Comparing Prototypal and Classical Inheritance

Programmers familiar with the classic model of inheritance are often confused by JavaScript’s prototypal inheritance when they first encounter it. To help illustrate the differences in the two models, I’ll implement what might appear to be the same thing in both Ruby and JavaScript.

In each example, I’ll define a class Foo and a class Bar which inherits from Foo. It’s worth noting that in JavaScript I am defining constructor functions, not classes (JavaScript has no class system). Then I’ll create two instances of Bar. With the first instance, baz, I will

  1. push the value 1 into its a property (an array)
  2. directly modify its a property to be the array [1,2,3]
  3. delete its a property entirely

After each step I will print out the state of each instance, at which point some drastic differences in the two programs will become apparent.

First, here’s a simple example of class inheritance and instantiation in Ruby:

And here’s what running that Ruby script outputs:

I create two instances, baz and qux. When I modify properties on baz, properties on qux are not changed. This makes sense.

Here is a comparable JavaScript program:

The JavaScript program has a different output due to the fundamental differences in the two inheritance models.

Here are the differences highlighted.

So, this raises a few questions:

  • Why does baz.a.push(1); also modify qux.a in JavaScript?
  • Subsequently, why does baz.a = [1,2,3]; not modify qux.a?
  • And why is baz.a still 1 after delete baz.a;?

Well, baz and qux do not have their own property a. Instead, they share a from their prototype.

  • When I call baz.a, JavaScript first looks at baz for a, then it begins traversing the prototype chain upwards looking for a property a until it finds one, or returns undefined if it cannot.
  • In the next step, I directly define a property a on baz which equals [1,2,3].
  • When I try to access baz.a, JavaScript finds a directly on baz and returns it.
  • Then, after delete baz.a, when baz.a is called again, JavaScript walks back up the prototype chain and finds the shared a on the prototype.

Unlike a class, which is like a blueprint for an instance, an object’s prototype is itself another object.

In the image above, baz and qux are created in the web inspector console and modified. Each instance has a visible __proto__ reference pointing to its prototype, an instance of Foo. The a property is visible on the prototype, and it is visible on baz after it is directly defined as [1,2,3].

A prototype, being just an object, can be accessed and modified at runtime.

In the image above, the prototype is directly modified at runtime.

I hope this helps shed some light on what makes prototypal inheritance different from classical inheritance. Though some programmers want to emulate classical inheritance patterns in JavaScript, it is best to embrace prototypal inheritance, which offers its own expressive power.

Further Reading