在Python中对子进程进行非阻塞读取

I’m using the subprocess module to start a subprocess and connect to it’s output stream (stdout). I want to be able to execute non-blocking reads on its stdout. Is there a way to make .readline non-blocking or to check if there is data on the stream before I invoke .readline? I’d like this to be portable or at least work under Windows and Linux.

here is how I do it for now (It’s blocking on the .readline if no data is avaible):

p = subprocess.Popen('myprogram.exe', stdout = subprocess.PIPE)
output_str = p.stdout.readline()

fcntl, select, asyncproc won’t help in this case.

A reliable way to read a stream without blocking regardless of operating system is to use Queue.get_nowait():

import sys
from subprocess import PIPE, Popen
from threading  import Thread

try:
    from queue import Queue, Empty
except ImportError:
    from Queue import Queue, Empty  # python 2.x

ON_POSIX = 'posix' in sys.builtin_module_names

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

p = Popen(['myprogram.exe'], stdout=PIPE, bufsize=1, close_fds=ON_POSIX)
q = Queue()
t = Thread(target=enqueue_output, args=(p.stdout, q))
t.daemon = True # thread dies with the program
t.start()

# ... do other things here

# read line without blocking
try:  line = q.get_nowait() # or q.get(timeout=.1)
except Empty:
    print('no output yet')
else: # got line
    # ... do something with line

I have often had a similar problem; Python programs I write frequently need to have the ability to execute some primary functionality while simultaneously accepting user input from the command line (stdin). Simply putting the user input handling functionality in another thread doesn’t solve the problem because readline() blocks and has no timeout. If the primary functionality is complete and there is no longer any need to wait for further user input I typically want my program to exit, but it can’t because readline() is still blocking in the other thread waiting for a line. A solution I have found to this problem is to make stdin a non-blocking file using the fcntl module:

import fcntl
import os
import sys

# make stdin a non-blocking file
fd = sys.stdin.fileno()
fl = fcntl.fcntl(fd, fcntl.F_GETFL)
fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)

# user input handling thread
while mainThreadIsRunning:
      try: input = sys.stdin.readline()
      except: continue
      handleInput(input)

In my opinion this is a bit cleaner than using the select or signal modules to solve this problem but then again it only works on UNIX…

Python 3.4 introduces new provisional API for asynchronous IO — asyncio module.

The approach is similar to twisted-based answer by @Bryan Ward — define a protocol and its methods are called as soon as data is ready:

#!/usr/bin/env python3
import asyncio
import os

class SubprocessProtocol(asyncio.SubprocessProtocol):
    def pipe_data_received(self, fd, data):
        if fd == 1: # got stdout data (bytes)
            print(data)

    def connection_lost(self, exc):
        loop.stop() # end loop.run_forever()

if os.name == 'nt':
    loop = asyncio.ProactorEventLoop() # for subprocess' pipes on Windows
    asyncio.set_event_loop(loop)
else:
    loop = asyncio.get_event_loop()
try:
    loop.run_until_complete(loop.subprocess_exec(SubprocessProtocol, 
        "myprogram.exe", "arg1", "arg2"))
    loop.run_forever()
finally:
    loop.close()

See “Subprocess” in the docs.

There is a high-level interface asyncio.create_subprocess_exec() that returns Process objects that allows to read a line asynchroniosly using StreamReader.readline() coroutine (with async/await Python 3.5+ syntax):

#!/usr/bin/env python3.5
import asyncio
import locale
import sys
from asyncio.subprocess import PIPE
from contextlib import closing

async def readline_and_kill(*args):
    # start child process
    process = await asyncio.create_subprocess_exec(*args, stdout=PIPE)

    # read line (sequence of bytes ending with b'\n') asynchronously
    async for line in process.stdout:
        print("got line:", line.decode(locale.getpreferredencoding(False)))
        break
    process.kill()
    return await process.wait() # wait for the child process to exit


if sys.platform == "win32":
    loop = asyncio.ProactorEventLoop()
    asyncio.set_event_loop(loop)
else:
    loop = asyncio.get_event_loop()

with closing(loop):
    sys.exit(loop.run_until_complete(readline_and_kill(
        "myprogram.exe", "arg1", "arg2")))

readline_and_kill() performs the following tasks:

• start subprocess, redirect its stdout to a pipe
• read a line from subprocess’ stdout asynchronously
• kill subprocess
• wait for it to exit

Each step could be limited by timeout seconds if necessary.

