

  print "Done"

if __name__ == "__main__":
    main()

I haven't figured the whole thing out, but the break in if process.returncode is None: means that you won't look at other process queues until the first process exits completely. 我没有想到整个事情,但是if process.returncode is None:则中断if process.returncode is None:意味着在第一个进程完全退出之前你不会查看其他进程队列。 And I'm not sure where you got that multi-queue polling thing from, but its absolutely horrible. 而且我不确定你从哪里获得多队列轮询,但它绝对可怕。

This problem is best solved with a single return queue used by all of the worker threads. 使用所有工作线程使用的单个返回队列可以最好地解决此问题。 The workers pass tuples of (process, line) and the main thread does a blocking wait for data from all of the workers. 工作者传递(进程,行)的元组,主线程阻塞等待来自所有工作者的数据。

This is pseudocode really, but it would look like: 这实际上是伪代码,但它看起来像:

STREAMER_URL = 'rtmp://127.0.0.1/app'
RTMPDUMP_EXECUTEABLE = 'rtmpdump'

def enqueue_output(process, queue):
    """read process stdout in small chunks and queue for processing"""
    for line in iter(lambda: out.read(16), b''):
        queue.put((process, line))
    process.wait()
    queue.put((process, None))

def download_rtmp(media, filename):
  # Create parameters
  args=[RTMPDUMP_EXECUTEABLE, '-r', media[0], '-y', media[1], '-o', filename]

  # Send rtmpdump any extra arguments
  # if len(sys.argv) > 2: no need for the if in list comprehension
  args.extend(sys.argv[2:])

  # Execute rtmpdump
  p = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT )
  t = Thread(target=enqueue_output, args=(p, return_q))
  t.daemon = True # thread dies with the program
  t.start()
  return (p, t)

def main():
  THREADS = []
  BUFS = dict()

  # actual data is from somewhere else on the internet
  for (name, playpath, filepath) in data:
    print 'Spawning %s download...' % name
    process, thread = download_rtmp((STREAMER_URL, playpath), filepath)
    BUFS[process] = ''
    THREADS.append(thread)

  # all processes write to return_q and we process them here
  while BUFS:
    process, line = return_q.get()
    readdata = BUFS[process] + (line or '')
    if line is None:
        del BUFS[process]
    # I didn't try to figure this part out... basically, when line is
    # None, process is removed from BUFS so you know your end condition
    # and the following stuff should do its final processing.
    lines = readdata.split('\n')
    if len(lines) > 1:
      for line in lines[:-1]:
        print 'Line: %s' % line
    if '\r' in lines[-1]:
      lines = readdata.split('\r')
      for line in lines[:-1]:
        print 'Line2: %s' % line
    if line is not None:
        BUFS[process] = lines[-1]

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM