Multicast with a selector argument, as a sandbox

[00:00] The argument given to multicast can be a subject or a subjectFactory. Then we get as an output an observable which is shared for multiple observers. There's actually one more feature in multicast, and that's when you give a second argument here, which is a function, the so-called selector function. The selector function takes as an input the shared observable. Here, the body of the selector function acts as a sandbox.

[00:29] When you use the second argument, the selector function, multicast will not return any more a connectable observable. That means that what multicast returns here is just a normal observable. It doesn't have, for instance, refCount anymore, and this here won't be a shared observable it will just be a result observable.

[00:50] Let's see how this works. What we have here as an input for the selector function is basically an observable that represents this source, but it's shared, so we can use multiple observers of this one here inside the sandbox. This is basically our sandbox, the body of the selector function. The idea is that we can move our observers of the common shared execution into the sandbox. Only inside the sandbox would we have this shared execution, and not outside of it.

[01:25] Let's look at an easier case for now. We have the source observable here that emits events, and we're going to map each of those events to a random number, like this. Let's ignore multicasting for now just to see how it would be without multicasting. The result is an observable of six random numbers. We can delay each of those numbers just by doing result.delay by 500 milliseconds for instance.

[01:55] Then we can merge the original, the result, with the delayed. Now if I subscribe to this, and I say get the x event and put it in the console. Let's run this. What happens is that once we subscribe to the merged, it will subscribe to result and will subscribe to result delayed. There are two different subscribes here, but when we subscribe to result delayed, it will also subscribe to result. There will be two separate subscriptions of result here, and that's why there's going to be two separate executions of result.

[02:41] Let's see that happening in practice, and there we go. We have source zero and source zero. It means that the source observable emitted zero twice. There's actually two executions going on there. This is just what we had before, no sharing going on that's why we have two executions, there's no shared execution.

[03:03] With multicast here, with the selector function, we get a sandbox. Shared now is basically this observable, but shared, so there's just one execution of this observable. That means that we have just one execution of these six random numbers. For instance, we can make the shared delayed to be shared delay, by 500 milliseconds, and then we could make the merged to be the shared merged with the shared delayed, and then we need to return the merged.

[03:52] Once we've returned the merged, that will be the result of what multicast returns. Multicast will return this observable. Essentially result here is merged. If we subscribe to result now, it will subscribe to this merged, but this merged depends on this shared observable, and the shared observable is this here.

[04:21] Let's see that working. As you can see, the source emits one zero only. There's no two instances of zero. That same random value was delayed by 500 milliseconds. As you can see, the shared refers to this observable here that's ticking, but with the sandbox here, we were able to delay each of these by 500 milliseconds. It's not generating another random number, and it's not executing this do in the separate instance.

[04:59] Multicast with a selector function is quite different than a multicast without one. Without a selector, this will return a connectable observable, so you either have to connect it manually, or you have to use refCount. With the selector, you use this sandbox function to hook up some observables together that depended on the shared observable, and in the end, you need to return an observable like we're doing here.

[05:26] The output of multicast, in this case, is not a connectable observable, it's just a normal observable, which is, after all, the merged observable. What the selector sandbox here returns is the output of Multicast, which means merged is the result.

[05:44] You may be asking yourself, "When does this connect?", because all of the shared observables are backed by a subject. That happens when we subscribe to this result. When we subscribe to the result, it will subscribe to this multicasted thing, which runs the sandbox, and then it will subscribe to this source, using the subject that we create from the factory.

[06:09] That's when it connects. When does it disconnect, or when does it stop? When does the shared execution stop? It happens when you unsubscribe from this subscription here. This subscription represents the connect that happened here, and when we unsubscribe from this one, we will stop the shared execution that happened here.

[06:30] The takeaway is you should use a selector function in multicast when you want to create, let's say, a diamond-shaped dependency. Here we have a bifurcation. As you see we have shared, and it's used in two parts, and then we converge those two parts together to return one observable. That's kind of like a diamond shape, where we bifurcate, and then we converge.

[06:56] That is one case where you almost always want to use a selector function in multicast. If you don't, then usually we use just multicast with a refCount. That's quite common to use.