Node.js in Practice (2015)

Part 1. Node fundamentals

Chapter 4. Events: Mastering EventEmitter and beyond

This chapter covers

·        Using Node’s EventEmitter module

·        Managing errors

·        How third-party modules use EventEmitter

·        How to use domains with events

·        Alternatives to EventEmitter

Node’s events module currently includes just a single class: EventEmitter. This class is used throughout both Node’s built-in modules and third-party modules. It contributes to the overall architecture of many Node programs. Therefore it’s important to understand EventEmitter and how to use it.

It’s a simple class, and if you’re familiar with DOM or jQuery events, then you shouldn’t have much trouble understanding it. The major consideration when using Node is in error handling, and we’ll look at this in technique 21.

EventEmitter can be used in various ways—it’s generally used as a base class for solving a wide range of problems, from building network servers to architecting application logic. In view of the fact that it’s used as the basis for key classes in popular Node modules like Express, learning how it works can be useful for writing idiomatic code that plays well alongside existing modules.

In this chapter you’ll learn how to use EventEmitter to make custom classes, and how it’s used within Node and open source modules. You’ll also learn how to solve problems found when using EventEmitter, and see some alternatives to it.

4.1. Basic usage

To use EventEmitter, the base class must be inherited from. This section includes techniques for inheriting from EventEmitter and mixing it into other classes that already inherit from another base class.

Technique 19 Inheriting from EventEmitter

This technique demonstrates how to create custom classes based on EventEmitter. By understanding the principles in this technique, you’ll learn how to use EventEmitter, and how to better use modules that are built with it.


You want to use an event-based approach to solve a problem. You have a class that you’d like to operate when asynchronous events occur.

Web, desktop, and mobile user interfaces have one thing in common: they’re event-based. Events are a great paradigm for dealing with something inherently asynchronous: the input from human beings. To show how EventEmitter works, we’ll use a music player as an example. It won’t really play music, but the underlying concept is a great way to learn how to use events.


The canonical example of using events in Node is inheriting from EventEmitter. This can be done by using a simple prototype class—just remember to call EventEmitter’s constructor from within your new constructor.

The first listing shows how to inherit from EventEmitter.

Listing 4.1. Inheriting from EventEmitter


The combination of a simple constructor function and util.inherits is the easiest and most common way to create customized event-based classes. The next listing extends the previous listing to show how to emit and bind listeners using on.

Listing 4.2. Inheriting from EventEmitter

This might not seem like much, but suppose we need to do something else when play is triggered—perhaps the user interface needs to be updated. This can be supported simply by adding another listener to the play event. The following listing shows how to add more listeners.

Listing 4.3. Adding multiple listeners

Listeners can be removed as well. emitter.removeListener removes a listener for a specific event, whereas emitter.removeAllListeners removes all of them. You’ll need to store the listener in a variable to be able to reference it when removing a specific listener, which is similar to removing timers with clearTimeout. The next listing shows this in action.

Listing 4.4. Removing listeners

util.inherits works by wrapping around the ES5 method Object.create, which inherits the properties from one prototype into another. Node’s implementation also sets the superconstructor in the super_ property. This makes accessing the original constructor a lot easier—after usingutil.inherits, your prototype class will have access to EventEmitter through YourClass.super_.

You can also respond to an event once, rather than every time it fires. To do that, attach a listener with the once method. This is useful where an event can be emitted multiple times, but you only care about it happening a single time. For example, you could update listing 4.3 to track if the play event has ever been triggered:

musicPlayer.once('play', {

  this.audioFirstStarted = new Date();


When inheriting from EventEmitter, it’s a good idea to use events.EventEmitter .call(this) in your constructor to run EventEmitter’s constructor. The reason for this is because it’ll attach the instance to a domain if domains are being used. To learn more about domains, seetechnique 22.

The methods we’ve covered here—on, emit, and removeListener—are fundamental to Node development. Once you’ve mastered EventEmitter, you’ll find it cropping up everywhere: in Node’s built-in modules and beyond. Creating TCP/IP servers with net.createServer will return a server based on EventEmitter, and even the process global object is an instance of EventEmitter. In addition, popular modules like Express are based around EventEmitter—you can actually create an Express app object and call app.emit to send messages around an Express project.

Technique 20 Mixing in EventEmitter

Sometimes inheritance isn’t the right way to use EventEmitter. In these cases, mixing in EventEmitter may work.


This is an alternative option to technique 19. Rather than using EventEmitter as a base class, it’s possible to copy its methods into another class. This is useful when you have an existing class and can’t easily rework it to inherit directly from EventEmitter.


