How to combine two serial flows which input depends on other output? - android

I have sophisticated scenario where a set of mutually dependent coroutine flows depends on each other and chained:
viewModelScope.launch {
repository.cacheAccount(person)
.flatMapConcat { it->
Log.d(App.TAG, "[2] create account call (server)")
repository.createAccount(person)
}
.flatMapConcat { it ->
if (it is Response.Data) {
repository.cacheAccount(it.data)
.collect { it ->
// no op, just execute the command
Log.d(App.TAG, "account has been cached")
}
}
flow {
emit(it)
}
}
.catch { e ->
Log.d(App.TAG, "[3] get an exception in catch block")
Log.e(App.TAG, "Got an exception during network call", e)
state.update { state ->
val errors = state.errors + getErrorMessage(PersonRepository.Response.Error.Exception(e))
state.copy(errors = errors, isLoading = false)
}
}
.collect { it ->
Log.d(App.TAG, "[4] collect the result")
updateStateProfile(it)
}
}
cache an account on the local disk
create an account on the backend
in positive scenario, cache the newly create account in the local disk
Now I have to add more calls to a new API endpoint and the scenario become even more sophisticated. This endpoint is a ethereum chain.
4a. In the positive scenario, put in the local disk (cache) initiated transaction cacheRepository.createChainTx()
4b. In the negative scenario, just emit further the response from the backend
4a.->5. Register user on the 2nd endpoint repository.registerUser()
The response from 2nd endpoint put in the cache by updating existing row. Even negative case except of exception should be cached to update status of tx.
viewModelScope.launch {
lateinit var newTx: ITransaction
cacheRepository.createChainTxAsFlow(RegisterUserTransaction(userWalletAddress = userWalletAddress))
.map { it ->
newTx= it
repository.registerUserOnSwapMarket(userWalletAddress)
}
.onEach { it -> preProcessResponse(it, newTx) }
.flowOn(backgroundDispatcher)
.collect { it -> processResponse(it) }
}
This a scenario which should be integrated into the 1st Flow chain.
The issue is I do not see how to do it clear in Flow chain. I can rewrite code without chaining, but it also bring variety if else statements.
How would you do this scenario in human readable way?

I'll ended up with this code for transition period:
viewModelScope.launch(backgroundDispatcher) {
try {
var cachedPersonProfile = repository.cacheAccount(person)
var createAccountResponse = repository.createAccount(person)
when(createAccountResponse) {
is Response.Data -> {
repository.cacheAccount(createAccountResponse.data)
val cachedTx = cacheRepository.createChainTx(RegisterUserTransaction(userWalletAddress = person.userWalletAddress))
val chainTx = walletRepository.registerUserOnSwapMarket(userWalletAddress = person.userWalletAddress)
when(chainTx) {
is ru.home.swap.core.network.Response.Data -> {
if (chainTx.data.isStatusOK()) {
cachedTx.status = TxStatus.TX_MINED
} else {
cachedTx.status = TxStatus.TX_REVERTED
}
}
is ru.home.swap.core.network.Response.Error.Message -> {
cachedTx.status = TxStatus.TX_EXCEPTION
}
is ru.home.swap.core.network.Response.Error.Exception -> {
cachedTx.status = TxStatus.TX_EXCEPTION
}
}
cacheRepository.createChainTx(cachedTx)
withContext(Dispatchers.Main) {
state.update { state ->
if (cachedTx.status == TxStatus.TX_MINED) {
state.copy(
isLoading = false,
profile = createAccountResponse.data,
status = StateFlagV2.PROFILE
)
} else {
val txError = "Failed register the profile on chain with status ${TxStatus.TX_MINED}"
state.copy(
isLoading = false,
errors = state.errors + txError
)
}
}
}
}
else -> { updateStateProfile(createAccountResponse) }
}
} catch (ex: Exception) {
withContext(Dispatchers.Main) {
state.update { state ->
val errors = state.errors + getErrorMessage(PersonRepository.Response.Error.Exception(ex))
state.copy(errors = errors, isLoading = false)
}
}
}
}
If you have a better alternative, please share it in the post as an answer.

