Click here if you are not redirected.

A to Z, F15

Twitter Bots (and server-side programming)

Examples (node basics)

Node basics

Node.js is a JavaScript framework for writing server-side applications. In its simplest form it allows you to trigger small JavaScript programs from the command line without any browser involved. For example, assuming node is installed if you write a JavaScript program in a file called hello.js:

console.log('Node works!');

And then execute it from terminal with node hello.js you’ll see:

$ node hello.js
Node works!

Node Package Manager (npm) comes with node and allows you to install and work with node packages. Node packages are like libraries or add-ons and you can work with them in the same way that you might use a Processing or JavaScript library.

Though it’s not required, it’s generally good practice to include a package.json file with each node project. This file stores metadata about your project like author, license, as well as any dependencies on other node packages. You can manually create the file, but the command npm init will prompt you with a series of questions.

$ npm init

After answering the prompts you’ll see something like the following in a new package.json file:

  "name": "project-name",
  "version": "1.0.0",
  "description": "This is a test project.",
  "main": "server.js",
  "scripts": {
    "test": "echo \"Error: no test specified\" && exit 1"
  "author": "Daniel Shiffman",
  "license": "ISC"

Here is a link to documentation of package.json contents.

Now let’s say you are creating a node application where you want to use the Twit package. Twit is a package for connecting to the Twitter API and I’ll user it in examples you’ll find later on this page. To install a package, simply type npm install packagename. You can also add flags which can modify how a package is installed. One that I’ll use here is --save. This flag saves a reference to the package to package.json which can come in handy later when deploying your application.

$ npm install Twit --save

Express basics

Express is a popular, simple web framework for node. It includes hooks for the usual things you want to do with a web server, like hosting files and getting query input from a user. Although Express is not necessary for the ultimate goal of writing twitter bots (no need to be a web server), I’m going to take a little time on this page to walk through some basics related to other kinds of server-side functionality you might need for projects in this course.

First thing you need to do is install express.

$ npm install express --save

Once it’s installed, the way to access code from a node package is through the require() function.

var express = require('express');

With a reference to express, you can then create an express “app”:

var app = express();

The app is now a variable that holds an express instance, and allows you to implement functionality on that instance. For example, you might listen to conenctions on a certain port. The callback is triggered when the

var server = app.listen(3000);

You can also serve static files. For example, if you have a p5.js sketch (with an index.html file) and you place it in a folder called public, you can say:


Now anytime you enter the url to your server (if you are running it on your machine, this will be http://localhost:3000), the contents of the public folder will be hosted. (By the way, this is exactly what the p5.js editor is doing when you hit the run button.)

Routes with express

Beyond hosting static files, one of the most useful things you can do with server-side programming is execute different blocks of code based on the users’ “route”. A route is a path on the server like With static files, this is just a folder structure, but new possibilities open up when you programatically handle a route. For example, the following code specifies a function to call whenever a user goes to

app.get('/someroute', sayHello);

With the above code, you then need to do define the sayHello() function (which of course could have been named anything you want.)

function sayHello(request, response) {

Notice how the callback sayHello() is defined with two parameters: request and response. These variables refer to the HTTP request-response protocol. The user made a request and the server provides a response.

So in the above example, you can see that the server is just sending back the work “Hello!” in the response. If you passed a variable full of HTML back through send() you’d be serving up a webpage (generated programmtically!).

You can also look at the data associated with the user’s request. For example, if the request was made with a query string like, the value “Olympia” can be accessed via the query property of request.

function sayHello(request, response) {
  var name =;
  res.send("Hello " + name + "!");

RESTful routes

REST (Representational_state_transfer) is a common style of web architecture that is used by many APIs. For example, as you may have seeen in previous examples, a request to the Wordnik API looks something like: Note how this is different than using a url query string like Instead of a query string, the API pulls out commands (“definitions”) and parameters (“unicorn”) from the route itself. This translates to “Please send the definitons for the word unicorn.”

You can implement this style in node using app.get() as above. The difference is the following.

  1. “Variables” (i.e. values that are filled by the user) and notated with a colon (“:”).
  2. Those variables are accessed via the params property of request rather than query.
app.get('/hello/:name/:age', sayHello);

A valid url for the above might then be: You could handle the above saying:

function sayHello(request, response) {
  var name =;
  var age = request.params.age;
  res.send("Hello " + name + ", " + age + ' years old!');

Making your own API

The scenarios above demonstrate how the code works, but don’t really produce any useful results. Let’s take a scenario where you have a server that analyzes a massive corpus of text which you want to visualize from the browser. In previous weeks, we looked at analyzing text right in the client JavaScript code itself, but this was really only well-suited for smaller datasets. If the server handles it, the client can be saved a lot of computation time and simple “query” the results from the server and visualize them.