Using a for-in loop is sufficient for copying the properties from one prototype to another. In this way you can copy the necessary properties from EventEmitter.


This example might seem a little contrived, but sometimes it really is useful to copy EventEmitter’s properties rather than inherit from it in the usual way. This approach is more akin to a mixin, or multiple inheritance; see this demonstrated in the following listing.

Listing 4.5. Mixing in EventEmitter

One example of multiple inheritance in the wild is the Connect framework.[1] The core Server class inherits from multiple sources, and in this case the Connect authors have decided to make their own property copying method, shown in the next listing.

1 See

Listing 4.6. utils.merge from Connect

exports.merge = function(a, b){

  if (a && b) {

    for (var key in b) {

      a[key] = b[key];



  return a;


This technique may be useful when you already have a well-established class that could benefit from events, but can’t easily be a direct descendant of EventEmitter.

Once you’ve inherited from EventEmitter you’ll need to handle errors. The next section explores techniques for handling errors generated by EventEmitter classes.

4.2. Error handling

Although most events are treated equally, error events are a special case and are therefore treated differently. This section looks at two ways of handling errors: one attaches a listener to the error event, and the other uses domains to collect errors from groups of EventEmitter instances.

Technique 21 Managing errors

Error handling with EventEmitter has its own special rules that must be adhered to. This technique explains how error handling works.


You’re using an EventEmitter and want to gracefully handle when errors occur, but it keeps raising exceptions.


To prevent EventEmitter from throwing exceptions whenever an error event is emitted, add a listener to the error event. This can be done with custom classes or any standard class that inherits from EventEmitter.


To handle errors, bind a listener to the error event. The following listing demonstrates this by building on the music player example.

Listing 4.7. Event-based errors

This example is perhaps simple, but it’s useful because it should help you realize how EventEmitter handles errors. It feels like a special case, and that’s because it is. The following excerpt is from the Node documentation:

When an EventEmitter instance experiences an error, the typical action is to emit an error event. Error events are treated as a special case in Node. If there is no listener for it, then the default action is to print a stack trace and exit the program.

You can try this out by removing the 'error' handler from listing 4.7. A stack trace should be displayed in the console.

This makes sense semantically—otherwise the absence of an error handler would lead to potentially dangerous activity going unnoticed. The event name, or type as it’s referred to internally, has to appear exactly as error—extra spaces, punctuation, or uppercase letters won’t be considered an error event.

This convention means there’s a great deal of consistency across event-based error-handling code. It might be a special case, but it’s one worth paying attention to.

Technique 22 Managing errors with domains

Dealing with errors from multiple instances of EventEmitter can feel like hard work ... unless domains are used!


You’re dealing with multiple non-blocking APIs, but are struggling to effectively handle errors.


Node’s domain module can be used to centralize error handling for a set of asynchronous operations, and this includes EventEmitter instances that emit unhandled error events.


Node’s domain API provides a way of wrapping existing non-blocking APIs and exceptions with error handlers. This helps centralize error handling, and is particularly useful in cases where multiple interdependent I/O operations are being used.

Listing 4.8 builds on the music player example by using two EventEmitter descendants to show how a single error handler can be used to handle errors for separate objects.

Listing 4.8. Managing errors with domain

Domains can be used with EventEmitter descendants, networking code, and also the asynchronous file system methods.

To visualize how domains work, imagine that the callback wraps around your code, even when the code inside the callback triggers events that occur outside of it. Any errors that are thrown will still be caught by the domain. Figure 4.1 illustrates this process.

Figure 4.1. Domains help catch errors and handle them with an EventEmitter-style API.

Without a domain, any errors raised using throw could potentially place the interpreter in an unknown state. Domains avoid this and help you handle errors more gracefully.

Now that you know how to inherit from EventEmitter and handle errors, you should be starting to see all kinds of useful ways that it can be used. The next section broadens these techniques by introducing some advanced usage patterns and higher-level solutions to program structure issues relating to events.

4.3. Advanced patterns

This section offers some best practice techniques for solving structural issues found when using EventEmitter.

Technique 23 Reflection

Sometimes you need to dynamically respond to changes to an instance of an EventEmitter, or query its listeners. This technique explains how to do this.


You need to either catch when a listener has been added to an emitter, or query the existing listeners.


To track when listeners are added, EventEmitter emits a special event called new-Listener. Listeners added to this event will receive the event name and the listener function.


