stompest 1.0.1

STOMP client library for Python including both synchronous and Twisted implementations.

stomp, stomper, stompest!

Stompest is a no-nonsense Stomp implementation for Python including both synchronous and Twisted clients.

Modeled after the Perl Net::Stomp module, the synchronous client is dead simple. It does not assume anything about your concurrency model (thread vs process) or force you to use it any particular way. It gets out of your way and lets you do what you want.

The Twisted client is a full-featured STOMP protocol client built on top of the stomper library. It supports destination-specific message handlers, concurrent message processing, graceful shutdown, "poison pill" error handling, and connection timeout.

This module is thoroughly unit tested and production hardened for the functionality used by Mozes: persistent queueing on ActiveMQ. Minor enhancements may be required to use this STOMP adapter with other brokers or certain feature sets like transactions or binary messages.

Examples
========

Simple Producer
---------------

from stompest.simple import Stomp

QUEUE = '/queue/simpleTest'

stomp = Stomp('localhost', 61613)
stomp.connect()
stomp.send(QUEUE, 'test message1')
stomp.send(QUEUE, 'test message2')
stomp.disconnect()

Simple Consumer
---------------

from stompest.simple import Stomp

QUEUE = '/queue/simpleTest'

stomp = Stomp('localhost', 61613)
stomp.connect()
stomp.subscribe(QUEUE, {'ack': 'client'})

while(True):
frame = stomp.receiveFrame()
print "Got message frame: %s" % frame
stomp.ack(frame)

stomp.disconnect()

Twisted Producer
----------------

import logging
import simplejson
from twisted.internet import reactor, defer
from stompest.async import StompConfig, StompCreator

class Producer(object):

QUEUE = '/queue/testIn'

def __init__(self, config=None):
if config is None:
config = StompConfig('localhost', 61613)
self.config = config

@defer.inlineCallbacks
def run(self):
#Establish connection
stomp = yield StompCreator(self.config).getConnection()
#Enqueue 10 messages
try:
for x in range(10):
stomp.send(self.QUEUE, simplejson.dumps({'count': x}))
finally:
#Give the reactor time to complete the writes
reactor.callLater(1, reactor.stop)

if __name__ == '__main__':
logging.basicConfig(level=logging.DEBUG)
Producer().run()
reactor.run()

Twisted Transformer
-------------------

import logging
import simplejson
from twisted.internet import reactor, defer
from stompest.async import StompConfig, StompCreator

class IncrementTransformer(object):

IN_QUEUE = '/queue/testIn'
OUT_QUEUE = '/queue/testOut'
ERROR_QUEUE = '/queue/testTransformerError'

def __init__(self, config=None):
if config is None:
config = StompConfig('localhost', 61613)
self.config = config

@defer.inlineCallbacks
def run(self):
#Establish connection
stomp = yield StompCreator(self.config).getConnection()
#Subscribe to inbound queue
headers = {
#client-individual mode is only supported in AMQ >= 5.2 but necessary for concurrent processing
'ack': 'client-individual',
#this is the maximum messages the broker will let you work on at the same time
'activemq.prefetchSize': 100,
}
stomp.subscribe(self.IN_QUEUE, self.addOne, headers, errorDestination=self.ERROR_QUEUE)

def addOne(self, stomp, frame):
"""
NOTE: you can return a Deferred here
"""
data = simplejson.loads(frame['body'])
data['count'] += 1
stomp.send(self.OUT_QUEUE, simplejson.dumps(data))

if __name__ == '__main__':
logging.basicConfig(level=logging.DEBUG)
IncrementTransformer().run()
reactor.run()

Twisted Consumer
----------------

import logging
import simplejson
from twisted.internet import reactor, defer
from stompest.async import StompConfig, StompCreator

class Consumer(object):

QUEUE = '/queue/testOut'
ERROR_QUEUE = '/queue/testConsumerError'

def __init__(self, config=None):
if config is None:
config = StompConfig('localhost', 61613)
self.config = config

@defer.inlineCallbacks
def run(self):
#Establish connection
stomp = yield StompCreator(self.config).getConnection()
#Subscribe to inbound queue
headers = {
#client-individual mode is only supported in AMQ >= 5.2 but necessary for concurrent processing
'ack': 'client-individual',
#this is the maximum messages the broker will let you work on at the same time
'activemq.prefetchSize': 100,
}
stomp.subscribe(self.QUEUE, self.consume, headers, errorDestination=self.ERROR_QUEUE)

def consume(self, stomp, frame):
"""
NOTE: you can return a Deferred here
"""
data = simplejson.loads(frame['body'])
print "Received msg with count %s" % data['count']

if __name__ == '__main__':
logging.basicConfig(level=logging.DEBUG)
Consumer().run()
reactor.run()

Features
========

Simple
------
Basically the same as Net::Stomp

Twisted
-------
* Graceful shutdown - On disconnect or error, the client stops processing new messages and waits for all outstanding message handlers to finish before issuing the DISCONNECT command
* Error handling - the client assumes that if an exception is not handled by a message handler, it's not necessarily safe to continue processing so it shutdowns down gracefully. By passing the errorDestination parameter to the subscribe() method, it can be configured to send a copy of offending message to a configured destination. This will allow you to avoid the "poison pill" scenario where the broker keeps redelivering a bad message infinitely.
* Fully unit-tested including a simulated STOMP broker
* Configurable connection timeout


Caveats
=======
* Does not support binary data - support for the content-length header is not (yet) implemented
* Tested with ActiveMQ versions 4.3 and above. Mileage may vary with other STOMP implementations

Twisted
-------
* Written before the advent of defer.inlineCallbacks so it could be simpler