Before I get to the actual server code, one thing that may prove useful is looking at how node “modules” work. In client-side JavaScript, if you wanted to put some code in another JS file, you simply made something called, say, file.js, typed some code and included a reference in the HTML file. With a node app, there is no HTML file gluing everything together. Instead the “main” JS file (maybe called “server.js”) has to refer to another file using require(). For example, the following code assumes a file called utilities.js.

var utilities = require('./utilities');

The objects and functions you want to have access to must be included inside something called module.exports.

module.exports = {
  Concordance: function() {
    // Everything from week 5 concordance example

Now in your made server.js you can then create a Concordance object, i.e.

var wordcounts = new utilities.Concordance();

So, let’s say the node app then reads in lots of text and fills the wordcounts object on startup. (Files can be read using the node file system module.) Then when the user requests a route, say “all”:

// Route for sending all the concordance data
app.get('/all', showAll);

The response can be the JavaScript object wordcounts itself!

function showAll(req, res) {
  // Send the entire concordance

The magic of using express is that the send() function will automatically turn the JavaScript object passed into JSON! The what the end user sees is:

Now, a p5 sketch that is stored in, say, the statically hosted “public” folder can query this newly made API!

function setup() {
  loadJSON('/all', gotData);

function gotData(data) {
  var words = data.keys;
  for (var i = 0; i < words.length; i++) {
    var word = words[i];
    var count = data.dict[word];
    // Do something with the data!


Another scenario server-side programming can be useful for is scraping HTML from another site. You can’t make an HTTP request to another server from client-side JavaScript, but you can ask your own server to make this request for you! To do this, you can use the http node package.

var request = require('request');

You could specify a route and get the url via a query string. For example

// A router to load a URL
app.get('/load', loadURL);

// Callback
function loadURL(req, res) {
  // Get the URL from the user
  var url = req.query.url;

  // Execute the HTTP Request
  request(url, loaded);

  // Callback for when the request is complete
  function loaded(error, response, body) {
    // Check for errors
    if (!error && response.statusCode == 200) {
      // The raw HTML is in body
    } else {

Note how there are nested callbacks in the code above. First loadURL() runs when the user requests the “load” route. Then the HTTP request is made with request(). When that request is finished the loaded() callback runs and sends back the body parameter, which is the raw HTML content itself.

Now this can be called directly from p5’s loadStrings() method!

function setup() {
  loadStrings('load?url=', gotData);

The Twitter API

Some APIs require a level of authentification that can’t be done from client-side JavaScript. A prime example of this is the Twitter API which requires something known as OAuth. For basic API querying, you don’t need to go to deep into the inner workings of authenticating. With node, you can pick one of many node packages that help you connet and query the API. The one I’m using for these examples is called Twit.

Before you can use Twit, you need to follow a few steps.

  1. Create a twitter account. You might already have one, but depending on what you are doing, you may want to create a new account to associate with your app. For example, I am using this account for all of these examples.

  2. Create an “APP”. This can be done by on the manage your apps developer page. One tricky thing here is that to do this you are required to associate a phone number with the account in order to use the API. You can temporarily enter your own number or else sign up for a free one with Twilio or Google Voice.

To create the app, you’ll need to enter an app name, description, and associated website (can be anything for now). You can leave the “callback” field blank.

  1. Once the app is created, navigate to “Keys and Access Tokens”. There you’ll find your “Consumer Key (API Key)” and “Consumer Secret (API Secret)”.

Also, scroll down and click the “Generate Access Token” button. Now you also have an “Access Token” and “Access Token Secret”.

Don’t worry, if you ever post these keys somewhere by accident (like in an image file on a tutorial) you can always regenerate new ones.

That’s it, you’re done and ready to use Twit!

Authenticating with Twit

To make any calls to the API you first need to install Twit and then include the package with require().

$ npm install Twit --save
var Twit = require('twit');

And then you’ll need to authenticate with all those secret keys. There are a variety of ways to do this. The easiest is just typing your keys into your code:

var T = new Twit({
  consumer_key:         'YOURCONSUMERKEY',
  consumer_secret:      'YOURCONSUMERSECRET',
  access_token:         'YOURACCESSTOKEN',
  access_token_secret:  'YOURACCESSTOKENSECRET'

This may cause you some problems down the road, however, if you want to publish your code. One way you can get around this is by putting your keys in a separate file called, say, config.js.

module.exports = {
  consumer_key:         'YOURCONSUMERKEY',
  consumer_secret:      'YOURCONSUMERSECRET',
  access_token:         'YOURACCESSTOKEN',
  access_token_secret:  'YOURACCESSTOKENSECRET'

Now you can access this also with require().

var config = require('./config.js');
var T = new Twit(config);

This way if you want to share or publish your code, you can do so leaving out the config.js file. This is what I’ve done in this repo here, though I’ve included a config-empty.js file with some comments on how to enter the tokens.

