RxJS Subjects in Depth

RxJS Subjects in Depth

In this article, I want to explore the topic of RxJS’s implementation of Subjects, a utility that is increasingly getting awareness and love from the community. In the past, I have used Subjects in a variety of ways, but sometimes not fully...

In this article, I want to explore the topic of RxJS’s implementation of Subjects, a utility that is increasingly getting awareness and love from the community.

In the past, I have used Subjects in a variety of ways, but sometimes not fully understanding what they are internally and what are the main differences with Observables.

This is what the article will cover:

  • What is a Subject?
  • Multicasting and Unicasting
  • Other types of Subject: AsyncSubject, ReplaySubject, and BehaviorSubject
What is a Subject?

Let’s start with a simple question: what is a Subject?
According to Rx’s website:

A Subject is a special type of Observable that allows values to be multicasted to many Observers.
Subjects are like EventEmitters.

If this is unclear, hang on, by the end of the article you’ll have a much clearer understanding of what a Subject is and all the way you can make use of them.

The definition stated by the Rx documentation initially struck me: in fact, I always thought of Subjects as purely a way to both pull and push values using streams. It turns out, I didn’t really know them full well even after having used them daily for about 5 years.

Subject

Subject is a class that internally extends Observable. A Subject is both an Observable and an Observer that allows values to be multicasted to many Observers, unlike Observables, where each subscriber owns an independent execution of the Observable.

That means:

  • you can subscribe to a Subject to pull values from its stream
  • you can feed values to the stream by calling the method next()
  • you can even pass a Subject as an Observer to an Observable: as mentioned above, a Subject is also an Observer, and as a such, it implements the methods next, error and complete

Let’s see a quick example:

const subject$ = new Subject();
// Pull values
subject$.subscribe(
  console.log, 
  null, 
  () => console.log('Complete!')
);
// Push values
subject$.next('Hello World');
// Use Subject as an Observer
const numbers$ = of(1, 2, 3);
numbers$.subscribe(subject$);
/* Output below */
// Hello Word
// 1
// 2
// 3
// Complete!
The Internals of a Subject

Internally, every Subject maintains a registry (as an array) of observers. This is how internally a Subject works, in a nutshell:

  • every time a new observer subscribes, the Subject will store the observer in the observers' array
  • when a new item is emitted (i.e. the method next() was called), the Subject will loop through the observers and emit the same value to each one of them (multicasting). The same will happen when it errors or completes
  • when a Subject completes, all the observers will be automatically unsubscribed
  • when a Subject is unsubscribed, instead, the subscriptions will still be alive. The observers’ array is nullified, but it doesn’t unsubscribe them. If you attempt to emit a value from an unsubscribed Subject, it will actually throw an error. The best course of action should be to complete your subjects when you need to dispose of them and their observers
  • when one of the observers is unsubscribed, it will then be removed from the registry
Multicasting

Passing a Subject as an Observer allows to convert the Observable’s behavior from unicast to multicast. Using a Subject is, indeed, the only way to make an Observable multicast, which means they will share the same execution with multiple Observers.

Wait, though: what does sharing execution actually mean? Let’s see two examples to understand the concept better.

Let’s use the observable interval as an example: we want to create an observable that emits every 1000 ms (1 second), and we want to share the execution with all the subscribers, regardless of when they subscribed.

const subject$ = new Subject<number>();
const observer = {
  next: console.log
};
const observable$ = interval(1000);
// subscribe after 1 second
setTimeout(() => {
  console.log("Subscribing first observer");    
  subject$.subscribe(observer);
}, 1000);
// subscribe after 2 seconds
setTimeout(() => {
  console.log("Subscribing second observer");
  subject$.subscribe(observer);
}, 2000);
// subscribe using subject$ as an observer
observable$.subscribe(subject$);

Let’s summarize the snippet above

  • we create a subject called subject$ and an observer which simply logs the current value after each emission
  • we create an observable that emits every 1 second (using interval)
  • we subscribe respectively after 1 and 2 seconds
  • finally, we use the subject as an observer and subscribe to the interval observable

Let’s see the output:

As you can see in the image above, even if the second observable subscribed after 1 second, the values emitted to the 2 observers are exactly the same. Indeed, they share the same source observable.

Another common example that shows the usefulness of multicasting is subscribing to an observable that executes an HTTP request, a scenario that happens often in frameworks such as Angular: by multicasting the observable, you can avoid executing multiple requests and share the execution with multiple subscribers, that will receive the same value.

AsyncSubject

I personally find AsyncSubject the least known type of Subject, simply because I never really needed it, or more likely I didn’t know I could have needed it.

