CODEWITHSUNDEEP
rx-java
prom85
prom85

Reputation: 17828

RXJava - hot observables - safe cpu resources

Here's a simple observable example:

observable
            .filter(...)
            .buffer(50, TimeUnit.MILLISECONDS)
            .doOnNext(/* this is executed all the time... */)
            .filter(data -> data.size() > 0)
            .doOnNext(/* this is not executed because of the filter above... */)
            .flatMap(data -> Observable.from(data).distinctUntilChanged())
            .subscribeOn(Schedulers.io())
            .observeOn(AndroidSchedulers.mainThread())
            .subscribe();

Problem/Question

The observable that I get from the buffer(...) function is emitting result after result, mostly empty ones.

Is filtering the correct way to handle this problem with observables like this? Will having a lot of such subscriptions at the same time cost performance? Or should this be handled differently?

Upvotes: 1

Views: 90

Answers (1)

akarnokd
akarnokd

Reputation: 69997

Generally, yes, filter out what you don't need downstream.

Will having a lot of such subscriptions at the same time cost performance?

Technically: yes. More timers, more active sequences on the computation thread. However, the overhead of the sequence may be far less than the business logic you execute.

To save on some overhead, I'd use flatMapIterable:

observable
    .filter(...)
    .buffer(50, TimeUnit.MILLISECONDS)
    .doOnNext(/* this is executed all the time... */)
    .filter(data -> data.size() > 0)
    .doOnNext(/* this is not executed because of the filter above... */)
    .flatMapIterable(data -> data)               // <-----------------------------
    .distinctUntilChanged()
    .subscribeOn(Schedulers.io())
    .observeOn(AndroidSchedulers.mainThread())
    .subscribe();

Upvotes: 1

Related Questions