Part 2 of our series discusses iteration protocols, for loops, and generator functions

This is Part 2 of Javascript mastery — and probably the most exciting parts of the language. ( Until Part 3 comes around, anyway ;) )

Part 1 covered the language basics, and here we’ll cover iteration protocol(s), their use in for loops, and generator functions.

Why generator functions in the mix? If you think that’s a random addition, read on! Generators are linked to iteration!

For Loops

Well, you know the basic for loop, right?

for (let i = 0; i < arr.length; i++)

You’d use this to access elements in an array.

You’d use something similar to access the properties / values of an object:

for ( let i = 0; i < Object.keys(obj).length; i++)

And again, something similar for map, set and any other custom object you define. When you want just the values / properties, writing this loop can lead to mistakes. You might be using the length property wrong, you might be making off by one errors or you might think Object.keys(obj).length is just plain ugly (I do).

Since there ought to be one best way to do things, here we have the for...of and for...in loops! … One best thing, right?

Well, yes. They both are loops to iterate over something, but that’s where the similarity ends, as we’ll see below.

For…of loop

Let’s start with trying to iterate over values in an object.

To access elements in an array: for (let val of arr)

To access values of an object: for (let var of Object.values(obj))

Beautiful, ain’t it? It begets the question though, why doesn’t for (let var of obj) simply work?

Let’s dive deeper into how this works and where all can you use the for…of loop. Most importantly, how can your classes / objects make use of this.

Welcome to the world of iteration protocols.

First, a short note about protocols.

If you’ve dealt with OOP before, then you probably know what an interface is: It’s a description of the actions that an object can do, like a contract. If you want to do X, you need to have a function defined in the contract which does X. For example, doX(a,b,c) which takes in parameters a,b,c . In the same way, protocols are interfaces in Javascript.

We have 2 iteration protocols in Javascript:

Iterable Protocol

This protocol enables JS objects to determine their iteration behavior. It’s enabling an object to be iterated over. It also determines what exactly is iterated. The interface demands an [Symbol.iterator] method somewhere up the prototype chain.

tzVvXusT0bOcZiXmqE6r32LjQLZ2LEjioZxh
MDN documentation

Iterator Protocol

This protocol determines the way our iterable protocol must return iterated values. Eh? An example would make this clearer.

The way I like to see it, the iterator protocol defines the class interface for an iterator. (If you look at the name again, this would seem pretty obvious, yeah? Iterator Protocol = Iterator Interface. Look ma, I can JS now.)

Going back to our dear documentation:

Jx4mFiYHWkzInSy6Hw0ZRCFRpr6kkju6yeAJ
Source

So, our iterator interface is determined completely by the existence of the next() function on an object.

One key point to make here is, it’s considered best practice to implement both the iterator and iterable protocols, since some functions / syntax may expect the former, while some the latter. Doing so enables you to use both with your iterator. Here’s a wonderful example:

const iteratorObject = {
 next() {
     const value = Math.random();
     if ( value < this.threshold ) {
         return { done: false, value}; 
     }
     return { done: true};
 },
 [Symbol.iterator]: function() {
     return this;
 },
 threshold: 0.7
}

The beauty lies in the [Symbol.iterator] part of the iterator. By defining this, we allow our iterator to be exposed to a variety of functions and syntaxes that need an iterable protocol, not just an iterator protocol. What can you do with this?

Remember the spread operator? — That accepts an iterable protocol as well!

>[...iteratorObject] 
[0.03085962239970308, 0.20649861146804716]

And of course, works with for...of, where this story began.

>for (let val of iteratorObject) {
    console.log(val);
}
0.6234680935767514
0.525812241023621

Under the hood, we can now understand what is happening: All these methods are using the [Symbol.iterator] to generate an iterator, and iterating over that using next!

>const iter = iteratorObject[Symbol.iterator]()
undefined
>iter.next();
{done: false, value: 0.04474940944875905}
>iter.next();
{done: true}

Sure makes things easier when you don’t have to do that yourself. There’s one bit we haven’t touched on, that goes hand in hand with for...of loops, which is: for...in. What’s the difference? Let’s dive in, starting with our example!

For…In Loops

>for (const val in iteratorObject) {
    console.log(val);
}
next
threshold

On a simple glance, the difference seems obvious: for...in gets the properties, while for...of gets the values! Why is [Symbol.iterator] missing then? Well, there are 2 reasons.

There exists an enumerable property descriptor over properties. This determines whether the given property is enumerable, configurable or writable.

> Object.getOwnPropertyDescriptors(iteratorObject)
{ next:
   { value: [Function: next],
     writable: true,
     enumerable: true,
     configurable: true },
  threshold:
   { value: 0.7,
     writable: true,
     enumerable: true,
     configurable: true },
  [Symbol(Symbol.iterator)]:
   { value: [Function: [Symbol.iterator]],
     writable: true,
     enumerable: true,
     configurable: true } }

The for...in loop loops over properties whose enumerable descriptor is set to true, as well as non-symbol properties. That explains it, right? Just to confirm, you could add a new property to the object, with enumerable set to false, and it wouldn’t show up in the for...in loop.

Object.defineProperty(iteratorObject, "newHiddenProperty", {
    enumerable: false,
    value: "hidden",
})

Sure enough, it still isn’t there. Object.keys() uses the exact same methodology.

>for(const val in iteratorObject) {
    console.log(val);
}
next
threshold

Coming back to the question that made us go down this rabbit hole — Why doesn’t for(let val of obj) simply work? Now you know, right? Because there doesn’t exist an iterable protocol on the Object prototype!