In short, the AsyncSubject will:

  • emit only once it completes
  • emit only the latest value it received
const asyncSubject$ = new AsyncSubject();
asyncSubject$.next(1);
asyncSubject$.next(2);
asyncSubject$.next(3);
asyncSubject$.subscribe(console.log);
// ... nothing happening!
asyncSubject$.complete();
// 3

As you can see, even if we subscribed, nothing happened until we called the method complete.

ReplaySubject

Before introducing a ReplaySubject, let’s see a common situation where normal Subjects are used:

  • we create a subject
  • somewhere in our app, we start pushing values to the subject, but there’s no subscriber yet
  • at some point, the first observer subscribes
  • we expect the observer to emit the values (all of them? or only the last one?) that were previously pushed through the subject
  • … nothing happening! In fact, a Subject has no memory
const subject$ = new Subject();
// somewhere else in our app
subject.next(/* value */);
// somewhere in our app
subject$.subscribe(/* do something */);
// nothing happening

This is one of the situations when a ReplaySubject can help us: in fact, a Subject will record the values emitted and will push to the observer all the values emitted when a subscribed.

Let’s get back to the question above: does a ReplaySubject replay all the emissions or only the latest one?

Well, by default, the subject will replay all the items emitted, but we can provide an argument called bufferSize. This argument defines the number of emissions that the ReplaySubject should keep in its memory:

const subject$ = new ReplaySubject(1);
subject$.next(1);
subject$.next(2);
subject$.next(3);
subject$.subscribe(console.log);
// Output
// 3

There’s also a second argument that can be passed to ReplaySubject in order to define how long the old values should be stored in memory.

const subject$ = new ReplaySubject(100, 250);
setTimeout(() => subject$.next(1), 50);
setTimeout(() => subject$.next(2), 100);
setTimeout(() => subject$.next(3), 150);
setTimeout(() => subject$.next(4), 200);
setTimeout(() => subject$.next(5), 250);
setTimeout(() => {
  subject$.subscribe(v => console.log('SUBCRIPTION A', v));
}, 200);
setTimeout(() => {
  subject$.subscribe(v => console.log('SUBCRIPTION B', v));
}, 400);
  • we create a ReplaySubject whose bufferSize is 100 and windowTime 250
  • we emit 5 values every 50ms
  • we subscribe the first time after 200ms and the second time after 400ms

Let’s analyze the output:

SUBCRIPTION A 1
SUBCRIPTION A 2
SUBCRIPTION A 3
SUBCRIPTION A 4
SUBCRIPTION A 5
SUBCRIPTION B 4
SUBCRIPTION B 5

The subscription A was able to replay all the items, but the subscription B was only able to replay items 4 and 5, as they were the only ones emitted within the window time specified.

BehaviorSubject

BehaviorSubject is probably the most well-known subclass of Subject. This kind of Subject represents the “current value”.

Interestingly, the Combine framework named it CurrentValueSubject

Similarly to ReplaySubject, it will also replay the current value whenever an observer subscribes to it.

In order to use BehaviorSubject we need to provide a mandatory initial value when this gets instantiated.

const subject$ = new BehaviorSubject(0); // 0 is the initial value
subject$.next(1);
setTimeout(() => {
  subject$.subscribe(console.log);
}, 200);
// 1

Whenever a new value is emitted, the BehaviorSubject will store the value in the property value which can also be publicly accessed.

Final Words

Rx Subjects are quite powerful tools, and like any powerful tool in software engineering, they can also be easily abused. The concept of unicasting and multicasting is a striking distinction that you need to take into account when working with Rx.

Understanding how Subjects work internally can be fairly beneficial to avoid common pitfalls and bugs, but also to understand when you need them and when, instead, you don’t.

If you need any clarifications, or if you think something is unclear or wrong, do please leave a comment!

Trick or Unsubscribe in RxJS: a Custom Angular Decorator

Trick or Unsubscribe in RxJS: a Custom Angular Decorator

One of the first concepts of Angular is Functional Reactive Programming via Observables. Angular extensively use Observables through RxJS library that introduces an implementation of the Observable type.

Background

Why Observables might be dangerous for your application? What are the options to reduce the risks? As you may have already guessed I’m going to talk about the “unsubscribe()” and I’ll be honored to present you my custom solution that is saving my lifetime and might save yours.

Introduction to Observable’s world

One of the first concepts of Angular is Functional Reactive Programming via Observables. Angular extensively use Observables through RxJS library that introduces an implementation of the Observable type. I won’t elaborate on the subject of Reactive Programming in the Angular or the RxJS library, I will just cover a few high-level principles.