Try the asyncproc module. For example:

import os
from asyncproc import Process
myProc = Process("myprogram.app")

while True:
    # check to see if process has ended
    poll = myProc.wait(os.WNOHANG)
    if poll != None:
        break
    # print any new output
    out = myProc.read()
    if out != "":
        print out

The module takes care of all the threading as suggested by S.Lott.

You can do this really easily in Twisted. Depending upon your existing code base, this might not be that easy to use, but if you are building a twisted application, then things like this become almost trivial. You create a ProcessProtocol class, and override the outReceived() method. Twisted (depending upon the reactor used) is usually just a big select() loop with callbacks installed to handle data from different file descriptors (often network sockets). So the outReceived() method is simply installing a callback for handling data coming from STDOUT. A simple example demonstrating this behavior is as follows:

from twisted.internet import protocol, reactor

class MyProcessProtocol(protocol.ProcessProtocol):

    def outReceived(self, data):
        print data

proc = MyProcessProtocol()
reactor.spawnProcess(proc, './myprogram', ['./myprogram', 'arg1', 'arg2', 'arg3'])
reactor.run()

The Twisted documentation has some good information on this.

If you build your entire application around Twisted, it makes asynchronous communication with other processes, local or remote, really elegant like this. On the other hand, if your program isn’t built on top of Twisted, this isn’t really going to be that helpful. Hopefully this can be helpful to other readers, even if it isn’t applicable for your particular application.

Use select & read(1).

import subprocess     #no new requirements
def readAllSoFar(proc, retVal=''): 
  while (select.select([proc.stdout],[],[],0)[0]!=[]):   
    retVal+=proc.stdout.read(1)
  return retVal
p = subprocess.Popen(['/bin/ls'], stdout=subprocess.PIPE)
while not p.poll():
  print (readAllSoFar(p))

For readline()-like:

lines = ['']
while not p.poll():
  lines = readAllSoFar(p, lines[-1]).split('\n')
  for a in range(len(lines)-1):
    print a
lines = readAllSoFar(p, lines[-1]).split('\n')
for a in range(len(lines)-1):
  print a

One solution is to make another process to perform your read of the process, or make a thread of the process with a timeout.

Here’s the threaded version of a timeout function:

http://code.activestate.com/recipes/473878/

However, do you need to read the stdout as it’s coming in? Another solution may be to dump the output to a file and wait for the process to finish using p.wait().

f = open('myprogram_output.txt','w')
p = subprocess.Popen('myprogram.exe', stdout=f)
p.wait()
f.close()


str = open('myprogram_output.txt','r').read()

Disclaimer: this works only for tornado

You can do this by setting the fd to be nonblocking and then use ioloop to register callbacks. I have packaged this in an egg called tornado_subprocess and you can install it via PyPI:

easy_install tornado_subprocess

now you can do something like this:

import tornado_subprocess
import tornado.ioloop

    def print_res( status, stdout, stderr ) :
    print status, stdout, stderr
    if status == 0:
        print "OK:"
        print stdout
    else:
        print "ERROR:"
        print stderr

t = tornado_subprocess.Subprocess( print_res, timeout=30, args=[ "cat", "/etc/passwd" ] )
t.start()
tornado.ioloop.IOLoop.instance().start()

you can also use it with a RequestHandler

class MyHandler(tornado.web.RequestHandler):
    def on_done(self, status, stdout, stderr):
        self.write( stdout )
        self.finish()

    @tornado.web.asynchronous
    def get(self):
        t = tornado_subprocess.Subprocess( self.on_done, timeout=30, args=[ "cat", "/etc/passwd" ] )
        t.start()

Existing solutions did not work for me (details below). What finally worked was to implement readline using read(1) (based on this answer). The latter does not block:

from subprocess import Popen, PIPE
from threading import Thread
def process_output(myprocess): #output-consuming thread
    nextline = None
    buf = ''
    while True:
        #--- extract line using read(1)
        out = myprocess.stdout.read(1)
        if out == '' and myprocess.poll() != None: break
        if out != '':
            buf += out
            if out == '\n':
                nextline = buf
                buf = ''
        if not nextline: continue
        line = nextline
        nextline = None

        #--- do whatever you want with line here
        print 'Line is:', line
    myprocess.stdout.close()

myprocess = Popen('myprogram.exe', stdout=PIPE) #output-producing process
p1 = Thread(target=process_output, args=(dcmpid,)) #output-consuming thread
p1.daemon = True
p1.start()

