AsyncSubject: representing a computation that yields a final value

[00:00] Let's look at one more type of subject that also has some sort of replay logic in it. That would be the async subject. So far, we saw that the replay subject is able to replay many values to a late observer, regardless of when that late observer arrived. It could have arrived before or after completion, it doesn't matter.

[00:19] A behavior subject will only replay one value to a late observer, and it only does that while the subject has not completed yet. The async subject will replay just one value, just like the behavior subject, but it only does that after or when the subject completes.

[00:37] Actually, before the subject completes, it won't emit any value at all. Let's take a look at how that works. Let's replace this one with an async subject. It takes no arguments here in the constructor. Now when we run this, we're going to see that A saw just three, and done, and then B saw three and the completion.

[01:00] It actually works like this, that when A subscribed somewhere here, it didn't see any of these values because it only sees values from the subject when the subject completes. When it completes here, it will get the last value, and then it's done.

[01:18] Then B which subscribed somewhere here, it also receives the last value, the replayed value, three, and then it completes. As we can see, the async subject will only emit one value. That value is exactly the last value from this execution here.

[01:38] Actually, in this sense, async subject works like the last operator. If you've seen the last operator, it only emits the last value in the sequence. Here, we only know what the last value is once we complete. That's the only way that we can know what the last value is.

[01:56] When is an async subject useful? That would be for heavy computations that eventually give you some value. Let's imagine that this would take a lot of CPU, and RAM, and maybe network requests, and eventually gives us a value.

[02:10] Because it is heavy, we don't want to repeat that for every single observer. We want to have two properties. We want to share that result, and we want to remember that result. Sharing exists in all of the subjects that we saw so far. They all see the same execution.

[02:28] Remembering is a property of replaying. This is what the async subject gives you. It allows us to remember what the last value in that computation was, the final result of this heavy computation, so that other observers will get that from memory instead of recalculating it all over again.

[02:49] In this sense, an async subject is like a promise because promises also, they eventually resolve with a value. Then if you do a then on that promise, you will get that value R, and you won't recalculate this all over again.

[03:08] Promises can be considered as eventual async values. That's also the case for an async subject. They are eventual async values. We know that they are the last ones. In fact, they are actually rarely used, to be honest.

[03:26] Most of the times, you're going to want a behavior subject, or a replay subject. In many cases, the replay subject can replace the use case for an async subject. Just keep that in mind. Don't worry so much about trying to use the async subject.

[03:41] It's important to know that all of these three variants exist. These are actually all of those that exist, besides the subject. We have only the subject that has no replay. Then we have these three that have some replay logic in it. That's pretty much all of the spectrum of subjects that there is in RxJS.