According to official docs - “Observables are lazy Push collections of multiple values”. Other words to say, it is a data stream - a sequence of any values in time. So, an Observable is some kind of advanced Promise that pushes (resolves) multiple values over time to callbacks instead of only one value.

In order to notify the Observable when to send data and also react to new data in the future, we need to subscribe to it, by simply calling the “subscribe()” method. As I have mentioned above, the Observable is some kind of a stream itself what means that after subscribing to it, its execution will be infinite. And in order to cancel/complete it and “sleep like a baby”, we have simply to call an “unsubscribe()” method. Easygoing, right?

However, here’s the most common mistake, especially among juniors, when a developer simply forgets to unsubscribe from a stream and moves further. And an Observable that isn’t used anymore still would be producing values. That directly leads to tremendous memory leaks and unpredictable behavior of your application in the future.

What are the "advanced" options to unsubscribe?

As I have mentioned above, if you don’t want to shoot yourself in a leg - you always should remember to unsubscribe! The most common place to do it in Angular is inside of “ngOnDestroy” lifecycle hook that is executed by Angular once the component is not used anymore.

This is the easiest solution when you have one or two subscriptions but in the real Angular application, you have dozens of subscriptions. And definitely, it would be tedious each time to unsubscribe “manually”. What to do then? Let’s consider some "advanced" built-in ways of unsubscribing from multiple Observables:

1. Chained subscriptions:

As a Subscription is a Class that essentially has an “unsubscribe()” method, it also has an “add()” method. It allows "adding" one Subscription into another - a child subscription to a parent subscription. Thus, you need to call an unsubscribe() method only once - a parent Subscription unsubscribes all child Subscriptions. Have a look at the example below.

export class HomeComponent implements OnInit, OnDestroy {
  sub: Subscription = new Subscription();

constructor(
private invoicesService: InvoicesService,
private productsService: ProductsService,
private customersService: CustomersService,
) {
}
ngOnInit() {
this.sub
.add(
this.invoicesService.invoices$
.subscribe(invoices => console.log(invoices))
)
.add(
this.productsService.products$
.subscribe(products => console.log(products))
)
.add(
this.customersService.products$
.subscribe(products => console.log(customers))
);
}
ngOnDestroy() {
this.sub.unsubscribe();
}

However, there’s an adverse effect within chaining - in case one of the chained subscriptions completes, e.g. the products$ stream throws an error, then its further descendant, I mean the customers$ stream, won’t be executed. Thus, I’d suggest avoiding chaining.

2. An array of Subscriptions:

Firstly, we create a variable with type “Subscription[]”, e.g. “subscriptions” with initial value as an empty Array. Then we create a setter in order not to wrap manually each Subscription in a “push” construct. Afterward, in the ngOnDestroy lifecycle hook we simply call the forEach() method on our Array and call an unsubscribe() method on each subscription inside of it. Check out the code example:

export class HomeComponent implements OnInit, OnDestroy {

subscriptions: Subscription[] = [];

private set sub (sub: Subscription) {
this.subscriptions.push(sub);
}

constructor(
private invoicesService: InvoicesService,
private productsService: ProductsService,
) {
}

ngOnInit() {
this.sub = this.invoicesService.invoices$
.subscribe(invoices => console.log(invoices));

this.sub = this.productsService.products$
  .subscribe(products =&gt; console.log(products));

}
ngOnDestroy() {
this.subscriptions.forEach(sub => sub.unsubscribe());
}
}

3. RxJS “Subject” and “takeUntil” operator:

Firstly, we create a variable/stream, e.g. unsubscribe$ with a new instance of the RxJS Subject. Then inside of the pipe chain of any other stream, we declare the “takeUntil” operator to which we simply pass our unsubscribe$ stream. Afterward, in the ngOnDestroy lifecycle hook, we call next() and complete() callbacks on our Subject. It means that all subscribers automatically stop receiving future values when our Component would be destroyed because our Subject would be completed. Let me provide you with a code example:

export class HomeComponent implements OnInit, OnDestroy {

unsubscribe$: Subject<void> = new Subject();

constructor(
private invoicesService: InvoicesService,
private productsService: ProductsService,
) {
}

ngOnInit() {
this.invoicesService.invoices$
.pipe(
takeUntil(this.unsubscribe$)
)
.subscribe(invoices => console.log(invoices));

this.productsService.products$
  .pipe(
    takeUntil(this.unsubscribe$)
  )
  .subscribe(products =&gt; console.log(products));

