Dive into Async apps with Kotlin Coroutines

This talk highlights the challenges underpinning some of the paradigms of asynchronous programming, in particular, the callback-based approach. The talk will address how Kotlin aims to solve this problem with coroutines by providing an asynchronous interface to the developer.

- Coroutines Generator API - Sequences or iterables in an asynchronous manner
- Coroutine Scopes and how to manage them
- How to use Coroutines to do network calls from the app.
- Actors, easing concurrency for Android Developers
- Sneak peek at channels, broadcast channel, flow
- Using Coroutines and Rx2 together

Transcript

1. Dive into Async apps with Kotlin Coroutines
2. @aditlal Adit Lal Product Engineer @rivuchakraborty Rivu Chakraborty Android Engineer
3. Problem • Run multiple tasks • Async is easy doable
4. Asynchronous • Threading • AsyncTask • Callbacks • Observable • Future/Promise • Coroutines
5. Threads • Threads are expensive • occupies about 1-2 mb • Requires context switching • Thread management • Race conditions • Thread Pools
6. Callbacks • Difficulty of nested callbacks • Callback Hell • Error Handling is a mess
7. Future/Promise • Similar to callbacks • Composition model with chained calls • Complicated Exception handling • Returns Promise
8. Reactive • Observable streams • Constant API across platforms • Great error handling
9. Coroutines • Launch a long-running operation w/o blocking thread • Functions are now suspend-able • Platform independent
10. Coroutines simplify asynchronous programming by providing possibility to write code in direct style (sequentially).
11. ./fetch_data
12. Blocking Thread Time Function A Blocked Blocked Function B
13. Suspended fun Thread Time Function A Suspended Function B
14. fun loadData() = GlobalScope.launch(uiContext) { showLoading() // ui thread val result = dataProvider.fetchData() // non ui thread, suspend until finished showText(result) // ui thread hideLoading() // ui thread } Suspend
15. fun loadData() = GlobalScope.launch(uiContext) { showLoading() // ui thread val result = dataProvider.fetchData() // non ui thread, suspend until finished showText(result) // ui thread hideLoading() // ui thread }
16. fun loadData() = GlobalScope.launch(uiContext) { showLoading() // ui thread val result = dataProvider.fetchData() // non ui thread, suspend until finished showText(result) // ui thread hideLoading() // ui thread }
17. fun loadData() = GlobalScope.launch(uiContext) { showLoading() // ui thread val result = dataProvider.fetchData() // non ui thread, suspend until finished showText(result) // ui thread hideLoading() // ui thread }
18. fun loadData() = GlobalScope.launch(uiContext) { showLoading() // ui thread val result = dataProvider.fetchData() // non ui thread, suspend until finished showText(result) // ui thread hideLoading() // ui thread }
19. CoroutineScope CoroutineContext Coroutine Builder launch
20. Context fun main() = runBlocking<Unit> { println("My context is: $coroutineContext") } My context is: [CoroutineId(1), "coroutine#1":BlockingCoroutine{Active}@73c6c3b2, BlockingEventLoop@48533e64]
21. Scope interface CoroutineScope { val coroutineContext: CoroutineContext }
22. ‘Dat Sequential suspend fun doSomethingUsefulOne(): Int { delay(1000L) //some fancy magic return 13 } suspend fun doSomethingUsefulTwo(): Int { delay(1000L) //some fancy magic again return 29 }
23. ‘Dat Sequential val time = measureTimeMillis { val one = doSomethingUsefulOne() val two = doSomethingUsefulTwo() println("The answer is ${one + two}") } println("Completed in $time ms") The answer is 42 Completed in 2007 ms
24. ‘Dat async val time = measureTimeMillis { val one = async(IO){ doSomethingUsefulOne() } val two = async(IO) { doSomethingUsefulTwo() } println("The answer is ${one.await() + two.await()}") } println("Completed in $time ms") The answer is 42 Completed in 1014 ms
25. Dispatcher To specify where the coroutines should run, Kotlin provides three Dispatchers you can use for thread dispatch.