#--- do whatever here and then kill process and thread if needed
if myprocess.poll() == None: #kill process; will automatically stop thread
    myprocess.kill()
    myprocess.wait()
if p1 and p1.is_alive(): #wait for thread to finish
    p1.join()

Why existing solutions did not work:

1. Solutions that require readline (including the Queue based ones) always block. It is difficult (impossible?) to kill the thread that executes readline. It only gets killed when the process that created it finishes, but not when the output-producing process is killed.
2. Mixing low-level fcntl with high-level readline calls may not work properly as anonnn has pointed out.
3. Using select.poll() is neat, but doesn’t work on Windows according to python docs.
4. Using third-party libraries seems overkill for this task and adds additional dependencies.

Here is my code, used to catch every output from subprocess ASAP, including partial lines. It pumps at same time and stdout and stderr in almost correct order.

Tested and correctly worked on Python 2.7 linux & windows.

#!/usr/bin/python
#
# Runner with stdout/stderr catcher
#
from sys import argv
from subprocess import Popen, PIPE
import os, io
from threading import Thread
import Queue
def __main__():
    if (len(argv) > 1) and (argv[-1] == "-sub-"):
        import time, sys
        print "Application runned!"
        time.sleep(2)
        print "Slept 2 second"
        time.sleep(1)
        print "Slept 1 additional second",
        time.sleep(2)
        sys.stderr.write("Stderr output after 5 seconds")
        print "Eol on stdin"
        sys.stderr.write("Eol on stderr\n")
        time.sleep(1)
        print "Wow, we have end of work!",
    else:
        os.environ["PYTHONUNBUFFERED"]="1"
        try:
            p = Popen( argv + ["-sub-"],
                       bufsize=0, # line-buffered
                       stdin=PIPE, stdout=PIPE, stderr=PIPE )
        except WindowsError, W:
            if W.winerror==193:
                p = Popen( argv + ["-sub-"],
                           shell=True, # Try to run via shell
                           bufsize=0, # line-buffered
                           stdin=PIPE, stdout=PIPE, stderr=PIPE )
            else:
                raise
        inp = Queue.Queue()
        sout = io.open(p.stdout.fileno(), 'rb', closefd=False)
        serr = io.open(p.stderr.fileno(), 'rb', closefd=False)
        def Pump(stream, category):
            queue = Queue.Queue()
            def rdr():
                while True:
                    buf = stream.read1(8192)
                    if len(buf)>0:
                        queue.put( buf )
                    else:
                        queue.put( None )
                        return
            def clct():
                active = True
                while active:
                    r = queue.get()
                    try:
                        while True:
                            r1 = queue.get(timeout=0.005)
                            if r1 is None:
                                active = False
                                break
                            else:
                                r += r1
                    except Queue.Empty:
                        pass
                    inp.put( (category, r) )
            for tgt in [rdr, clct]:
                th = Thread(target=tgt)
                th.setDaemon(True)
                th.start()
        Pump(sout, 'stdout')
        Pump(serr, 'stderr')

        while p.poll() is None:
            # App still working
            try:
                chan,line = inp.get(timeout = 1.0)
                if chan=='stdout':
                    print "STDOUT>>", line, "<?<"
                elif chan=='stderr':
                    print " ERROR==", line, "=?="
            except Queue.Empty:
                pass
        print "Finish"

if __name__ == '__main__':
    __main__()

I add this problem to read some subprocess.Popen stdout. Here is my non blocking read solution:

import fcntl

def non_block_read(output):
    fd = output.fileno()
    fl = fcntl.fcntl(fd, fcntl.F_GETFL)
    fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)
    try:
        return output.read()
    except:
        return ""

# Use example
from subprocess import *
sb = Popen("echo test && sleep 1000", shell=True, stdout=PIPE)
sb.kill()

# sb.stdout.read() # <-- This will block
non_block_read(sb.stdout)
'test\n'

This version of non-blocking read doesn’t require special modules and will work out-of-the-box on majority of Linux distros.

