TL;DR, take me to the code already!
Rx programming is hard. It’s hard at the beginning and then it gets even harder. In Future Mind we have begun our journey with RxJava more than two years ago and even now we learn new things every day. I’ve seen many developers having difficulties with fully grasping relatively simple problems concerning reactive programming. One of those problems is proper reactive task scheduling.
getUsers()
.subscribeOn(Schedulers.io())
.observeOn(AndroidSchedulers.mainThread())
.subscribe({ showUsers(it) })
You probably know exactly what happens here. Just a reminder: getUsers()
gets called on a background thread (Schedulers.io()
), and showUsers()
method is called on the main UI thread (AndroidSchedulers.mainThread()
). Easy?
What if you have a scenario where you first react to user pressing a button (mainThread()
), then load a list from the web (io
()
)? Write each item from the list to database (io
()
) and wait a few milliseconds (computation()
) before notifying view about a change (mainThread()
). Finally, notify the view again about the list having been successfully processed (mainThread
()
). Where do you put all those observeOn()
and subscribeOn
()
? You know? Congratulations, you don’t need the cheat sheet!
Well, not everybody is a smartass like you. That’s why we’ve put together a quick code snippet along with a cheat sheet, so that we have it close at hand whenever someone has any trouble with schedulers.
Download as a super duper handy PDF
It’s written in Kotlin and uses RxJava2 but these principles apply to all ReactiveX flavours. It's also much more readable in the PDF form above, because it has colours and everything. Without the colours, you have to be, like, super knowledgeable about dogs being mammals, crocodiles being reptiles etc.
Completable.fromCallable { animalsMaker.makeCat(getThread()) }
.doOnComplete { animalsMaker.makeDog(getThread()) }
.observeOn(scheduler(FISH))
.andThen(Completable.fromCallable { animalsMaker.makeShark(getThread()) })
.andThen(
Completable.fromCallable { animalsMaker.makeChicken(getThread()) }
.subscribeOn(scheduler(BIRDS))
.andThen(Completable.fromCallable { animalsMaker.makeDuck(getThread()) })
)
.andThen(Completable.fromCallable { animalsMaker.makePenguin(getThread()) })
.subscribeOn(scheduler(MAMMALS))
.subscribeOn(scheduler(AMPHIBIANS))
.observeOn(scheduler(REPTILES))
.doOnComplete { animalsMaker.makeCrocodile(getThread()) }
observeOn
()
always works downstream — it defines the Scheduler for the tasks following it.
subscribeOn
()
generally defines on what scheduler the first task in the chain starts (e.g. makeCat
()
is called on MAMMALS
).
A nested chain is by default subscribed on the preceding Scheduler, but this can be changed with a call to subscribeOn
()
. In other words, makeChicken
()
would be subscribed on FISH
by default, but is explicitly subscribed on BIRDS
instead.
The Scheduler of a nested chain defines the Scheduler for the following tasks of outer chain. That is why makePenguin
()
is called on BIRDS
Scheduler.
Some operators (e.g. delay()
, timer()
) have a default Scheduler, which can affect the following parts of the chain quietly.
In the example above the AMPHIBIANS
Scheduler is never used, because it’s overridden by MAMMALS
, but there are special cases where multiple subscribeOn
()
make sense, e.g. when using doOnSubscribe()
— you can explore them with Maciek Górski’s presentation.
The cheat sheet is based on a code that we’ve made available to you to play with. Have fun and pay special attention not to mess crocodile and dog on the same scheduler (brrr).