notes.txt

var TwitterStreamChannels = require('twitter-stream-channels');
var credentials = require('your.twitter.credentials.json');

var client = new TwitterStreamChannels(credentials);

var channels = {
  "languages" : ['javascript','php','java','python','perl'],
  "js-frameworks" : ['angularjs','jquery','backbone','emberjs'],
  "web" : ['javascript','nodejs','html5','css','angularjs']
};

var stream = client.streamChannels(channels);

stream.on('channels/languages',function(tweet){
  console.log(tweet);//any tweet with 'javascript','php','java','python','perl'
});

stream.on('channels/js-frameworks',function(tweet){
  console.log(tweet);//any tweet with 'angularjs','jquery','backbone','emberjs'
});

stream.on('channels/web',function(tweet){
  console.log(tweet);//any tweet with 'javascript','nodejs','html5','css','angularjs'
});

stream.on('channels',function(tweet){
  console.log(tweet);//any tweet with any of the keywords above
});

//sugar methods
setTimeout(function(){
    //those methods will modify the events emitted like "channels/channelName", so will be needed to :
    // - inspect useless listener (by the lib) to remove them
    // - update the needed listener for the new channels added (by the user)
    stream.addChannels(someChannels);
    stream.removeChannels(arrayOfChannelNames);
    stream.setChannels(someChannels);
},30000);

//also forward basic events like :
- connected
- disconnected
- reconnect
- tweet -> forward to channels (is that really that ? watch out for duplicates, normalization/denormalization)

//whatever needs to close stream some time
setTimeout(function(){
  stream.stop();//closes the stream connected to Twitter 
},100000);

----------

client.streamChannels(channels);

channels needs to be normalized to be passed to the track param
but also need to be kept in memory when processing tweet.text to choose to which event to emit

stream.on('channels/nameOfChannel')

before adding the listener check if the channel exists

add $keywords in tweets
add $channel in tweets

---------

for full project (next)

//finally, joining and leaving channels may be a bad idea, since if the user switch between tabs, he will loose a timelapsed infos, maybe better to
//emit all channels from the server and listen on only one room from the client, passing for each tweet to which channel it belongs
//so, the frontend service will manage and persist all data between views all along the visit (and keep updating infos even if the user isn't on the tab)
//there will be some work on the frontend service on persistence + recovery from infos that wasn't processed (displayed) yet
use socket.io channels http://socket.io/docs/rooms-and-namespaces/  -  http://stackoverflow.com/questions/16423150/socket-io-subscribe-to-multiple-channels

stream.on('channels/languages',function(tweet){
  io.sockets.in('languages').emit('message', { room: room, tweet: tweet });
});

//better as

//for the emmit part from the server
for(var channel in channels){
  (function(channelName){
    stream.on('channels/'+channelName,function(tweet){
      io.sockets.in(channelName).emit(processTweet(tweet));//processTweet is here to remove all the none necessary parts of the tweet
    });
  })(channel);
}

//for the join part from the client
socket.leave('javascript');//onchanging route
socket.join('languages');//on landing on new controller

//check the exact syntax according to socket io version (0.9.* or 1.*)

---------

better with no join or leave room + optimize to send only once all the channels (TODO)
for(var channel in channels){
  (function(channelName){
    stream.on('channels/'+channelName,function(tweet){
      io.sockets.emit({channel:channelName,tweet:processTweet(tweet)});//processTweet is here to remove all the none necessary parts of the tweet + each response has a tweet AND the channel from where it comes
    });
  })(channel);
}

---------

I saw curious number in the unit-tests on the events, at first sight, I thought I had an optimization problem between :
channels
channels/name
keywords/name
Each one were taking more time than the other one.
In fact, it was only that even if I captured only 10 tweets, even more were passing by (not captured by the listener, which is normal in real life - you have to wait until a tweet "fits" to your demands)