26. Dispatchers.Main Main thread on Android, interact with the UI and perform light work Great for : - Calling suspend functions - Call UI functions - Updating LiveData
27. Dispatchers.IO Optimized for disk and network IO off the main thread Great for : - Database* - Reading/writing files - Networking$**
28. Dispatchers.Default Optimized for CPU intensive work off the main thread Great for : - Sorting a list - Parsing JSON - DiffUtils
29. Where thou Coroutines • One Shot request • Stream request
30. class SearchViewModel(val repo: SearchRepository): ViewModel() { private val _sortedResults = MutableLiveData<List<SearchResults>>() val sortedResults: LiveData<List<SearchResults>> = _sortedResults fun onSortAscending() = sortSearchBy(sort = true) fun onSortDescending() = sortSearchBy(sort = false) fun sortSearchBy(sort: Boolean, query: String) { viewModelScope.launch { // suspend and resume make this database request main-safe // so our ViewModel doesn't need to worry about threading _sortedResults.value = repo.fetchSearchResults(query, sort) } } }
31. class SearchViewModel(val repo: SearchRepository): ViewModel() { private val _sortedResults = MutableLiveData<List<SearchResults>>() val sortedResults: LiveData<List<SearchResults>> = _sortedResults fun onSortAscending() = sortSearchBy(ascending = true) fun onSortDescending() = sortSearchBy(ascending = false) fun sortSearchBy(sort: Boolean, query: String) { viewModelScope.launch { // suspend and resume make this database request main-safe // so our ViewModel doesn't need to worry about threading _sortedResults.value = repo.fetchSearchResults(query, sort) } } }
32. class SearchViewModel(val repo: SearchRepository): ViewModel() { private val _sortedResults = MutableLiveData<List<SearchResults>>() val sortedResults: LiveData<List<SearchResults>> = _sortedResults fun onSortAscending() = sortSearchBy(ascending = true) fun onSortDescending() = sortSearchBy(ascending = false) fun sortSearchBy(sort: Boolean, query: String) { viewModelScope.launch { // suspend and resume make this database request main-safe // so our ViewModel doesn't need to worry about threading _sortedResults.value = repo.fetchSearchResults(query, sort) } } }
33. class SearchViewModel(val repo: SearchRepository): ViewModel() { private val _sortedResults = MutableLiveData<List<SearchResults>>() val sortedResults: LiveData<List<SearchResults>> = _sortedResults fun onSortAscending() = sortSearchBy(ascending = true) fun onSortDescending() = sortSearchBy(ascending = false) fun sortSearchBy(sort: Boolean, query: String) { viewModelScope.launch { // suspend and resume make this database request main-safe // so our ViewModel doesn't need to worry about threading _sortedResults.value = repo.fetchSearchResults(query, sort) } } }
34. As a general pattern, start coroutines in the ViewModel.
35. class SearchRepository(val api: SearchAPI) { suspend fun fetchSearchResults( query: String, sort: Boolean ): List<SearchResults> { api.fetchResults(query, sort) .map(::mapToList) } }
36. class SearchRepository(val api: SearchAPI) { suspend fun fetchSearchResults( query: String, sort: Boolean ): List<SearchResults> { api.fetchResults(query, sort) .map(::mapToList) } }
37. class SearchRepository(val api: SearchAPI) { suspend fun fetchSearchResults( query: String, sort: Boolean ): List<SearchResults> { api.fetchResults(query, sort) .map(::mapToList) } }
38. //Rx fun fetchResults( query: String, sort: Boolean ): Single<SearchResponse> //Coroutine suspend fun fetchResults( query: String, sort: Boolean ): SearchResponse
39. //Rx fun fetchResults( query: String, sort: Boolean ): Single<SearchResponse> //Coroutine suspend fun fetchResults( query: String, sort: Boolean ): SearchResponse
40. //Rx fun fetchResults( query: String, sort: Boolean ): Single<SearchResponse> //Coroutine suspend fun fetchResults( query: String, sort: Boolean ): SearchResponse
41. //Fast and local fetch fun foo(param: Param) : Result //Slow and IO suspend fun foo(param: Param) : Result //BG fun CoroutineScope.foo(param: Param): Result
42. Asynchronicity Comparison • Observables in RxJava Observable.create<Int> { emitter -> for (i in 1..5) { emitter.onNext(i) } emitter.onComplete() }
43. Asynchronicity Comparison • Channels val channel = Channel<Int>() launch { // this might be heavy CPU-consuming computation or async logic //we'll just send five squares for (x in 1..5) channel.send(x) } // here we print five received integers: repeat(5) { println(channel.receive()) } println("Done!")
44. Asynchronicity Comparison • Channels val channel = Channel<Int>() launch { // this might be heavy CPU-consuming computation or async logic //we'll just send five squares for (x in 1..5) channel.send(x) channel.close() } // here we print five received integers: repeat(5) { println(channel.receive()) } println("Done!")
45. Asynchronicity Comparison • Channels val channel = Channel<Int>() launch { // this might be heavy CPU-consuming computation or async logic //we'll just send five squares for (x in 1..5) channel.send(x) channel.close() } // here we print five received integers: repeat(5) { println(channel.receive()) } println("Done!")
46. Asynchronicity Comparison • Channels val channel = Channel<Int>() launch { // this might be heavy CPU-consuming computation or async logic //we'll just send five squares for (x in 1..5) channel.send(x) channel.close() } // here we print five received integers: repeat(5) { println(channel.receive()) } println("Done!")
47. Asynchronicity Comparison • Channels val sequence = produceSequence() sequence.consumeEach { println(it) } println("Done!")
