Sean Voisen

Learning FRP with Bacon.js by way of Underscore

December 30, 2014

Functional Reactive Programming (FRP) can be difficult subject matter to grasp, particularly for programmers new to both functional and reactive concepts.

Using it effectively requires something of a radical shift not only in the way you think about both the architecture of your program, but also in the very act of computation itself. While traditional programming encourages thinking of computation as the linear execution of statements in the standard Von Neumann style, FRP – particularly the reactive part of FRP – encourages a higher-level, more abstract approach. In reactive programming, thinking about computation requires, first and foremost, thinking about the flow of data through a system.

Having explored FRP on and off for the past year or so, I continue to remain enthralled by the concept and evangelize for it. I do this because I remain convinced that – especially for developers focused on software domains fraught with the perils of asynchronicity (for instance, user interfaces) – it can lead to cleaner, easier-to-understand, and less error-prone code.

Because of this, I recently gave a short presentation introducing Functional Reactive Programming to our small group of engineers in XD at Adobe. My goal was to introduce the concept using a single programming language (JavaScript) and a single FRP library (Bacon.js), through analogy and comparison with a library that most of the attendees were already familiar with, namely Underscore.

The beauty of using Underscore lies in the fact that it provides many of the fundamental primitives for functional programming in a relatively easy-to-use package. In our group at Adobe, we regularly employ Underscore (or it’s competitor, Lo-Dash) in many of the JavaScript-based applications we develop. Most JavaScript developers are already familiar with its varied and many uses.

To start, let’s consider one of the most fundamental: map.

Map and EventStream

Much of functional programming involves the manipulation of data stored in a series of lists. Lists, often stored as arrays, are the most basic and foundational tool of the functional programmer. Underscore provides a wide variety of methods for working with lists (the documentation for these methods is filed under “arrays” and “collections”).

Consider the following array:

var arr = [1, 2, 3, 4, 5];

Let’s say we want to square each number in the array. The functional way to accomplish this task requires applying a function to the array that in turn applies a squaring function to each item in the array sequentially. Developers familiar with Underscore already know that such a function exists. This function is called map.

var squared =, function(x) {
  return x * x;

In FRP libraries like Bacon.js, an equivalent map primitive exists. But rather than operate on arrays, Bacon’s primitives operate on and return Observable objects like EventStreams.

What, exactly, is an EventStream? I wrote about this topic in more detail in a previous blog post. But, perhaps the easiest way to think about an EventStream is as an array that changes over time. With an EventStream, new items may be added to the stream at any time (that is, asynchronously) until the stream ends, and you can choose to be notified whenever this occurs.

For all of the subsequent examples, we will use Bacon.js to create an EventStream that emits the items in our array every 500ms. We will use the Bacon.sequentially method to do so. (It may help to note that EventStreams need not necessarily emit events at regular intervals, but I’m using these clock-like EventStreams for sake of simplicity.)

var stream = Bacon.sequentially(500, arr);

Once we have a an EventStream, we can use Bacon’s map method to create a new stream that squares each item as it arrives and then logs that value: {
  return x*x;
}).onValue(function(val) {

It may help to try this JSFiddle, which shows both the Underscore and Bacon.js examples in action.


Of course, many other primitives besides map exist. For instance, if we wanted to sum all of the items in the array we could use Underscore’s reduce function:

var sum = _.reduce(arr, function(x, acc) {
  return acc + x;
}, 0);

In Bacon.js, we have two primitives analogous to reduce. The first is fold.

stream.fold(0, function(x, acc) {
  return x + acc;
}).onValue(function(val) {

The challenge with fold is that it waits for the stream to end before a value is returned. In this example, the stream with 5 elements being pushed out every 500ms means that we have to wait the full 2.5s before onValue is executed. We can get around this problem by using the scan method instead, which will execute onValue as the sum is calculated in real-time:

stream.scan(0, function(x, acc) {
  return x + acc;
}).onValue(function(val) {

And, again, here is the JSFiddle showing it all in action.


“Zipping” is the act of taking two arrays and, item by item in alternating sequence, combining them into a single array. Consider the two arrays:

var arr1 = [1, 2, 3, 4, 5];
var arr2 = ['a', 'b', 'c', 'd', 'e'];

The Underscore code to produce the zipped combination of the two arrays is as follows:

var zipped =, arr2);
// Contains [1, 'a', 2, 'b', 3, 'c', 4, 'd', 5, 'e']

Bacon.js of course has an analogous zip function that produces a zipped EventStream, with the caveat that because EventStreams operate over time, it will wait until it has new items from both streams before outputting a value in the new stream:

var stream1 = Bacon.sequentially(500, arr1);
var stream2 = Bacon.sequentially(500, arr2); {

Here’s the JSFiddle.


Underscore also provides a filter method that takes a single array, and given a predicate function, outputs a new array that contains only the items for which the predicate holds true. For instance, if we wanted to filter an array containing all the numbers from 1-10 to an array containing only the even values:

var arr = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10];
var evens = _.filter(arr, function(x) {
  return x % 2 == 0;

Again, Bacon.js provides an analogous method operating on EventStreams:

var stream = Bacon.sequentially(500, arr);
stream.filter(function(x) {
  return x % 2 == 0;
}).onValue(function(val) {

And here’s the JSFiddle.

Creating EventStreams

Bacon.js provides a wide variety of ways for producing EventStreams. Bacon.sequentially is just the tip of the iceberg. All of the methods are too numerous to list here (see the documentation), but here are a few of the most useful:


The $.asEventStream method creates an EventStream from jQuery objects that emit events. This handy for quickly creating EventStreams from sources of user input, like buttons or text fields.

var stream = $('#myButton').asEventStream('click');


fromArray operates similarly to Bacon.sequentially, but will emit all items in an array immediately (as if Bacon.sequentially were provided with an interval of 0).

var arr = [1, 2, 3, 4, 5];
var stream = Bacon.fromArray(arr);


fromCallback will create an EventStream that will emit one event and then end the stream when a given callback is called. This can be quite useful for wrapping existing methods that expect a callback into something more friendly to a FRP-based approach. For instance, consider the following example that creates an EventStream from the setTimeout function

Bacon.fromCallback(function(callback) {
  setTimeout(function() {
  }, 1000);
}).onValue(function(val) {


Finally, fromPromise creates EventStreams from a Promise object. The stream will emit a single event when the promise succeeds, an error event if the promise fails, and will end immediately thereafter in either case. Here’s an example that creates an EventStream from a jQuery Ajax call to the Wikipedia API:

function doSearch(query) {
  var url = ''
    + '&format=json' 
    + '&search=' + encodeURI(query);
  return Bacon.fromPromise($.ajax({
doSearch('Aardvark').onValue(function(val) {

Here’s the JSFiddle for all of the previous examples.


Hopefully this short post shows that diving into FRP is not as difficult as it first may seem; many developers are already familiar with many of the concepts used in FRP through other functional libraries like Underscore.

Becoming familiar with the functional building blocks that FRP libraries like Bacon.js provide is an important first step to working effectively with FRP. The next step, of course, requires thinking in a functional reactive way. That is, how do we use these building blocks to make something truly useful? I provided one example in a previous blog post. In upcoming posts, I hope to expand on this with a few more. Stay tuned.

More posts in the archives →