Why not? The simple answer is — language design choice. Why did they choose this? Because a lot of objects inherit from the base Object. Having an iterable protocol on the base Object would mean making all those objects iterable. For example: Your date objects become iterable, which doesn’t make any sense.

ForEach Loop

This brings us to the last kind of for loops: the forEach loop. I’ve seen people get confused over why doesn’t forEach work everywhere ( like on Objects) and I’ll answer that question here.

Simple answer — Array.prototype.forEach().

The forEach loop is defined only for arrays! So, you can use them only with arrays. Now, forEach doesn’t care where that array comes from. It could be a simple native array, or an array generated by Objects, like Object.keys().

To end the loops section, one common gotcha.

When using objects in JS as maps (or dictionaries, hashmap), you can run into issues when some key coincides with a property up the prototype chain.

Consider this example:

You have an object with certain keys you want to loop over.

const baseObject = {
  a: 1,
  b: 2,
  someProperty: function() {
    return 4;
  }
}


const myObjectMap = Object.create(baseObject);

myObjectMap.c = 3; // key set in map for some reason.

for(let val in myObjectMap) { // this iterates up the chain!
  console.log(val);
}

> c
 a
 b
 someProperty

You probably just wanted to see c, the key you set. You can fix this via:

for (let val in myObjectMap) {
  if (myObjectMap.hasOwnProperty(val)) {
    console.log(val);
  }
}

> c

Thus, two rules to avoid this problem:

  1. Always use hasOwnProperty() to check if the key you’re looking for exists in the object ( and not up the proto chain)
  2. Never use the hasOwnProperty as key in your dictionaries / maps.

If you have overridden hasOwnProperty, there is still a way to use it, since it’s a method of the Object prototype.

myObjectMap.hasOwnProperty = 4;

for(let val in myObjectMap) {
    if (myObjectMap.hasOwnProperty(val)) {
        console.log(val);
    }
}
> Uncaught TypeError: myObjectMap.hasOwnProperty is not a function
    at <anonymous>:4:21

// instead, we can do: 
for(let val in myObjectMap) {
    if (Object.prototype.hasOwnProperty.call(myObjectMap, val)) {
        console.log(val);
    }
}

> c
  hasOwnProperty

Remember call and apply from the last part? This is one awesome way to use them.

Generator Functions

Generator functions allow on-demand entry and exit from a function. The entry and exit points are fixed. It’s like a multiple entry visa.

They are very powerful tools to get difficult things done.

The way I think of generator functions is this: They are useful to create a list of values on the fly, without the overhead of having an array.

Why not just iterate over an array of values? Well, generators save space. There is no array to begin with — just the computation (or I/O) necessary to get the next element from the “array”.

Let’s dive into the mechanics of it.

Calling a generator function doesn’t execute the body but returns an iterator object for the function. The body is executed when you call the iterator’s next() method. What about the fixed exit point? The entire body isn’t executed, but only until the next yield expression in the body.

This yield expression also specifies the value to be returned.

Let’s make this concept concrete with an example. Let’s do the tweet example from Part 1.

function * generateTweets(userID, numberOfTweets) {
    for(let i=0; i< numberOfTweets; i++) {
        const tweet = randomTweetGenerator(); // assume this gives you a string of words < 280 characters.
        yield { tweet, userID, tweetID: i};
    }
}

const tweetList = generateTweets('neilkakkar', 3);
for( let tweet of tweetList) {
	  console.log(tweet);
}

> {tweet: "hi", userID: "neilkakkar", tweetID: 0}
  {tweet: "how's it going?", userID: "neilkakkar", tweetID: 1}
  {tweet: "I'm automagic", userID: "neilkakkar", tweetID: 2}


console.log(tweetList.next());
>    {value: undefined, done: true}

Okay, there’s a lot going on here. Let’s break it down.

First, we have the function generator, which generates tweets based on userID and number of tweets to generate. This function would return an iterator object. Thus, that’s what tweetList is.

> tweetList
generateTweets {<suspended>}
    __proto__: Generator
    [[GeneratorLocation]]: VM2668:1
    [[GeneratorStatus]]: "suspended"
    [[GeneratorFunction]]: ƒ * generateTweets(userID, numberOfTweets)
    [[GeneratorReceiver]]: Window
    [[Scopes]]: Scopes[3]

Suspended means the generator isn’t closed/finished yet. So, there are values it can provide. We can access these via tweetList.next() - which would give us an object with two keys, value and done.

On the flip side, for...of loops understand the iteration protocol so they can iterate over the entire generator on their own!

That’s precisely why we can do the for...of on tweetList and get our tweets.

At this point, the generator is finished. The for...of loop consumes all values.

Common gotcha: If there is a break statement inside the for...of loop, the generator closes too. So, you can’t reuse it again. See: Don’t reuse generators in for..of loops.

We have here

> tweetList
generateTweets {<closed>}
    __proto__: Generator
    [[GeneratorLocation]]: VM2668:1
    [[GeneratorStatus]]: "closed"
    [[GeneratorFunction]]: ƒ * generateTweets(userID, numberOfTweets)
    [[GeneratorReceiver]]: Window

Thus, when we log the next value in the next line, we get done: true as we’d expect - and no values.

That’s all for the example.

But, the story doesn’t end here. You can have generators yielding to generators as well! You do this via yield *.

function * generateTweetsForSomeUsers(users, numberOfTweets) {
    for(let user of users) {
        yield * generateTweets(user, numberOfTweets)
    }
}

Generators can also return instead of yield. This causes the generator to finish.

Well, this has gone on long enough, I think I’ll save the other cool bits for the next parts. Fun fact? We will get rid of for loops altogether. Welcome to the world of Map, Filter and Reduce.

Read more of my blog posts on neilkakkar.com.