import os
import sys
import time
import fcntl
import subprocess

def async_read(fd):
    # set non-blocking flag while preserving old flags
    fl = fcntl.fcntl(fd, fcntl.F_GETFL)
    fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)
    # read char until EOF hit
    while True:
        try:
            ch = os.read(fd.fileno(), 1)
            # EOF
            if not ch: break                                                                                                                                                              
            sys.stdout.write(ch)
        except OSError:
            # waiting for data be available on fd
            pass

def shell(args, async=True):
    # merge stderr and stdout
    proc = subprocess.Popen(args, shell=False, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
    if async: async_read(proc.stdout)
    sout, serr = proc.communicate()
    return (sout, serr)

if __name__ == '__main__':
    cmd = 'ping 8.8.8.8'
    sout, serr = shell(cmd.split())

Here is a simple solution based on threads which:

• works on both Linux and Windows (not relying on select).
• reads both stdout and stderr asynchronouly.
• doesn’t rely on active polling with arbitrary waiting time (CPU friendly).
• doesn’t use asyncio (which may conflict with other libraries).
• runs until the child process terminates.

printer.py

import time
import sys

sys.stdout.write("Hello\n")
sys.stdout.flush()
time.sleep(1)
sys.stdout.write("World!\n")
sys.stdout.flush()
time.sleep(1)
sys.stderr.write("That's an error\n")
sys.stderr.flush()
time.sleep(2)
sys.stdout.write("Actually, I'm fine\n")
sys.stdout.flush()
time.sleep(1)

reader.py

import queue
import subprocess
import sys
import threading


def enqueue_stream(stream, queue, type):
    for line in iter(stream.readline, b''):
        queue.put(str(type) + line.decode('utf-8'))
    stream.close()


def enqueue_process(process, queue):
    process.wait()
    queue.put('x')


p = subprocess.Popen('python printer.py', stdout=subprocess.PIPE, stderr=subprocess.PIPE)
q = queue.Queue()
to = threading.Thread(target=enqueue_stream, args=(p.stdout, q, 1))
te = threading.Thread(target=enqueue_stream, args=(p.stderr, q, 2))
tp = threading.Thread(target=enqueue_process, args=(p, q))
te.start()
to.start()
tp.start()

while True:
    line = q.get()
    if line[0] == 'x':
        break
    if line[0] == '2':  # stderr
        sys.stdout.write("\033[0;31m")  # ANSI red color
    sys.stdout.write(line[1:])
    if line[0] == '2':
        sys.stdout.write("\033[0m")  # reset ANSI code
    sys.stdout.flush()

tp.join()
to.join()
te.join()

Adding this answer here since it provides ability to set non-blocking pipes on Windows and Unix.

All the ctypes details are thanks to @techtonik’s answer.

There is a slightly modified version to be used both on Unix and Windows systems.

• Python3 compatible (only minor change needed).
• Includes posix version, and defines exception to use for either.

This way you can use the same function and exception for Unix and Windows code.

# pipe_non_blocking.py (module)
"""
Example use:

    p = subprocess.Popen(
            command,
            stdout=subprocess.PIPE,
            )

    pipe_non_blocking_set(p.stdout.fileno())

    try:
        data = os.read(p.stdout.fileno(), 1)
    except PortableBlockingIOError as ex:
        if not pipe_non_blocking_is_error_blocking(ex):
            raise ex
"""


__all__ = (
    "pipe_non_blocking_set",
    "pipe_non_blocking_is_error_blocking",
    "PortableBlockingIOError",
    )

import os


if os.name == "nt":
    def pipe_non_blocking_set(fd):
        # Constant could define globally but avoid polluting the name-space
        # thanks to: https://stackoverflow.com/questions/34504970
        import msvcrt

        from ctypes import windll, byref, wintypes, WinError, POINTER
        from ctypes.wintypes import HANDLE, DWORD, BOOL

        LPDWORD = POINTER(DWORD)

        PIPE_NOWAIT = wintypes.DWORD(0x00000001)

        def pipe_no_wait(pipefd):
            SetNamedPipeHandleState = windll.kernel32.SetNamedPipeHandleState
            SetNamedPipeHandleState.argtypes = [HANDLE, LPDWORD, LPDWORD, LPDWORD]
            SetNamedPipeHandleState.restype = BOOL

            h = msvcrt.get_osfhandle(pipefd)

            res = windll.kernel32.SetNamedPipeHandleState(h, byref(PIPE_NOWAIT), None, None)
            if res == 0:
                print(WinError())
                return False
            return True

        return pipe_no_wait(fd)

    def pipe_non_blocking_is_error_blocking(ex):
        if not isinstance(ex, PortableBlockingIOError):
            return False
        from ctypes import GetLastError
        ERROR_NO_DATA = 232

        return (GetLastError() == ERROR_NO_DATA)

    PortableBlockingIOError = OSError
else:
    def pipe_non_blocking_set(fd):
        import fcntl
        fl = fcntl.fcntl(fd, fcntl.F_GETFL)
        fcntl.fcntl(fd, fcntl.F_SETFL, fl | os.O_NONBLOCK)
        return True

    def pipe_non_blocking_is_error_blocking(ex):
        if not isinstance(ex, PortableBlockingIOError):
            return False
        return True

    PortableBlockingIOError = BlockingIOError

To avoid reading incomplete data, I ended up writing my own readline generator (which returns the byte string for each line).