Related

Fragment popbackstack trigger lifecyclescope collect

Situation
I submit data setTripDeliver, the collect works fine (trigger LOADING and then SUCCESS). I pressed a button go to next fragment B (using replace). After that, I press back button (using popbackstack). the collect SUCCESS triggered.
Codes Related
These codes at the FragmentA.kt inside onViewCreated.
private fun startLifeCycle() {
viewLifecycleOwner.lifecycleScope.launch {
viewLifecycleOwner.lifecycle.repeatOnLifecycle(Lifecycle.State.STARTED) {
launch {
collectTripDeliver()
}
launch {
collectTripReattempt()
}
}
}
}
These codes when to submit data at a button setOnClickListener.
viewLifecycleOwner.lifecycleScope.launchWhenStarted {
viewModel.setTripDeliver(
verificationCode,
remark
)
}
Method to collect flow collectTripReattempt()
private suspend fun collectTripReattempt() {
viewModel.tripReattempt.collect {
when (it) {
is Resource.Initialize -> {
}
is Resource.Loading -> {
Log.i("???","collectTripReattempt loading")
handleSaveEarly()
}
is Resource.Success -> {
val error = it.data?.error
if (error == null) {
Tools.showToast(requireContext(), "Success Reattempt")
Log.i("???","collectTripReattempt Success")
} else {
Tools.showToast(requireContext(), "$error")
}
handleSaveEnding()
}
is Resource.Error -> {
handleSaveEnding()
}
}
}
}
Below codes are from ViewModel.
private val _tripDeliver =
MutableStateFlow<Resource<TripDeliverResponse>>(Resource.Initialize())
val tripDeliver: StateFlow<Resource<TripDeliverResponse>> = _tripDeliver
This method to call repository.
suspend fun setTripDeliver(
verificationCode: String?,
remark: String?
) {
_tripDeliver.value = Resource.Loading()
try {
val result = withContext(ioDispatcher) {
val tripDeliverParameter = DeliverParameter(
verificationCode,
remark
)
val response = appRepository.setTripDeliver(tripDeliverParameter)
Resource.getResponse { response }
}
_tripDeliver.value = result
} catch (e: Exception) {
when (e) {
is IOException -> _tripDeliver.value =
Resource.Error(messageInt = R.string.no_internet_connection)
else -> _tripDeliver.value =
Resource.Error("Trip Deliver Error: " + e.message)
}
}
}
Logcat
2021-07-09 19:56:10.946 7446-7446/com.package.app I/???: collectTripReattempt loading
2021-07-09 19:56:11.172 7446-7446/com.package.app I/???: collectTripReattempt Success
2021-07-09 19:56:17.703 7446-7446/com.package.app I/???: collectTripReattempt Success
As you can see, the last Success is called again AFTER I pressed back button (popbackstack)
Question
How to make it trigger once only? Is it the way I implement it is wrong? Thank you in advance.
This is not problem of your implementation this is happening because of stateIn() which use used in your viewModel to convert regular flow into stateFlow
If according to your code snippet the success is triggered once again, then why not loading has triggered?
as per article, it is showing the latest cached value when you left the screen and came back you got the latest cached value on view.
Resource:
https://medium.com/androiddevelopers/migrating-from-livedata-to-kotlins-flow-379292f419fb
The latest value will still be cached so that when the user comes back to it, the view will have some data immediately.
I have found the solution, thanks to #Nurseyit Tursunkulov for giving me a clue. I have to use SharedFlow.
At the ViewModel, I replace the initialize with these:
private val _tripDeliver = MutableSharedFlow<Resource<TripDeliverResponse>>(replay = 0)
val tripDeliver: SharedFlow<Resource<TripDeliverResponse>> = _tripDeliver
At the replay I have to use 0, so this SharedFlow will trigger once. Next, change _tripDeliver.value to _tripDeliver.emit() like the codes below:
fun setTripDeliver(
verificationCode: String?,
remark: String?
) = viewModelScope.launch {
_tripDeliver.emit(Resource.Loading())
if (verificationCode == null && remark == null) {
_tripDeliver.emit(Resource.Error("Remark cannot be empty if verification is empty"))
return#launch
}
try {
val result = withContext(ioDispatcher) {
val tripDeliverParameter = DeliverParameter(
verificationCode,
remark,
)
val response = appRepository.setTripDeliver(tripDeliverParameter)
Resource.getResponse { response }
}
_tripDeliver.emit(result)
} catch (e: Exception) {
when (e) {
is IOException -> _tripDeliver.emit(Resource.Error(messageInt = R.string.no_internet_connection))
else -> _tripDeliver.emit(Resource.Error("Trip Deliver Error: " + e.message))
}
}
}
I hope this answer will help the others also.
I think this is because of coldFlow, you need a HotFlow. Another option is to try to hide and show fragment, instead of replacing. And yet another solution is to keep this code in viewModel.
In my opinion, I think your way of using coroutines in lifeScope is incorrect. After the lifeScope status of FragmentA is at Started again, the coroutine will be restarted:
launch {
collectTripDeliver()
}
launch {
collectTripReattempt()
}
So I think: You need to modify this way:
private fun startLifeCycle() {
viewLifecycleOwner.lifecycleScope.launch {
launch {
collectTripDeliver()
}
launch {
collectTripReattempt()
}
}
}