Querying the Twitter API

Full documentation of the various Twit methods is available on github, but I’ll highlight what you need here for a basic p5.js sketch or bot.

There are three main calls you can make with Twit: get(), post(), and stream().

get() is useful for looking at particular tweets. For example, if you want to just search for tweets that match a given keyword, are from a particular user, etc. you can use get(). The following asks for the 10 most recent tweets with the keyword JavaScript. The callback for receiving the data is gotData().

T.get('search/tweets', { q: 'JavaScript', count: 10 }, gotData);

All of the arguments you can pass to the search can be found here. In addition to the search text, you can search for tweets from a particular location or from a certain time period, and more.

What you get back is just a lot of JSON as outlined here. So for example, if you look at the text of the tweets you would say:

function gotData(err, data) {
  var tweets = data.statuses;
  for (var i = 0; i < tweets.length; i++) {

The post() function is used to post an actual tweet itself.'statuses/update', { status: 'I am tweeting via the API!' }, tweeted);

The tweeted() callback is where you can check whether the tweet was successfully posted or not.

Included in this week’s examples is an example that expands on the above. Features include (1) searching the Twitter API and returning the results as JSON for client-side JavaScript to parse and (2) receiving a tweet via a URL query string and passing it along via the API to be posted.

Writing a bot

There’s not much left to say about writing a bot. All of the pieces you need are in the above sections. And none of the additional complexity of a web server via express is needed. The bot is simply just a process that executes every so often and tweets something based on some sort of algorithm or data source.

The easiest way to “schedule” a bot is to use JavaScript’s setInterval() function. For example, take the following:

// Set up Twit
var Twit = require('twit');
var config = require('./config.js');
var T = new Twit(config);

// Once every hour
setInterval(tweeter, 60*60*1000);

With the above code, the function tweeter() is triggered once per hour (an hour is 60 minutes, a minute is 60 seconds, a second is 1,000 milliseconds). All that is left to do is have tweeter() tweet something. For example:

// A function that tweets a random number!
function tweeter() {
  var num = Math.floor(Math.random()*100);
  var tweet = 'Here\'s a random number between 0 and 100: ' + num;'statuses/update', { status: tweet }, tweeted);

That’s it! Your first twitter bot. Of course you might want to do something more involved, like generate from a markov chain, or mix up text from some other API, etc. But for any of these ideas the above structure will work.

Here’s an example of a twitter bot using a Context-Free Grammar to generate tweets.

Using the streaming API

The above bot scenario involved tweeting every N milliseconds. This is what you think of when you think of some sort of autonomous robot tweeting. It operates on its own, never gets tired, and tweets at a precise time. But it is also possible to create a bot that participates in the social activity of twitter itself. This can be accomplished using stream(). The Twitter streaming API allows you to execute callbacks when certain events occur – like when someone follows you.

There are different kinds of streams – a “public” stream, a “user” stream, and a “site” stream (documentation). Here, I’ll use the “user” stream to demonstrate a “follow event”.

// A user stream
var stream ='user');

// When someone follows the user
stream.on('follow', followed);

// In this callback we can see the name and screen name
function followed(event) {
  var name =;
  var screenName = event.source.screen_name;
  console.log('I was followed by: ' + name + ' ' + screenName);

Now you could just add a post() to tweet back "Thanks for following me!" + screenName!

You can also trigger an events whenever the user is mentioned in a tweet.

stream.on('tweet', tweetEvent);

Now this event is triggered both when the user is mentioned and when the user itself tweets. So if you want to only get @replies you could add a check in the callback like:

function tweetEvent(tweet) {
  var reply_to = tweet.in_reply_to_screen_name;
  // Check to see if this was, in fact, a reply to you
  if (reply_to === 'yourusername') {
    // Get the username and content of the tweet
    var name = tweet.user.screen_name;
    var txt = tweet.text;
    // Tweet back a reply?

Here’s an example of a bot that replies to all @mentions with the same exact text in reverse.

The same search I used in the get() examples can also be accessed as a stream. For example, to get a continuous stream of all tweets from a certain geolocation:

// A geo location
var sanfran = [ '-122.75', '36.8', '-121.75', '37.8' ]
var stream ='statuses/filter', { locations: sanfran })
stream.on('tweet', tweetEvent);