Its a generator so you can for example…

def non_blocking_readlines(f, chunk=1024):
    """
    Iterate over lines, yielding b'' when nothings left
    or when new data is not yet available.

    stdout_iter = iter(non_blocking_readlines(process.stdout))

    line = next(stdout_iter)  # will be a line or b''.
    """
    import os

    from .pipe_non_blocking import (
            pipe_non_blocking_set,
            pipe_non_blocking_is_error_blocking,
            PortableBlockingIOError,
            )

    fd = f.fileno()
    pipe_non_blocking_set(fd)

    blocks = []

    while True:
        try:
            data = os.read(fd, chunk)
            if not data:
                # case were reading finishes with no trailing newline
                yield b''.join(blocks)
                blocks.clear()
        except PortableBlockingIOError as ex:
            if not pipe_non_blocking_is_error_blocking(ex):
                raise ex

            yield b''
            continue

        while True:
            n = data.find(b'\n')
            if n == -1:
                break

            yield b''.join(blocks) + data[:n + 1]
            data = data[n + 1:]
            blocks.clear()
        blocks.append(data)

I have the original questioner’s problem, but did not wish to invoke threads. I mixed Jesse’s solution with a direct read() from the pipe, and my own buffer-handler for line reads (however, my sub-process – ping – always wrote full lines < a system page size). I avoid busy-waiting by only reading in a gobject-registered io watch. These days I usually run code within a gobject MainLoop to avoid threads.

def set_up_ping(ip, w):
# run the sub-process
# watch the resultant pipe
p = subprocess.Popen(['/bin/ping', ip], stdout=subprocess.PIPE)
# make stdout a non-blocking file
fl = fcntl.fcntl(p.stdout, fcntl.F_GETFL)
fcntl.fcntl(p.stdout, fcntl.F_SETFL, fl | os.O_NONBLOCK)
stdout_gid = gobject.io_add_watch(p.stdout, gobject.IO_IN, w)
return stdout_gid # for shutting down

The watcher is

def watch(f, *other):
print 'reading',f.read()
return True

And the main program sets up a ping and then calls gobject mail loop.

def main():
set_up_ping('192.168.1.8', watch)
# discard gid as unused here
gobject.MainLoop().run()

Any other work is attached to callbacks in gobject.

Things are a lot better in modern Python.

Here’s a simple child program, “hello.py”:

#!/usr/bin/env python3

while True:
    i = input()
    if i == "quit":
        break
    print(f"hello {i}")

And a program to interact with it:

import asyncio


async def main():
    proc = await asyncio.subprocess.create_subprocess_exec(
        "./hello.py", stdin=asyncio.subprocess.PIPE, stdout=asyncio.subprocess.PIPE
    )
    proc.stdin.write(b"bob\n")
    print(await proc.stdout.read(1024))
    proc.stdin.write(b"alice\n")
    print(await proc.stdout.read(1024))
    proc.stdin.write(b"quit\n")
    await proc.wait()


asyncio.run(main())

That prints out:

b'hello bob\n'
b'hello alice\n'

Note that the actual pattern, which is also by almost all of the previous answers, both here and in related questions, is to set the child’s stdout file descriptor to non-blocking and then poll it in some sort of select loop. These days, of course, that loop is provided by asyncio.

The select module helps you determine where the next useful input is.

However, you’re almost always happier with separate threads. One does a blocking read the stdin, another does wherever it is you don’t want blocked.

why bothering thread&queue? unlike readline(), BufferedReader.read1() wont block waiting for \r\n, it returns ASAP if there is any output coming in.