RxJava ConcatArrayDelayError and filters: returning an error only if both sources fail

I'm new to RxJava and after a few days of trying everything I could find online I see that I really need help with this one.
I fetch a member in my repository with local and remote sources. I added some operators to return my remote source in priority (via debounce), and to filter out errors so it would return only 1 of the 2 if either remote is not available or the database is empty.
It works fine as long as something is returned by one of my 2 sources, but the problem occurs if both sources returns errors: as I filter out the errors, it doesn't return anything, and my subscribe is never called.
Maybe there is a simple solution but I have not found it so far, could someone help?
Here is my fetchMember() in my Repository:
override fun fetchMember(): Observable<MemberModel?> {
return Observable.concatArrayDelayError(memberLocalSource.fetchMember(), memberRemoteSource.fetchMember())
.doOnNext { member ->
saveMember(member!!)
}
.materialize()
.filter { !it.isOnError }
.dematerialize { it -> it }
.debounce(400, TimeUnit.MILLISECONDS)
}
And here is my viewmodel:
fun fetchToken(username: String, password: String) {
val loginDisposable = authApiService.loginWithJWT(username, password)
.flatMap {
isAuthenticated = isTokenValid(username, password, it)
sharedPreferences.setHasValidCredentials(isAuthenticated)
memberRepository.fetchMember()
}
.subscribeOn(Schedulers.io())
.observeOn((AndroidSchedulers.mainThread()))
.doOnError { throwable ->
throwable.printStackTrace()
}
.subscribe(
{ member ->
memberLiveData.value = member
this.memberId = member!!.id.toString()
this.memberName = member.name.split(" ")[0]
if(isAuthenticated) {
authenticationState.value = AuthenticationState.AUTHENTICATED_VALID_MEMBER
} else {
authenticationState.value = AuthenticationState.UNAUTHENTICATED_VALID_MEMBER
}
},
{ error ->
if(isAuthenticated) {
authenticationState.value = AuthenticationState.AUTHENTICATED_INVALID_MEMBER
} else {
authenticationState.value = AuthenticationState.INVALID_AUTHENTICATION
}
})
disposable.add(loginDisposable)
}
private fun isTokenValid(username: String, password: String, authResponse: AuthModel): Boolean {
return if (authResponse.data != null) {
false
} else {
tokenInterceptor.token = authResponse.token
val tokenWithCredentials = AuthModel(authResponse.token, null, null, username, password)
tokenRepository.saveToken(tokenWithCredentials)
true
}
}
In the end I managed to make it work by adding:
.defaultIfEmpty(MemberModel(-1))
and checking against id == -1.