48. Asynchronicity Comparison • Channels ? Observable ? When you consume an object from the channel no other coroutine will be able to get the same object
49. Asynchronicity Comparison • Channels ~ Race Conditions ? val channel = Channel<Int>() launch { val value1 = channel.receive() } launch { val value2 = channel.receive() } launch { channel.send(1) }
50. Asynchronicity Comparison • Rx Subjects val subject: PublishSubject<Int> = PublishSubject.create() subject.subscribe { consumeValue(it) } subject.subscribe { println(it) }
51. Asynchronicity Comparison • BroadcastChannel val channel = BroadcastChannel<Int>(2) val observer1Job = launch { channel.openSubscription().use { channel -> for (value in channel) { consumeValue(value) } // subscription will be closed } }
52. Asynchronicity Comparison • BroadcastChannel val observer2Job = launch { channel.consumeEach { value -> consumeValue(value) } }
53. Asynchronicity Comparison Produce = Coroutine + Channel
54. Asynchronicity Comparison • Produce val publisher = produce(capacity = 2) { for (i in 1..5) send(i) } • Only the code inside produce can send elements to the channel
55. Asynchronicity Comparison Produce is a pretty useful way to create custom operators
56. Asynchronicity Comparison • Actor An actor also creates a coroutine with a channel built in. Produce implements the ReceiveChannel<E> interface whereas actor implements the SendChannel<E> interface.
57. Asynchronicity Comparison • Actor val actor = actor<Int>(CommonPool) { for (int in channel) { // iterate over received Integers } } launch { actor.send(2) }
58. Asynchronicity Comparison
59. Flow Flow is a Stream of Data, which can emit multiple values and complete with or without an error
60. Flow val myFlow: Flow<Int> = flow { emit(1) emit(2) delay(100) emit(3) } flow.collect { value -> println("Received $value") } Received 1 Received 2 Received 3 $$==>
61. Flow + Channels • Channels are Hot whereas Flows are cold • One Emits values from 1 side, receives from other side • A bit complex to work with
62. Flow + Channels • Hot + Cold = Hot • Ex: Flow from a channel
63. Flow + Channels object Action { private val channel = BroadcastChannel<Unit>(1) button.onClick { channel.send(Unit) } fun getFlow(): Flow<Unit> = channel.asFlow() } Action.getFlow().collect { println(“Received Emission”) }
64. Flow + Channels object Action { private val channel = BroadcastChannel<Unit>(1) button.onClick { channel.send(Unit) } fun getFlow(): Flow<Unit> = channel.asFlow() } Action.getFlow().collect { println(“Received Emission”) }
65. Flow + Channels object Action { private val channel = BroadcastChannel<Unit>(1) button.onClick { channel.send(Unit) } fun getFlow(): Flow<Unit> = channel.asFlow() } Action.getFlow().collect { println(“Received Emission”) }
66. Flow + Channels object Action { private val channel = BroadcastChannel<Unit>(1) button.onClick { channel.send(Unit) } fun getFlow(): Flow<Unit> = channel.asFlow() } Action.getFlow().collect { println(“Received Emission”) }
67. Flow and Rx compositeDisposable.add(api.search(query, sort) .map{response -> response.toList() } .flatMap { searchResults -> cache.saveAndReturn(searchResults) } .subscribeOn(Schedulers.io()) .observeOn(AndroidSchedulers.mainThread()) .subscribe( { searchResults -> displayResults(searchResults) }, { error -> Log.e("TAG", "Failed to search", error) } ))
68. Flow and Rx try { flow { emit(api.search(query, sort))} .map{response ->response.toList()} .flatMapConcat { searchResults -> cache.saveAndReturn(searchResults) } .flowOn(Dispatchers.IO) .collect { searchResultList -> withContext(Dispatchers.Main) { showSearchResult(searchResultList) } } } catch(e: Exception) { e.printStackTrace() }
69. Rx / Suspend //Rx Observable.interval(1, TimeUnit.SECONDS) .subscribe { textView.text = "$it seconds have passed" } //Coroutine GlobalScope.launch { var i = 0 while (true){ textView.text = "${it++} seconds have passed" delay(1000) } }
70. Rx / Suspend publishSubject .debounce(300, TimeUnit.MILLISECONDS) .distinctUntilChanged() .switchMap { searchQuery -> api.fetchResults(query, sort) } .subscribeOn(Schedulers.io()) .observeOn(AndroidSchedulers.mainThread()) .subscribe({ ... })
71. Rx / Suspend launch { broadcast.consumeEach { delay(300) val results = api.fetchResults(it) .await() displayResultsLogic() ) } }
72. Rx / Suspend //Rx searchQuery.addTextChangedListener(object: TextWatcher { override fun afterTextChanged(s: Editable?) { publishSubject.onNext(s.toString()) } } //Coroutines searchQuery.addTextChangedListener(object: TextWatcher { override fun afterTextChanged(s: Editable?) { broadcast.send(s.toString()) } }
73. bit.ly/droidcon_in_async Thats all folks! @aditlal Adit Lal adit.dev @rivuchakraborty Rivu chakraborty rivu.dev