}

ngOnDestroy() {
this.unsubscribe$.next();
this.unsubscribe$.complete();
}
}

4. RxJS “AsyncPipe”:

This is the last, however, the most reliable, neat and correct built-in option for unsubscribing within Observables. An “AsyncPipe” automatically subscribes to an Observable, returns the latest value it has emitted and also unsubscribes when a Component is destroyed. Thus, we don’t need to do anything. All the cleanup logic for avoiding the memory leaks is done under the hood. It’s amazing! Just take a look at an example below:

export class InvoicesComponent implements OnInit {

invoices$: Observable<Invoice[]>;

constructor(
private invoicesService: InvoicesService,
) {
}

ngOnInit() {
this.invoices$ = this.invoicesService.invoices$;
}
}
<main class="invoices-main">

&lt;mat-table [dataSource]='invoices$ | async'&gt;

....
</mat-table>

<main/>

Why have I come to a custom solution and what are the decorators itself?

The AsyncPipe is reliable and works well, however, very often we have to not just simply subscribe to an Observable and render the output, we need to put some logic in a subscribe() method. Thus, every time we’ll have to repeat the implementation in our Components one of those advanced unsubscribing options mentioned above.

So, after a while, I’ve decided that I don’t want to do a “monkey job” inside of many Components manually. I thought that it would be great to put out all the unsubscribing logic somewhere in one place and just reuse it when I would need, additionally to make my code cleaner and maintainable. And, thanks to the Typescript, I’ve found the right, neat and “Angularish” place - a Decorator. You might already know that Decorators are extensively used throughout an Angular, but if you don’t know what are the Decorators itself and asking yourself what is the magic under the hood, let me explain it very briefly.

In general, the main idea of Decorator is that you can dynamically attach to the object additional functionality. And if to be more precise, in a Typescript, the Decorator is a pure function with arguments that is called by @ sign and can be attached to:

  • Classes;
  • Methods;
  • Properties;
  • Parameters;
  • Accessor.

Just in case, here’s a simple example within a Class:

function Log() {
console.log(arguments);
}

@Log
export class HomeComponent {
...
}
// printed to console:
// {'0': [Function: HomeComponent]}

All in all, Decorators simply help to customize the thing they are attached to at design time. Let’s move further where I would be glad to present and describe my own Decorator for unsubscribing from Observables that I’ve called - “DestroySubscribers”.

My custom @DestroySubscribers() decorator

I’m really delighted with RxJS, but I’ve decided to automate the unsubscribe process and clean my code with the help of a Class Decorator and an “Array of Subscriptions” approach implementation.

Check out the “DestroySubscribers” Decorator itself:

export function DestroySubscribers(params?) {

return function (target) {
params = {
destroyFunc: 'ngOnDestroy',
...params
};
const unsubscribableLike: {subscriptions: Unsubscribable[], unsubscribe: () => void} = {
subscriptions: [],
unsubscribe,
};
const subscriber: string = Reflect.getMetadata('subscription:name', target.prototype, 'subscriber');

Object.defineProperty(target.prototype, subscriber ? subscriber : 'subscriber', {
get: () => unsubscribableLike,
set: subscription => unsubscribableLike.subscriptions.push(subscription),
});

if (typeof target.prototype[params.destroyFunc] !== 'function') {
throw new Error(${target.prototype.constructor.name} must implement ${params.destroyFunc}() lifecycle hook);
}

target.prototype[params.destroyFunc] = ngOnDestroyDecorator(target.prototype[params.destroyFunc]);

function ngOnDestroyDecorator(f) {
return function () {
unsubscribe();
return f.apply(this, arguments);
};
}

function unsubscribe() {
do {
const sub: Unsubscribable = unsubscribableLike.subscriptions.shift();
if ( sub && typeof sub.unsubscribe === 'function') { sub.unsubscribe(); }
} while (unsubscribableLike.subscriptions.length);
}

return target;
};
}

export function CombineSubscriptions(params?) {
return function (target, propertyKey: string | symbol) {
Reflect.defineMetadata('subscription:name', propertyKey, target, 'subscriber');
};
}

As you can see from the code above - the “@DestroySubscribers()” Decorator represents an “Array of subscriptions” approach extended with the “@CombineSubscriptions()” Decorator, and everything is done under the hood now. Let me briefly describe its main code points.

First, I’ve created an object with an empty array for future Subscriptions and custom unsubscribe method in order to have an ability to unsubscribe from all subscriptions at a time manually. Then with the help of reflect-metadata library and “@CombineSubscriptions” Decorator, I‘ve got the current property name from the Class or assign the “subscriber” as a default name and create getter and setter methods. Afterward, I’ve created another version of ngOnDestroy lifecycle hook that firstly unsubscribes from all subscriptions in the array, secondly invokes and returns the original ngOnDestroy method by default or another “destroying function” specified in the passed config to Decorator. That’s it - quite concise and easy to use.