#!/usr/bin/python
from subprocess import Popen, PIPE, STDOUT
import io

def __main__():
    try:
        p = Popen( ["ping", "-n", "3", "127.0.0.1"], stdin=PIPE, stdout=PIPE, stderr=STDOUT )
    except: print("Popen failed"); quit()
    sout = io.open(p.stdout.fileno(), 'rb', closefd=False)
    while True:
        buf = sout.read1(1024)
        if len(buf) == 0: break
        print buf,

if __name__ == '__main__':
    __main__()

In my case I needed a logging module that catches the output from the background applications and augments it(adding time-stamps, colors, etc.).

I ended up with a background thread that does the actual I/O. Following code is only for POSIX platforms. I stripped non-essential parts.

If someone is going to use this beast for long runs consider managing open descriptors. In my case it was not a big problem.

# -*- python -*-
import fcntl
import threading
import sys, os, errno
import subprocess

class Logger(threading.Thread):
    def __init__(self, *modules):
        threading.Thread.__init__(self)
        try:
            from select import epoll, EPOLLIN
            self.__poll = epoll()
            self.__evt = EPOLLIN
            self.__to = -1
        except:
            from select import poll, POLLIN
            print 'epoll is not available'
            self.__poll = poll()
            self.__evt = POLLIN
            self.__to = 100
        self.__fds = {}
        self.daemon = True
        self.start()

    def run(self):
        while True:
            events = self.__poll.poll(self.__to)
            for fd, ev in events:
                if (ev&self.__evt) != self.__evt:
                    continue
                try:
                    self.__fds[fd].run()
                except Exception, e:
                    print e

    def add(self, fd, log):
        assert not self.__fds.has_key(fd)
        self.__fds[fd] = log
        self.__poll.register(fd, self.__evt)

class log:
    logger = Logger()

    def __init__(self, name):
        self.__name = name
        self.__piped = False

    def fileno(self):
        if self.__piped:
            return self.write
        self.read, self.write = os.pipe()
        fl = fcntl.fcntl(self.read, fcntl.F_GETFL)
        fcntl.fcntl(self.read, fcntl.F_SETFL, fl | os.O_NONBLOCK)
        self.fdRead = os.fdopen(self.read)
        self.logger.add(self.read, self)
        self.__piped = True
        return self.write

    def __run(self, line):
        self.chat(line, nl=False)

    def run(self):
        while True:
            try: line = self.fdRead.readline()
            except IOError, exc:
                if exc.errno == errno.EAGAIN:
                    return
                raise
            self.__run(line)

    def chat(self, line, nl=True):
        if nl: nl = '\n'
        else: nl = ''
        sys.stdout.write('[%s] %s%s' % (self.__name, line, nl))

def system(command, param=[], cwd=None, env=None, input=None, output=None):
    args = [command] + param
    p = subprocess.Popen(args, cwd=cwd, stdout=output, stderr=output, stdin=input, env=env, bufsize=0)
    p.wait()

ls = log('ls')
ls.chat('go')
system("ls", ['-l', '/'], output=ls)

date = log('date')
date.chat('go')
system("date", output=date)

My problem is a bit different as I wanted to collect both stdout and stderr from a running process, but ultimately the same since I wanted to render the output in a widget as its generated.

I did not want to resort to many of the proposed workarounds using Queues or additional Threads as they should not be necessary to perform such a common task as running another script and collecting its output.

After reading the proposed solutions and python docs I resolved my issue with the implementation below. Yes it only works for POSIX as I’m using the select function call.

I agree that the docs are confusing and the implementation is awkward for such a common scripting task. I believe that older versions of python have different defaults for Popen and different explanations so that created a lot of confusion. This seems to work well for both Python 2.7.12 and 3.5.2.

The key was to set bufsize=1 for line buffering and then universal_newlines=True to process as a text file instead of a binary which seems to become the default when setting bufsize=1.

class workerThread(QThread):
   def __init__(self, cmd):
      QThread.__init__(self)
      self.cmd = cmd
      self.result = None           ## return code
      self.error = None            ## flag indicates an error
      self.errorstr = ""           ## info message about the error

   def __del__(self):
      self.wait()
      DEBUG("Thread removed")

