DEV Community

Shawn Reisner
Shawn Reisner

Posted on • Originally published at itnext.io on

Here's Why Mapping a Constructed Array in JavaScript Doesn't Work

Why Mapping a Constructed Array Doesn’t Work in JavaScript

And How To Do It Correctly

Scenario

For the sake of demonstration, suppose you need to generate an array of numbers from 0 to 99. How might you do this? Here’s one option:

const arr = [];

for (let i = 0; i < 100; i++) {
  arr[i] = i;
}

If you’re like me, seeing a traditional for-loop in JavaScript makes you slightly uncomfortable. In fact, I haven’t written a traditional for-loop in years thanks to higher-order functions such as forEach, map, filter, and reduce. Declarative functional programming for the win!

Maybe you haven’t yet drunk the functional programming Kool-aid and you’re thinking the above solution seems perfectly fine. It technically is, but after you’ve had a taste of the magic of higher-order functions, you’re probably thinking, “There must be a better way.

My first reaction to this problem was, “I know! I’ll create an empty array of length 100 and map the indices to each element!” JavaScript allows you create an empty array of length n with the Array constructor, like this:

const arr = Array(100);

Perfect, right? I have an array of length 100, so now I just need to map the index to each element.

const arr = Array(100).map((_, i) => i);

console.log(arr[0] === undefined); // true

What the h*ck!? The first element of the array should be 0, but it’s actually undefined!

Explanation

There’s an important technical distinction I need to make here in order to explain why this happened. Internally, JavaScript arrays are objects, with numbers as keys. For example:

['a', 'b', 'c']

is essentially equivalent to this object:

{
  0: 'a',
  1: 'b',
  2: 'c',
  length: 3
}

When you access the element at index 0 in an array, you’re really just accessing an object property whose key is 0. This is important because when you think of arrays as objects in conjunction with the way these higher-order functions are implemented (don’t worry, I did that part for you), the cause of our problem makes perfect sense.

When you create a new array with the Array constructor, it creates a new array object with its length property set to the value you passed in, but otherwise the object is a vacuum. There are no index keys in the object representation of the array whatsoever.

{
  //no index keys!
  length: 100
}

You get undefined when you try to access the array value at index 0, but it’s not that the value undefined is stored at index 0, it’s that the default behavior in JavaScript is to return undefined if you try to access the value of an object for a key that does not exist.

It just so happens that the higher-order functions map, reduce, filter, and forEach iterate over the index keys of the array object from 0 to length, but the callback is only executed if the key exists on the object. This explains why our callback is never called and nothing happens when we call the map function on the array — there are no index keys!

Solution

As you now know, what we need is an array whose internal object representation contains a key for each number from 0 to length. The best way to do this is to spread the array out into an empty array.

const arr = [...Array(100)].map((_, i) => i);

console.log(arr[0]);
// 0

Spreading the array into an empty array results in an array that’s filled with undefined at each index.

{
  0: undefined,
  1: undefined,
  2: undefined,
  ...
  99: undefined,
  length: 100
}

This is because the spread operator is simpler than the map function. It simply loops through the array (or any iterable, really) from 0 to length and creates a new index key in the enclosing array with the value returned from the spreading array at the current index. Since JavaScript returns undefined from our spreading array at each of its indices (remember, it does this by default because it doesn’t have the index key for that value), we end up with a new array that’s actually filled with index keys, and therefore map-able (and reduce-able, filter-able, and forEach-able).

Conclusion

We discovered some of the implications of arrays being internally represented as objects in Javascript, and learned the best way to create arrays of arbitrary length filled with whatever values you need.

As always, leave your comments, questions, and feedback below!

Happy coding! 👍

Follow me!

If you liked this post, follow me! Or at least throw me a clap or two. You can spare one measly clap, right?

Medium : @shawn.webdev
Twitter : @ReisnerShawn
dev.to : @sreisner
Instagram : @shawn.webdev

Check out more of my work


Top comments (16)

Collapse
 
elarcis profile image
Elarcis • Edited
Array.from({ length: 100 }, (_, i) => i);
Collapse
 
loilo profile image
Florian Reuschel

Whaaat.