And the decorator's implementation is even simpler. Check it out:

@DestroySubscribers({
destroyFunc: 'ngAfterViewInit',
})
export class HomeComponent implements OnInit, AfterViewInit {

/*
Within the @CombineSubscriptions Decorator, you can choose any custom name that you prefer.
Without the @CombineSubscriptions Decorator, the name by default is 'subscriber'.
*/
@CombineSubscriptions()
private subscriber: Unsubscribable;

constructor(
private invoicesService: InvoicesService,
private productsService: ProductsService,
) {
}

ngOnInit() {
this.subscriber = this.invoicesService.invoices$
.subscribe(invoices => console.log(invoices));

this.subscriber = this.productsService.products$
  .subscribe(products =&gt; console.log(products));

}

/*
This method must be declared, even if it's empty.
Otherwise, the Decorator would throw an Error.
*/
ngAfterViewInit() {
console.log('for unsubscribing');
}
}

The DestroySubscribers Decorator’s main definitions:

  • “subscriber” - a variable that represents the name by default for each subscription and conforms to an Unsubscribable Interface. Each time when you assign a Subscription to the "subscribe" variable - it's auto pushed to the array of Subscriptions under the hood. In addition, if you want to unsubscribe from all subscriptions at a time manually before a Component destroys, you can call an unsubscribe() method on the "subscriber" variable.
  • “@CombineSubscriptions()” Decorator - implement this Decorator in case you want to change the default variable's name("subscriber") of a subscription and use your own custom name, otherwise don't apply it.
  • {destroyFunc: '...' } - add this parameter to the “@DestroySubscribers” Decorator with the name of a hook for auto unsubscribing in case you want to change the default one - "ngOnDestroy" lifecycle hook, otherwise don't apply it. An ability to change the function called when a Component is destroyed gives you an opportunity to use this Decorator not only within an Angular.

The DestroySubscribers Decorator’s implementation steps:

Firstly, you have to annotate the Class with the “@DestroySubscribers()” Decorator.

Secondly, you need to create a variable called “subscriber” by default with the type Unsubscribable or if you want to use your own custom name - simply annotate that variable with “@CombineSubscriptions()” Decorator.

Thirdly, you should just assign to that variable each Subscription that you want to be unsubscribed from when the Component would be no longer in use.

The last thing, you must declare the ngOnDestroy lifecycle hook in a Component even if it's empty because of AOT compilation. Otherwise, the Decorator would throw an Error. In case you change the default lifecycle hook(ngOnDestroy) to another one(ngAfterViewInit) as in an example above, then this method must be declared in a Component, and the ngOnDestroy is obviously optional. I’ve told you, as easy as falling off a log!

Conclusion

All in all, I would like to outline that Decorators itself are nothing to be scared of, but rather are really amazing tools that you should use. They definitely would make your code more reusable, concise and readable!

In addition, thanks to the Angular community over time appeared many different solutions for unsubscribing. All of them are noteworthy and make our everyday lives easier! 

Thanks for reading. If you liked this post, share it with all of your programming buddies!

This post was originally published here

Angular and The Case for RxJS

Angular and The Case for RxJS

Angular and The Case for RxJS. When choosing Angular to build a front end app, it can seem like RxJS is just one more thing to learn, and there’s no easy transition coming from another framework. RxJS is a core part of the Angular framework used in everything from Reactive Forms to the HTTP module. evelopers can overcome the RxJS learning curve by focussing on core concepts like hot vs cold observables, piping operations, and a few of the most common operators.

When choosing Angular to build a front end app, it can seem like RxJS is just one more thing to learn, and there’s no easy transition coming from another framework. But RxJS is efficient and expressive! We declare relationships between entities, describing the what not the how. The basic sequence operators, map, filter, and reduce, are probably familiar from using them in array chains. RxJS is a core part of the Angular framework used in everything from Reactive Forms to the HTTP module.

Harnessing observables with Angular’s async pipe is essential to building clean and concise components. Developers can overcome the RxJS learning curve by focussing on core concepts like hot vs cold observables, piping operations, and a few of the most common operators.

Node vs Angular : Comparing Two Strong JavaScript Technologies

Just from being a simple client-side scripting language, JavaScript has evolved over the years to turn out to be a powerful programming language. Here Node.js is a cross-platform runtime environment while AngularJS is one of the top JavaScript framework. Angular helps the developers to build web applications which are dynamic in nature using HTML template language and following the MVC design pattern.