   def run(self):
      cmd_list = self.cmd.split(" ")   
      try:
         cmd = subprocess.Popen(cmd_list, bufsize=1, stdin=None
                                        , universal_newlines=True
                                        , stderr=subprocess.PIPE
                                        , stdout=subprocess.PIPE)
      except OSError:
         self.error = 1
         self.errorstr = "Failed to execute " + self.cmd
         ERROR(self.errorstr)
      finally:
         VERBOSE("task started...")
      import select
      while True:
         try:
            r,w,x = select.select([cmd.stdout, cmd.stderr],[],[])
            if cmd.stderr in r:
               line = cmd.stderr.readline()
               if line != "":
                  line = line.strip()
                  self.emit(SIGNAL("update_error(QString)"), line)
            if cmd.stdout in r:
               line = cmd.stdout.readline()
               if line == "":
                  break
               line = line.strip()
               self.emit(SIGNAL("update_output(QString)"), line)
         except IOError:
            pass
      cmd.wait()
      self.result = cmd.returncode
      if self.result < 0:
         self.error = 1
         self.errorstr = "Task terminated by signal " + str(self.result)
         ERROR(self.errorstr)
         return
      if self.result:
         self.error = 1
         self.errorstr = "exit code " + str(self.result)
         ERROR(self.errorstr)
         return
      return

ERROR, DEBUG and VERBOSE are simply macros that print output to the terminal.

This solution is IMHO 99.99% effective as it still uses the blocking readline function, so we assume the sub process is nice and outputs complete lines.

I welcome feedback to improve the solution as I am still new to Python.

I have created a library based on J. F. Sebastian’s solution. You can use it.

https://github.com/cenkalti/what

Working from J.F. Sebastian’s answer, and several other sources, I’ve put together a simple subprocess manager. It provides the request non-blocking reading, as well as running several processes in parallel. It doesn’t use any OS-specific call (that I’m aware) and thus should work anywhere.

It’s available from pypi, so just pip install shelljob. Refer to the project page for examples and full docs.

EDIT: This implementation still blocks. Use J.F.Sebastian’s answer instead.

I tried the top answer, but the additional risk and maintenance of thread code was worrisome.

Looking through the io module (and being limited to 2.6), I found BufferedReader. This is my threadless, non-blocking solution.

import io
from subprocess import PIPE, Popen

p = Popen(['myprogram.exe'], stdout=PIPE)

SLEEP_DELAY = 0.001

# Create an io.BufferedReader on the file descriptor for stdout
with io.open(p.stdout.fileno(), 'rb', closefd=False) as buffer:
  while p.poll() == None:
      time.sleep(SLEEP_DELAY)
      while '\n' in bufferedStdout.peek(bufferedStdout.buffer_size):
          line = buffer.readline()
          # do stuff with the line

  # Handle any remaining output after the process has ended
  while buffer.peek():
    line = buffer.readline()
    # do stuff with the line

I recently stumbled upon on the same problem I need to read one line at time from stream ( tail run in subprocess ) in non-blocking mode I wanted to avoid next problems: not to burn cpu, don’t read stream by one byte (like readline did ), etc

Here is my implementation https://gist.github.com/grubberr/5501e1a9760c3eab5e0a it don’t support windows (poll), don’t handle EOF, but it works for me well

This is a example to run interactive command in subprocess, and the stdout is interactive by using pseudo terminal. You can refer to: https://stackoverflow.com/a/43012138/3555925

#!/usr/bin/env python
# -*- coding: utf-8 -*-

import os
import sys
import select
import termios
import tty
import pty
from subprocess import Popen

command = 'bash'
# command = 'docker run -it --rm centos /bin/bash'.split()

# save original tty setting then set it to raw mode
old_tty = termios.tcgetattr(sys.stdin)
tty.setraw(sys.stdin.fileno())

# open pseudo-terminal to interact with subprocess
master_fd, slave_fd = pty.openpty()

# use os.setsid() make it run in a new process group, or bash job control will not be enabled
p = Popen(command,
          preexec_fn=os.setsid,
          stdin=slave_fd,
          stdout=slave_fd,
          stderr=slave_fd,
          universal_newlines=True)

while p.poll() is None:
    r, w, e = select.select([sys.stdin, master_fd], [], [])
    if sys.stdin in r:
        d = os.read(sys.stdin.fileno(), 10240)
        os.write(master_fd, d)
    elif master_fd in r:
        o = os.read(master_fd, 10240)
        if o:
            os.write(sys.stdout.fileno(), o)

# restore tty settings back
termios.tcsetattr(sys.stdin, termios.TCSADRAIN, old_tty)

This solution uses the select module to “read any available data” from an IO stream. This function blocks initially until data is available, but then reads only the data that is available and doesn’t block further.

Given the fact that it uses the select module, this only works on Unix.

The code is fully PEP8-compliant.

import select


def read_available(input_stream, max_bytes=None):
    """
    Blocks until any data is available, then all available data is then read and returned.
    This function returns an empty string when end of stream is reached.

