Interprocess communication in Python

What is a good way to communicate between two separate Python runtimes? Thing's I've tried:

• reading/writing on named pipes e.g. os.mkfifo (feels hacky)
• dbus services (worked on desktop, but too heavyweight for headless)
• sockets (seems too low-level; surely there's a higher level module to use?)

My basic requirement is to be able to run python listen.py like a daemon, able to receive messages from python client.py. The client should just send a message to the existing process and terminate, with return code 0 for success and nonzero for failure (i.e. a two-way communication will be required.)

The multiprocessing library provides listeners and clients that wrap sockets and allow you to pass arbitrary python objects.

Your server could listen to receive python objects:

from multiprocessing.connection import Listener

address = ('localhost', 6000)     # family is deduced to be 'AF_INET'
listener = Listener(address, authkey=b'secret password')
conn = listener.accept()
print 'connection accepted from', listener.last_accepted
while True:
    msg = conn.recv()
    # do something with msg
    if msg == 'close':
        conn.close()
        break
listener.close()

Your client could send commands as objects:

from multiprocessing.connection import Client

address = ('localhost', 6000)
conn = Client(address, authkey=b'secret password')
conn.send('close')
# can also send arbitrary objects:
# conn.send(['a', 2.5, None, int, sum])
conn.close()

Nah, zeromq is the way to go. Delicious, isn't it?

import argparse
import zmq

parser = argparse.ArgumentParser(description='zeromq server/client')
parser.add_argument('--bar')
args = parser.parse_args()

if args.bar:
    # client
    context = zmq.Context()
    socket = context.socket(zmq.REQ)
    socket.connect('tcp://127.0.0.1:5555')
    socket.send(args.bar)
    msg = socket.recv()
    print msg
else:
    # server
    context = zmq.Context()
    socket = context.socket(zmq.REP)
    socket.bind('tcp://127.0.0.1:5555')
    while True:
        msg = socket.recv()
        if msg == 'zeromq':
            socket.send('ah ha!')
        else:
            socket.send('...nah')

From my experience, rpyc is by far the simplest and most elegant way to go about it.

Based on @vsekhar's answer, here is a Python 3 version with more details and multiple connections:

Server

from multiprocessing.connection import Listener

listener = Listener(('localhost', 6000), authkey=b'secret password')
running = True
while running:
    conn = listener.accept()
    print('connection accepted from', listener.last_accepted)
    while True:
        msg = conn.recv()
        print(msg)
        if msg == 'close connection':
            conn.close()
            break
        if msg == 'close server':
            conn.close()
            running = False
            break
listener.close()

Client

from multiprocessing.connection import Client
import time

# Client 1
conn = Client(('localhost', 6000), authkey=b'secret password')
conn.send('foo')
time.sleep(1)
conn.send('close connection')
conn.close()

time.sleep(1)

# Client 2
conn = Client(('localhost', 6000), authkey=b'secret password')
conn.send('bar')
conn.send('close server')
conn.close()

I would use sockets; local communication was strongly optimized, so you shouldn't have performance problems and it gives you the ability to distribute your application to different physical nodes if the needs should arise.

With regard to the "low-level" approach, you're right. But you can always use an higher-level wrapper depending on your needs. XMLRPC could be a good candidate, but it is maybe overkill for the task you're trying to perform.

Twisted offers some good protocol simple implementations, such as LineReceiver (for simple line based messages) or the more elegant AMP (which was, by the way, standardized and implemented in different languages).

Check out a cross-platform library/server called RabbitMQ. Might be too heavy for two-process communication, but if you need multi-process or multi-codebase communication (with various different means, e.g. one-to-many, queues, etc), it is a good option.

Requirements:

$ pip install pika
$ pip install bson # for sending binary content
$ sudo apt-get rabbitmq-server # ubuntu, see rabbitmq installation instructions for other platforms

Publisher (sends data):

import pika, time, bson, os

connection = pika.BlockingConnection(pika.ConnectionParameters('localhost'))
channel = connection.channel()
channel.exchange_declare(exchange='logs', type='fanout')

i = 0
while True:
    data = {'msg': 'Hello %s' % i, b'data': os.urandom(2), 'some': bytes(bytearray(b'\x00\x0F\x98\x24'))}
    channel.basic_publish(exchange='logs', routing_key='', body=bson.dumps(data))
    print("Sent", data)
    i = i + 1
    time.sleep(1)

connection.close()

Subscriber (receives data, can be multiple):

import pika, bson

connection = pika.BlockingConnection(pika.ConnectionParameters(host='localhost'))
channel = connection.channel()

channel.exchange_declare(exchange='logs', type='fanout')

result = channel.queue_declare(exclusive=True)
queue_name = result.method.queue

channel.queue_bind(exchange='logs', queue=queue_name)

def callback(ch, method, properties, body):
    data = bson.loads(body)
    print("Received", data)

channel.basic_consume(callback, queue=queue_name, no_ack=True)
channel.start_consuming()

Examples based on https://www.rabbitmq.com/tutorials/tutorial-two-python.html

I would use sockets, but use Twisted to give you some abstraction, and to make things easy. Their Simple Echo Client / Server example is a good place to start.

You would just have to combine the files and instantiate and run either the client or server depending on the passed argument(s).