Firebase realtime snapshot listener using Coroutines

I want to be able to listen to realtime updates in Firebase DB's using Kotlin coroutines in my ViewModel.
The problem is that whenever a new message is created in the collection my application freezes and won't recover from this state. I need to kill it and restart app.
For the first time it passes and I can see the previous messages on the UI. This problem happens when SnapshotListener is called for 2nd time.
My observer() function
val channel = Channel<List<MessageEntity>>()
firestore.collection(path).addSnapshotListener { data, error ->
if (error != null) {
channel.close(error)
} else {
if (data != null) {
val messages = data.toObjects(MessageEntity::class.java)
//till this point it gets executed^^^^
channel.sendBlocking(messages)
} else {
channel.close(CancellationException("No data received"))
}
}
}
return channel
That's how I want to observe messages
launch(Dispatchers.IO) {
val newMessages =
messageRepository
.observer()
.receive()
}
}
After I replacing sendBlocking() with send() I am still not getting any new messages in the channel. SnapshotListener side is executed
//channel.sendBlocking(messages) was replaced by code bellow
scope.launch(Dispatchers.IO) {
channel.send(messages)
}
//scope is my viewModel
How to observe messages in firestore/realtime-dbs using Kotlin coroutines?
I have these extension functions, so I can simply get back results from the query as a Flow.
Flow is a Kotlin coroutine construct perfect for this purposes.
https://kotlin.github.io/kotlinx.coroutines/kotlinx-coroutines-core/kotlinx.coroutines.flow/-flow/
#ExperimentalCoroutinesApi
fun CollectionReference.getQuerySnapshotFlow(): Flow<QuerySnapshot?> {
return callbackFlow {
val listenerRegistration =
addSnapshotListener { querySnapshot, firebaseFirestoreException ->
if (firebaseFirestoreException != null) {
cancel(
message = "error fetching collection data at path - $path",
cause = firebaseFirestoreException
)
return#addSnapshotListener
}
offer(querySnapshot)
}
awaitClose {
Timber.d("cancelling the listener on collection at path - $path")
listenerRegistration.remove()
}
}
}
#ExperimentalCoroutinesApi
fun <T> CollectionReference.getDataFlow(mapper: (QuerySnapshot?) -> T): Flow<T> {
return getQuerySnapshotFlow()
.map {
return#map mapper(it)
}
}
The following is an example of how to use the above functions.
#ExperimentalCoroutinesApi
fun getShoppingListItemsFlow(): Flow<List<ShoppingListItem>> {
return FirebaseFirestore.getInstance()
.collection("$COLLECTION_SHOPPING_LIST")
.getDataFlow { querySnapshot ->
querySnapshot?.documents?.map {
getShoppingListItemFromSnapshot(it)
} ?: listOf()
}
}
// Parses the document snapshot to the desired object
fun getShoppingListItemFromSnapshot(documentSnapshot: DocumentSnapshot) : ShoppingListItem {
return documentSnapshot.toObject(ShoppingListItem::class.java)!!
}
And in your ViewModel class, (or your Fragment) make sure you call this from the right scope, so the listener gets removed appropriately when the user moves away from the screen.
viewModelScope.launch {
getShoppingListItemsFlow().collect{
// Show on the view.
}
}
What I ended up with is I used Flow which is part of coroutines 1.2.0-alpha-2
return flowViaChannel { channel ->
firestore.collection(path).addSnapshotListener { data, error ->
if (error != null) {
channel.close(error)
} else {
if (data != null) {
val messages = data.toObjects(MessageEntity::class.java)
channel.sendBlocking(messages)
} else {
channel.close(CancellationException("No data received"))
}
}
}
channel.invokeOnClose {
it?.printStackTrace()
}
}
And that's how I observe it in my ViewModel
launch {
messageRepository.observe().collect {
//process
}
}
more on topic https://medium.com/#elizarov/cold-flows-hot-channels-d74769805f9
Extension function to remove callbacks
For Firebase's Firestore database there are two types of calls.
One time requests - addOnCompleteListener
Realtime updates - addSnapshotListener
One time requests