In some ways, the difference between writing good Node code and great Node code comes down to a deep understanding of EventEmitter. Being able to correctly reflect on EventEmitter objects gives rise to a whole range of opportunities for creating more flexible and intuitive APIs. One dynamic way of doing this is through the new-Listener event, emitted when listeners are added using the on method. Interestingly, this event is emitted by using EventEmitter itself—it’s implemented by using emit.

The next listing shows how to track newListener events.

Listing 4.9. Keeping tabs on new listeners

Even though 'a listener' is never explicitly emitted in this example, the newListener event will still fire. Since the listener’s callback function is passed as well as the event name, this is a great way to create simplified public APIs for things that require access to the original listener function. Listing 4.10 demonstrates this concept by automatically starting a timer when listeners for pulse events are added.

Listing 4.10. Automatically triggering events based on new listeners

We can go a step further and query EventEmitter objects about their listeners by calling emitter.listeners(event). A list of all listeners can’t be returned in one go, though. The entire list is technically available within the this._events object, but this property should be considered private. The listeners method currently returns an Array instance. This could be used to iterate over multiple listeners if several have been added to a given event—perhaps to remove them at the end of an asynchronous process, or simply to check if any listeners have been added.

In cases where an array of events is available, the listeners method will effectively return this._events[type].slice(0). Calling slice on an array is a JavaScript shortcut for creating a copy of an array. The documentation states that this behavior may change in the future, so if you really want to create a copy of attached listeners, then call slice yourself to ensure you really get a copy and not a reference to a data structure within the emitter instance.

Listing 4.11 adds a stop method to the Pulsar class. When stop is called, it checks to see if there are any listeners; otherwise, it raises an error. Checking for listeners is a good way to prevent incorrect usage, but you don’t have to do this in your own code.

Listing 4.11. Querying listeners

Pulsar.prototype.stop = function() {

  if (this.listeners('pulse').length === 0) {

    throw new Error('No listeners have been added!');



var pulsar = new Pulsar(500, 5);


Technique 24 Detecting and exploiting EventEmitter

A lot of successful open source Node modules are built on EventEmitter. It’s useful to spot where EventEmitter is being used and to know how to take advantage of it.


You’re working on a large project with several components and want to communicate between them.


Look for the emit and on methods whenever you’re using either Node’s standard modules or open source libraries. For example, the Express app object has these methods, and they’re great for sending messages within an application.


Usually when you’re working on a large project, there’s a major component that’s central to your problem domain. If you’re building a web application with Express, then the app object is one such component. A quick check of the source shows that this object mixes in EventEmitter, so you can take advantage of events to communicate between the disparate components within your project.

Listing 4.12 shows an Express-based example where a listener is bound to an event, and then the event is emitted when a specific route is accessed.

Listing 4.12. Reusing EventEmitter in Express

This might seem contrived, but what if the route were defined in another file? In this case, you wouldn’t have access to the app object, unless it was defined as a global.

Another example of a popular project built on EventEmitter is the Node Redis client ( Instances of RedisClient inherit from EventEmitter. This allows you to hook into useful events, like the error event, as shown in the next listing.

Listing 4.13. Reusing EventEmitter in the redis module

In cases where the route separation technique has been used to store routes in several files, you can actually send events by calling This allows route handlers to communicate back to the app object itself.

This might seem like a highly specific Express example, but other popular open source modules are also built on EventEmitter—just look for the emit and on methods. Remember that Node’s internal modules like the process object and net.create-Server inherit from EventEmitter, and well-written open source modules tend to inherit from these modules as well. This means there’s a huge amount of scope for event-based solutions to architectural problems.

This example also highlights another benefit of building projects around EventEmitter—asynchronous processes can respond as soon as possible. If the hello-alert event performs a very slow operation like sending an email, the person browsing the page might not want to wait for this process to finish. In this case, you can render the requested page while effectively performing a slower operation in the background.

The Node Redis client makes excellent use of EventEmitter and the author has written documentation for what each of the methods do. This is a good idea—if somebody joins your project, they may find it hard to get an overall picture of the events that are being used.

Technique 25 Categorizing event names

Some projects just have too many events. This technique shows how to deal with bugs caused by mistyped event names.


You’re losing track of the events in your program, and are concerned that it may be too easy to write an incorrect event name somewhere causing a difficult-to-track bug.


The easiest way to solve this problem is to use an object to act as a central dictionary for all of the event names. This creates a centralized location of each event in the project.