    Args:
        input_stream: The stream to read from.
        max_bytes (int|None): The maximum number of bytes to read. This function may return fewer bytes than this.

    Returns:
        str
    """
    # Prepare local variables
    input_streams = [input_stream]
    empty_list = []
    read_buffer = ""

    # Initially block for input using 'select'
    if len(select.select(input_streams, empty_list, empty_list)[0]) > 0:

        # Poll read-readiness using 'select'
        def select_func():
            return len(select.select(input_streams, empty_list, empty_list, 0)[0]) > 0

        # Create while function based on parameters
        if max_bytes is not None:
            def while_func():
                return (len(read_buffer) < max_bytes) and select_func()
        else:
            while_func = select_func

        while True:
            # Read single byte at a time
            read_data = input_stream.read(1)
            if len(read_data) == 0:
                # End of stream
                break
            # Append byte to string buffer
            read_buffer += read_data
            # Check if more data is available
            if not while_func():
                break

    # Return read buffer
    return read_buffer

I also faced the problem described by Jesse and solved it by using “select” as Bradley, Andy and others did but in a blocking mode to avoid a busy loop. It uses a dummy Pipe as a fake stdin. The select blocks and wait for either stdin or the pipe to be ready. When a key is pressed stdin unblocks the select and the key value can be retrieved with read(1). When a different thread writes to the pipe then the pipe unblocks the select and it can be taken as an indication that the need for stdin is over. Here is some reference code:

import sys
import os
from select import select

# -------------------------------------------------------------------------    
# Set the pipe (fake stdin) to simulate a final key stroke
# which will unblock the select statement
readEnd, writeEnd = os.pipe()
readFile = os.fdopen(readEnd)
writeFile = os.fdopen(writeEnd, "w")

# -------------------------------------------------------------------------
def getKey():

    # Wait for stdin or pipe (fake stdin) to be ready
    dr,dw,de = select([sys.__stdin__, readFile], [], [])

    # If stdin is the one ready then read it and return value
    if sys.__stdin__ in dr:
        return sys.__stdin__.read(1)   # For Windows use ----> getch() from module msvcrt

    # Must finish
    else:
        return None

# -------------------------------------------------------------------------
def breakStdinRead():
    writeFile.write(' ')
    writeFile.flush()

# -------------------------------------------------------------------------
# MAIN CODE

# Get key stroke
key = getKey()

# Keyboard input
if key:
    # ... do your stuff with the key value

# Faked keystroke
else:
    # ... use of stdin finished

# -------------------------------------------------------------------------
# OTHER THREAD CODE

breakStdinRead()

Try wexpect, which is the windows alternative of pexpect.

import wexpect

p = wexpect.spawn('myprogram.exe')
p.stdout.readline('.')               // regex pattern of any character
output_str = p.after()

On Unix-like systems and Python 3.5+ there’s os.set_blocking which does exactly what it says.

import os
import time
import subprocess

cmd = 'python3', '-c', 'import time; [(print(i), time.sleep(1)) for i in range(5)]'
p = subprocess.Popen(cmd, stdout=subprocess.PIPE)
os.set_blocking(p.stdout.fileno(), False)
start = time.time()
while True:
    # first iteration always produces empty byte string in non-blocking mode
    for i in range(2):    
        line = p.stdout.readline()
        print(i, line)
        time.sleep(0.5)
    if time.time() > start + 5:
        break
p.terminate()

This outputs:

1 b''
2 b'0\n'
1 b''
2 b'1\n'
1 b''
2 b'2\n'
1 b''
2 b'3\n'
1 b''
2 b'4\n'

With os.set_blocking commented it’s:

0 b'0\n'
1 b'1\n'
0 b'2\n'
1 b'3\n'
0 b'4\n'
1 b''

Here is a module that supports non-blocking reads and background writes in python:

https://pypi.python.org/pypi/python-nonblock

Provides a function,

nonblock_read which will read data from the stream, if available, otherwise return an empty string (or None if the stream is closed on the other side and all possible data has been read)

You may also consider the python-subprocess2 module,

https://pypi.python.org/pypi/python-subprocess2

which adds to the subprocess module. So on the object returned from “subprocess.Popen” is added an additional method, runInBackground. This starts a thread and returns an object which will automatically be populated as stuff is written to stdout/stderr, without blocking your main thread.

Enjoy!