For one time requests there is an await extension function provided by the library org.jetbrains.kotlinx:kotlinx-coroutines-play-services:X.X.X. The function returns results from addOnCompleteListener.
For the latest version, see the Maven Repository, kotlinx-coroutines-play-services.
Resources
Using Firebase on Android with Kotlin Coroutines by Joe Birch
Using Kotlin Extension Functions and Coroutines with Firebase by Rosário Pereira Fernandes
Realtime updates
The extension function awaitRealtime has checks including verifying the state of the continuation in order to see whether it is in isActive state. This is important because the function is called when the user's main feed of content is updated either by a lifecycle event, refreshing the feed manually, or removing content from their feed. Without this check there will be a crash.
ExtenstionFuction.kt
data class QueryResponse(val packet: QuerySnapshot?, val error: FirebaseFirestoreException?)
suspend fun Query.awaitRealtime() = suspendCancellableCoroutine<QueryResponse> { continuation ->
addSnapshotListener({ value, error ->
if (error == null && continuation.isActive)
continuation.resume(QueryResponse(value, null))
else if (error != null && continuation.isActive)
continuation.resume(QueryResponse(null, error))
})
}
In order to handle errors the try/catch pattern is used.
Repository.kt
object ContentRepository {
fun getMainFeedList(isRealtime: Boolean, timeframe: Timestamp) = flow<Lce<PagedListResult>> {
emit(Loading())
val labeledSet = HashSet<String>()
val user = usersDocument.collection(getInstance().currentUser!!.uid)
syncLabeledContent(user, timeframe, labeledSet, SAVE_COLLECTION, this)
getLoggedInNonRealtimeContent(timeframe, labeledSet, this)
}
// Realtime updates with 'awaitRealtime' used
private suspend fun syncLabeledContent(user: CollectionReference, timeframe: Timestamp,
labeledSet: HashSet<String>, collection: String,
lce: FlowCollector<Lce<PagedListResult>>) {
val response = user.document(COLLECTIONS_DOCUMENT)
.collection(collection)
.orderBy(TIMESTAMP, DESCENDING)
.whereGreaterThanOrEqualTo(TIMESTAMP, timeframe)
.awaitRealtime()
if (response.error == null) {
val contentList = response.packet?.documentChanges?.map { doc ->
doc.document.toObject(Content::class.java).also { content ->
labeledSet.add(content.id)
}
}
database.contentDao().insertContentList(contentList)
} else lce.emit(Error(PagedListResult(null,
"Error retrieving user save_collection: ${response.error?.localizedMessage}")))
}
// One time updates with 'await' used
private suspend fun getLoggedInNonRealtimeContent(timeframe: Timestamp,
labeledSet: HashSet<String>,
lce: FlowCollector<Lce<PagedListResult>>) =
try {
database.contentDao().insertContentList(
contentEnCollection.orderBy(TIMESTAMP, DESCENDING)
.whereGreaterThanOrEqualTo(TIMESTAMP, timeframe).get().await()
.documentChanges
?.map { change -> change.document.toObject(Content::class.java) }
?.filter { content -> !labeledSet.contains(content.id) })
lce.emit(Lce.Content(PagedListResult(queryMainContentList(timeframe), "")))
} catch (error: FirebaseFirestoreException) {
lce.emit(Error(PagedListResult(
null,
CONTENT_LOGGED_IN_NON_REALTIME_ERROR + "${error.localizedMessage}")))
}
}
This is working for me:
suspend fun DocumentReference.observe(block: suspend (getNextSnapshot: suspend ()->DocumentSnapshot?)->Unit) {
val channel = Channel<Pair<DocumentSnapshot?, FirebaseFirestoreException?>>(Channel.UNLIMITED)
val listenerRegistration = this.addSnapshotListener { value, error ->
channel.sendBlocking(Pair(value, error))
}
try {
block {
val (value, error) = channel.receive()
if (error != null) {
throw error
}
value
}
}
finally {
channel.close()
listenerRegistration.remove()
}
}
Then you can use it like:
docRef.observe { getNextSnapshot ->
while (true) {
val value = getNextSnapshot() ?: continue
// do whatever you like with the database snapshot
}
}
If the observer block throws an error, or the block finishes, or your coroutine is cancelled, the listener is removed automatically.