It’s hard to keep track of event names littered throughout a project. One way to manage this is to keep each event name in one place. Listing 4.14 demonstrates using an object to categorize event names, based on the previous examples in this chapter.

Listing 4.14. Categorizing event names using an object

Although EventEmitter is an integral part of Node’s standard library, and an elegant solution to many problems, it can be the source of a lot of bugs in larger projects where people may forget the name of a given event. One way around this is to avoid writing events as strings. Instead, an object can be used with properties that refer to the event name strings.

If you’re writing a reusable, open source module, you should consider making this part of the public API so it’s easy for people to get a centralized list of event names.

There are other observer pattern implementations that avoid using string event names to effectively type check events. In the next technique we’ll look at a few that are available through npm.

Although EventEmitter provides a wide array of solutions when working on Node projects, there are alternative implementations out there. The next section includes some popular alternatives.

4.4. Third-party modules and extensions

EventEmitter is essentially an observer pattern implementation. There are other interpretations of this pattern, which can help scale Node programs to run across several processes or over a network. The next technique introduces some of the more popular alternatives created by the Node community.

Technique 26 Alternatives to EventEmitter

EventEmitter has a great API and works well in Node programs, but sometimes a problem requires a slightly different solution. This technique explores some alternatives to EventEmitter.


You’re trying to solve a problem that doesn’t quite fit EventEmitter.


Depending on the exact nature of the problem you’re trying to solve, there are several alternatives to EventEmitter: publish/subscribe, AMQP, and js-signals are some popular alternatives with good support in Node.


The EventEmitter class is an implementation of the observer pattern. A related pattern is publish/subscribe, where publishers send messages that are characterized into classes to subscribers without knowing the details of the subscribers themselves.

The publish/subscribe pattern is often useful in cases where horizontal scaling is required. If you need to run multiple Node processes on multiple servers, then technologies like AMQP and ØMQ can help implement this. They’re both specifically designed to solve this class of problem, but may not be as convenient as using the Redis publish/subscribe API if you’re already using Redis.

If you need to horizontally scale across a distributed cluster, then an AMQP implementation like RabbitMQ ( will work well. The rabbitmq-nodejs-client ( module has a publish/subscribe API. The following listing shows a simple example of RabbitMQ in Node.

Listing 4.15. Using RabbitMQ with Node

ØMQ ( is more popular in the Node community. Justin Tulloss and TJ Holowaychuk’s zeromq.node module ( is a popular binding. The next listing shows just how simple this API is.

Listing 4.16. Using ØMQ with Node

var zmq = require('zmq');

var push = zmq.socket('push');

var pull = zmq.socket('pull');



console.log('Producer bound to port 3000');

setInterval(function() {

  console.log('sending work');

  push.send('some work');

}, 500);

pull.on('message', function(msg) {

  console.log('work: %s', msg.toString());


If you’re already using Redis with Node, then it’s worth trying out the Pub/Sub API ( Listing 4.17 shows an example of this using the Node Redis client (

Listing 4.17. Using Redis Pub/Sub with Node

Finally, if publish/subscribe isn’t what you’re looking for, then you may want to take a look at js-signals ( This module is a messaging system that doesn’t use strings for the signal names, and dispatching or listening to events that don’t yet exist will raise errors.

Listing 4.18 shows how js-signals sends and receives messages. Notice how signals are properties of an object, rather than strings, and that listeners can receive an arbitrary number of arguments.

Listing 4.18. Using Redis Pub/Sub with Node

js-signals provides a way of using properties for signal names, as mentioned in technique 25, but in this case the module will raise an error if an unregistered listener is dispatched or bound to. This approach is more like “strongly typed” events, and is very different from most publish/subscribe and event observer implementations.

4.5. Summary

In this chapter you’ve learned how EventEmitter is used through inheritance and multiple inheritance, and how to manage errors with and without domains. You’ve also seen how to centralize event names, how open source modules build on EventEmitter, and some alternative solutions.

What you should take away from this chapter is that although EventEmitter is usually used as a base class for inheritance, it’s also possible to mix it into existing classes. Also, although EventEmitter is a great solution to many problems and used throughout Node’s internals, sometimes other solutions are more optimal. For example, if you’re using Redis, then you can take advantage of its publish/subscribe implementation. Finally, EventEmitter isn’t without its problems; managing large amounts of event names can cause bugs, and now you know how to avoid this by using an object with properties that act as event names.

In the next chapter we’ll look at a related topic: streams. Streams are built around an event-based API, so you’ll be able to use some of these EventEmitter techniques there as well.