Skip to content

Latest commit

 

History

History
190 lines (137 loc) · 5.97 KB

README.rst

File metadata and controls

190 lines (137 loc) · 5.97 KB
Raw

Twitsocket, twitter + WebSockets = ♥

A twitter wall / live stream for your conference / event / topic of interest, integrated in your Django website.

How it works

Twitsocket has 3 components:

  • A client for twitter's streaming API which you can use to track keywords and specific accounts.
  • A websocket server to broadcast new tweets to every connected client.
  • A tiny Django app to store the tweets in a DB and render them.

Installation

pip install https://github.com/brutasse/django-twitsocket/tarball/master

Requirements:

  • Python >=2.4
  • Django >= 1.0
  • python-oauth2
  • Jquery >= 1.4 enabled in your templates.

Configuration

You need to register an OAuth consumer on twitter.com (read AND write if you want to maintain the Twitter list) and get a token for it (using tweepy and this script for instance).

Add twitsocket to your INSTALLED_APPS, run syncdb and add a few settings:

CONSUMER_SECRET = 'your consumer secret'
CONSUMER_KEY = 'your consumer key'

TOKEN_SECRET = 'your token secret'
TOKEN_KEY = 'your token key'

# Optional, in case of spam
BANNED_USERS = ('list', 'of', 'banned', 'users')

TRACK_KEYWORDS = ('europython', 'python', 'java')
TRACK_USERS = (15324940, 15389140) # Only IDs here

# The WebSocket port is mandatory, no trailing slash.
WEBSOCKET_SERVER = 'ws://ws.example.com:8888'
WEB_SERVER = 'http://example.com'

# The log file tells you when new clients and new tweets arrive.
LOGFILE = '/path/to/logfile.log'

Create a view somewhere that renders a template and add this to the content:

{% load twitsocket_tags %}

{% render_tweets 30 %}

{% websocket_client %}

Extra template tags

The following template tags are available. They are not required but may be used to display extra information about the stream.

  • count: {% count %} will output the number of tweets saved in the database. The counter is automatically incremented when new tweets are received.
  • top_users: {% top_users <num> %} will display the top <num> users, people who are tweeting the most about the topic.
  • top_tweets: {% top_tweets <num> %} will display the top <num> tweets, ordered by number of retweets.
  • retweet_switch: {% retweet_switch %} will display a "show retweets" / "hide retweets" button if users want to see only original tweets. The user's preference is stored in the browser's localStorage and restored when the page is reloaded.

Note

For the top_users and top_tweets tags, you need to run the top_tweets management command every once a while to refreshed the cached values. It's up to you to decide on the frequency since the operation can be rather intensive with a large number of tweets.

Running the websocket server

./manage.py websockets

Ideally, you should put this line in a bash script and run it with supervisord to restart it automatically when Twitter cuts the connection.

Bonus

There is a lister management command which you can use to maintain a Twitter list of people tweeting about the subject you're tracking.

If you want to use the command, create the list on twitter and add it to your settings (prepend you twitter username):

TWITTER_LIST = 'brutasse/europython2010'

Running manage.py lister will add everyone to the list. Run it every once a while, Twitter has a rate-limit (and a max. of 500 ppl on a list). Note that only people who send an 'original' tweet (not a retweet) will appear on the list.

Styling

This example CSS code will add a nice look to your tweets (WTFPL, modify it to fit your needs):

#tweets {
  font: normal 18px Georgia, Times, serif; width: 600px;
  margin: 10px auto;                       padding: 10px;
  border: 1px solid #ccc;                  background-color: #eee;
}
#tweets a { text-decoration: none;     color: #4096ee; }
#tweets a.username { color: #73880a;   font-weight: bold; }
#tweets a:hover { text-decoration: underline; }
#tweets .tweet { color: #444; }
#tweets .tweet img {
  display: block;          float: left;
  background-color: #fff;  border: 1px solid #bbb;
  padding: 3px;            margin-right: 10px;
}
#tweets .tweet p { margin: 0; padding: 0; float: left; width: 500px; }
#tweets .clear {
  clear: both;             border-bottom: 1px solid #ccc;
  margin-bottom: 10px;     padding-bottom: 10px;
  font-size: 0.8em;        color: #aaa;
  text-align: right;       text-shadow: 0 1px 0 #fff;
}
#tweets .rt { color: #d01f3c; font-weight: bold; padding-right: 15px; }
.notice {
  width: 610px;            text-shadow: 0 1px 0 #fff;
  margin: 10px auto;       background-color: #FFFFaa;
  padding: 5px;            border: 1px solid #eecc55;
  color: #555;             font-size: 0.8em;
}

The flash hack

As you may know, not all browsers support WebSockets. They are implemented in Safari, Chrome and Firefox 4. There is a clever hack involving Flash that implements WebSockets for older browsers. To enable it, simply use contrib.staticfiles (or django-staticfiles with django 1.2). Add it to your INSTALLED_APPS, configure STATIC_ROOT and STATIC_URL and run manage.py collectstatic.

Then add to your <head> block (assuming you've loaded twitsocket_tags):

<head>
    <title> ... whatever you have </title>
    ...
    {% flash_hack %}
</head>

Note that because of some cross-domain security concerns, the flash hack will only if the media files are served on the same domain name as the website itself. No media.example.com for serving static files.

TODO

  • i18n for websocket error messages.
  • Try to decouple the Twitter consumer and the WebSocket server. Maybe with Redis and its Pub/Sub mechanism.