Testing suspend corotuine

I am trying to use coroutines to handle asynchronous code for my login service. Unfortunately, the implementation of the login service must accept callbacks when it completes. I do not want this login() function to complete until one of these callbacks occurs.
Here is what I have:
fun login(): Outcome = runBlocking {
suspendCoroutine<Outcome> { continuation ->
loginService.login(
onLoginSuccess = {
// do some stuff
continuation.resume(Outcome.SUCCESS)
},
onLoginFailure = {
// handle failure case
continuation.resume(Outcome.FAILURE)
}
)
}
}
My issue is my tests never complete. I think what is happening is that the continuation block itself isn't running. I tried wrapping the call to uut.login() in a runBlocking as well, but it didn't help. Here is my test code (using Spek):
describe("when login") {
val successCaptor: ArgumentCaptor<() -> Unit> = TestHelpers.argumentCaptorForClass()
val failureCaptor: ArgumentCaptor<() -> Unit> = TestHelpers.argumentCaptorForClass()
var result: Outcome? = null
beforeEachTest {
doNothing().whenever(mockLoginService)?.login(capture(successCaptor), capture(failureCaptor))
result = uut?.execute()
}
it("logs in with the login service") {
verify(mockLoginService)?.login(any(), any())
}
describe("and the login succeeds") {
beforeEachTest {
successCaptor.value.invoke()
}
// other tests...
it("returns an outcome of SUCCESS") {
expect(result).to.equal(Outcome.SUCCESS)
}
}
describe("and the login fails") {
beforeEachTest {
failureCaptor.value.invoke()
}
// other tests...
it("returns an outcome of FAILURE") {
expect(result).to.equal(Outcome.FAILURE)
}
}
}
Basically, I'd like to assert that the login() method returned either a SUCCESS or FAILURE outcome based on what occurred.
Any ideas?
Of course, I figured this out right after posting. If interested, here is what I did in the test:
describe("when login") {
val successCaptor: ArgumentCaptor<() -> Unit> = TestHelpers.argumentCaptorForClass()
val failureCaptor: ArgumentCaptor<() -> Unit> = TestHelpers.argumentCaptorForClass()
var result: Outcome? = null
describe("and the login succeeds") {
beforeEachTest {
whenever(mockLoginService?.login(capture(successCaptor), capture(failureCaptor))).thenAnswer {
successCaptor.value.invoke()
}
result = uut?.execute()
}
it("logs in with the login service") {
verify(mockLoginService)?.login(any(), any())
}
it("returns an outcome of SUCCESS") {
expect(result).to.equal(Outcome.SUCCESS)
}
}
describe("and the login fails") {
beforeEachTest {
whenever(mockLoginService?.login(capture(successCaptor), capture(failureCaptor))).thenAnswer {
failureCaptor.value.invoke()
}
result = uut?.execute()
}
it("logs in with the login service") {
verify(mockLoginService)?.login(any(), any())
}
// other tests
it("returns an outcome of FAILURE") {
expect(result).to.equal(Outcome.FAILURE)
}
}
}

retryWhen on RxJava with Rx2Apollo

I'm using Rx2Apollo to make a graphql call:
private fun registerCardToken(token: String): io.reactivex.Observable<RegisterCardTokenMutation.RegisterCreditCard> {
val apolloCall = apolloClient().mutate(RegisterCardTokenMutation.builder().token(token).build())
return Rx2Apollo.from(apolloCall).map {
(it.data() as RegisterCardTokenMutation.Data).registerCreditCard()
}.doOnError({ error ->
//Log.e("registerCardToke", error.message)
})
}
This works well, but I want to handle specific error and retry this onces. I have tried to work around this using retryWhen and retry , but not able to write any executable code yet.
The retry persons a token refresh before performing the actual retry. Here's the token refresh sample:
private fun refreshBearerToken(callback: OnCompleteListener<GetTokenResult>) {
FirebaseAuth.getInstance().currentUser?.getIdToken(true)?.addOnCompleteListener(callback)
}
First, you have to turn refreshBearerToken into an Observable
val refreshTokenSource = Observable.create({ emitter ->
FirebaseAuth.
getInstance().
currentUser?.
getIdToken(true)?.
addOnCompleteListener({ task ->
if (task.isSuccessful()) {
emitter.onNext(task.getResult())
emitter.onComplete()
} else {
emitter.onError(task.getException())
}
})
})
Second, use some external reference holding the current token and conditionally use it before calling registerCardToken:
val currentToken = AtomicReference<String>()
val registerCardTokenObservable = Observable.defer({
val token = currentToken.get()
if (token == null) {
return refreshTokenSource
.doOnNext({ currentToken.set(it) })
.flatMap({ registerCardToken(it) })
}
return registerCardToken(token)
})
.retry({ error ->
if ((error is IOException) || (error.getMessage().contains("network")) {
currentToken.set(null)
return true
}
return false
})

Categories

Resources