Question

I have an isolated python script that simply captures data from Twitter’s streaming API and then on the receipt of each message, using redis pubsub it publishes to the channel “tweets”. Here is that script:

def main():
    username = "username"
    password = "password"
    track_list = ["apple", "microsoft", "google"]

    with tweetstream.FilterStream(username, password, track=track_list) as stream:
        for tweet in stream:
            text = tweet["text"]
            user = tweet["user"]["screen_name"]
            message = {"text": text, "user": user}
            db.publish("tweets", message)

if __name__ == '__main__':
    try:
        print "Started..."
        main()
    except KeyboardInterrupt:
        print '\nGoodbye!'

My server side socket.io implementation is done using django-socketio (based off of gevent-socketio) https://github.com/stephenmcd/django-socketio which simply provides a few helper decorators as well as a broadcast_channel method. Because it’s done in django, I’ve simply put this code in views.py simply so that they’re imported. My views.py code:

def index(request):
    return render_to_response("twitter_app/index.html", {
    }, context_instance=RequestContext(request))

def _listen(socket):
    db = redis.Redis(host="localhost", port=6379, db=0)
    client = db.pubsub()
    client.subscribe("tweets")
    tweets = client.listen()

    while True:
        tweet = tweets.next()
        tweet_data = ast.literal_eval(tweet["data"])
        message = {"text": tweet_data["text"], "user": tweet_data["user"], "type": "tweet"}
        socket.broadcast_channel(message)

@on_subscribe(channel="livestream")
def subscribe(request, socket, context, channel):
    g = Greenlet.spawn(_listen, socket)

The client side socket.io JavaScript simply connects and subscribes to the channel “livestream” and captures any received messages to that channel:

var socket = new io.Socket();
socket.connect();
socket.on('connect', function() {
    socket.subscribe("livestream");
});
socket.on('message', function(data) {
    console.log(data);
});

The obvious problem with this code is that each time a new user or browser window is opened to the page, a new _listen method is spawned and the tweets get subscribed to and broadcast for each user resulting in duplicate messages being received on the client. My question is, where would the proper place be to put the _listen method so that it’s only created once regardless of the # of clients? Also, keeping in mind that the broadcast_channel method is a method of a socket instance.

Answer 1

The problem was that I was using socket.broadcast_channel when I should have been using socket.send.