I'm honestly baffled. Used Array.from() so often but never knew about the mapping argument. Well, today I learned. Thank you!

Collapse
 
sreisner profile image
Shawn Reisner

Well I'll be... Nice!

Collapse
 
jbristow profile image
Jon Bristow • Edited

Why not just [...Array(5).keys()];?

For history’s sake, it’s equivalent to:

Array.apply(null, Array(5)).map(function(_, i) {
  return i;
});
Collapse
 
thorstenhirsch profile image
Thorsten Hirsch • Edited

Shawn's point was not to find a solution that does not use a for loop. Shawn's point was to demonstrate why the code that one would expect to work does not:

Array(100).map((_, i) => i)

It looks simple, but does not work in Javascript, due to this array-is-an-object thingy. That's the point.

I'm not sure if I would go with Shawn's spread operator solution or if I'd just go back to a simple for loop. But I'm pretty sure that Array.apply(null, ...) and Elarcis' Array.from(...) are not solutions that are easy to understand (no offence), they both seem to be workarounds.

Collapse
 
loilo profile image
Florian Reuschel • Edited

To be honest, with regard to the article, I find it quite weird that Array(5).keys() does work. 🤔

Thread Thread
 
jbristow profile image
Jon Bristow

It doesn’t!

The ... fires first and produces Array.apply(null, Array(5)).keys()

Then that is expanded to Array(undefined, undefined, undefined, undefined, undefined).keys().

Since the Array was created by a list of items, keys returns [0,1,2,3,4]

JavaScript needs a range operator or a Texas range notation. Just so I don’t need to hide the splat under a function to increase readability.

Thread Thread
 
loilo profile image
Florian Reuschel

Makes sense, thanks. I struggle with splat operator execution order from time to time.

Thread Thread
 
jbristow profile image
Jon Bristow

In trying to rewrite the above code I ended up having to experiment a bit to figure out what was happening when... it’s certainly far from obvious.

Collapse
 
sreisner profile image
Shawn Reisner

This also works!

Collapse
 
ann0nip profile image
Juan Martin Gimenez

I think my question is out of the scope of the topic, but:

When we do:


Array(100);

If there are no index keys in the object representation of the array (the object is a vacuum), why we can spread it?? 🤯


const arr = [...Array(100)]

Thank you for this interesting post!

Salute from 🇦🇷!

Collapse
 
sreisner profile image
Shawn Reisner • Edited

Great question. The reason we can spread an array with no index keys is because spread's implementation doesn't require that any index keys exist, unlike map. It blindly loops from 0 to 99, asks the array for array[index], and places whatever's returned in the new array.

Collapse
 
belinde profile image
Franco Traversaro

I know this isn't a constructive comment, but... How the heck a map on a deconstructed array with an anonymous function can be more appealing then a for loop? Really, I don't want to start a flame, I just don't understand.

Collapse
 
sreisner profile image
Shawn Reisner

The for loop is an appealing construct. It's simple. You understand exactly what's happening algorithmically without much of a second thought. The appeal of the higher order map function is that it abstracts away the algorithmic details so that you're telling the code what to, not how to do it. In my experience, code written this way is easier to understand, debug, and maintain, and generally produces fewer bugs. This becomes more true when we need to chain the effects of these higher order functions, but even in cases like this we still benefit from the fact that there's almost nowhere for a bug to hide.

Collapse
 
belinde profile image
Franco Traversaro

I don't agree... You know, you've just written an article about this approach not working with a naive approach 😁

That was obviously a joke, but still for me it's very hard to understand what that code do. And I feel a bit of horror thinking about the CPU cycles wasted with so many calls...

Thread Thread
 
sreisner profile image
Shawn Reisner

Not quite. I added the bit about the for loop being perfectly fine at the beginning of the article on purpose - because it is! I'm glad you found it thought provoking. I appreciate the feedback! ❤️