When should I use `publishReplay` vs `shareReplay`

问题:

I already know that

• publish shares a single subscription and also returns a ConnectableObservable ( so we have to Connect())

• Share() is publish().refcount()

The Replay postfix is pretty obvious , it returns its last emission/s.

Let's take for example an Angular http request with present AND future subscription :

<p>{{ (person | async)?.id   }}</p> //present markup

<p *ngIf=”show”>{{ (person | async)?.userId }}</p> //future markup

If I don't want multiple http requests I can use :

publishReplay().Connect()

But I can also use : shareReplay() , but I'm sure that there is one here that is more correct to use than the other.

Question :

When should I use publishReplay vs shareReplay ? What will be the difference in terms of that Http present & future request ?

_{NB Why there's not documentation about shareReplay ?}

回答1:

publishReplay allows you to controls when the subscription starts. shareReplay will start automatically upon the first subscription.

Generally, if the observable is to be used in a template (html file) use shareReplay. The advantage being you won't have to worry about unsubscribing etc.

回答2:

shareReplay() is basically publishReplay().refCount()

Definitely not.

Both shareReplay and publishReplay (+ calling connect on it) will make the observable behind it hot.

But the very important difference between them is:

• shareReplay: won't stop emitting until it's completed, no matter if there are no subscriptions anymore or not.
• publishReplay: will stop after the last subscriber unsubscribes if used together with refCount

Imho this is a crucial information.

回答3:

shareReplay() is basically publishReplay().refCount()

Here is a great article explaingin just that in detail: "Angular Async Pipes  - Beware the share"

Edit:

The right thing to say is:

shareReplay() is similar to publishReplay().refCount()

see @DevRok's answer for more info why they are not exactly the same.