I this piece of code I would like to know instead of having convert() separate can I have anonymous in map {}
fun <A, B> LiveData<A>.map(function: (A) -> B): LiveData<B> = Transformations.map(this, function)
fun loadSettings() {
configLiveData.map { configFile ->
return#map convert(configFile)
}
}
fun convert(configFile: Response<ConfigFile>): MutableLiveData<Settings> {
val mutableData = MutableLiveData<Setting>()
when (configFile) {
is Response.Success<ConfigFile> -> {
mutableData.postValue(configFile.data.config?.settings)
}
is Response.Failure -> {
errorMessageMutableData.postValue(it.message)
}
}
return mutableData
}
final result I have
fun loadTheme(): LiveData<Response<Theme?>> {
return configLiveData.map { configFile ->
when (configFile) {
is Response.Success<ConfigFile> -> {
Response.Success(configFile.data.config?.theme)
}
is Response.Failure -> {
Response.Failure(configFile.message)
}
}
}
}
can I have anonymous in map {}
Yes, of course. You already do: { configFile -> return#map convert(configFile) } is a lambda (which could equally be written { configFile -> convert(configFile) } or { convert(it) }, or even ::convert). If you don't want to make convert a separate function, just inline it into the lambda:
configLiveData.map { configFile ->
val mutableData = MutableLiveData<Setting>()
when (configFile) {
is Response.Success<ConfigFile> -> {
mutableData.postValue(configFile.data.config?.settings)
}
is Response.Failure -> {
errorMessageMutableData.postValue(it.message)
}
}
mutableData // no need for return#map
}
But the problem is that if that's your real code, it probably doesn't actually do what you want, because
you create a LiveData<MutableLiveData<Settings>> (did you want switchMap instead of map?);
you then throw it away;
if you get a Response.Success, it's effectively ignored (because you just post data from it into an unobserved LiveData).
Related
I'm seeing some odd behavior. I have a simple StateFlow<Boolean> in my ViewModel that is not being collected in the fragment. Definition:
private val _primaryButtonClicked = MutableStateFlow(false)
val primaryButtonClicked: StateFlow<Boolean> = _primaryButtonClicked
and here is where I set the value:
fun primaryButtonClick() {
_primaryButtonClicked.value = true
}
Here is where I'm collecting it.
repeatOnOwnerLifecycle {
launch(dispatchProvider.io()) {
freeSimPurchaseFragmentViewModel.primaryButtonClicked.collect {
if (it) {
autoCompletePlacesStateFlowModel.validateErrors()
formValidated = autoCompletePlacesStateFlowModel.validateAddress()
if (formValidated) {
freeSimPurchaseFragmentViewModel
.sumbitForm(autoCompletePlacesStateFlowModel.getStateFlowCopy())
}
}
}
}
}
repeatOnOwnerLifecycle:
inline fun Fragment.repeatOnOwnerLifecycle(
state: Lifecycle.State = Lifecycle.State.RESUMED,
crossinline block: suspend CoroutineScope.() -> Unit
) {
viewLifecycleOwner.lifecycleScope.launch {
repeatOnLifecycle(state) {
block()
}
}
What am I doing wrong? The collector never fires.
Does this make sense?
val primaryButtonClicked: StateFlow<Boolean> = _primaryButtonClicked.asStateFlow()
Also I couldn't understand the inline function part, because under the hood seems you wrote something like this
viewLifecycleOwner.lifecycleScope.launch {
viewLifecycleOwner.repeatOnLifecycle(Lifecycle.State.RESUMED) {
launch(dispatchProvider.io()) {
freeSimPurchaseFragmentViewModel.primaryButtonClicked.collect {
if (it) {
autoCompletePlacesStateFlowModel.validateErrors()
formValidated = autoCompletePlacesStateFlowModel.validateAddress()
if (formValidated) {
freeSimPurchaseFragmentViewModel
.sumbitForm(autoCompletePlacesStateFlowModel.getStateFlowCopy())
}
}
}
}
}
}
Why are you launching one coroutine in another and collect the flow from IO dispatcher? You need to collect the values from the main dispatcher.
I have a situation where I have to execute 3 network requests one after the other collect their results (which are of different types).
Following is the relevant part of the code :
Resource.kt
sealed class Resource<T>(val data: T? = null, val message: String? = null) {
class Loading<T>(data: T? = null): Resource<T>(data)
class Success<T>(data: T?): Resource<T>(data)
class Error<T>(message: String, data: T? = null): Resource<T>(data, message)
}
Repository.kt
override fun getReportData(profileId: Int): Flow<Resource<ProfileReport>> =
flow {
emit(Resource.Loading<ProfileReport>())
var report: ProfileReport? = null
try {
// Api is available as a retrofit implementation
report = api.getReport(profileId).toProfileReport()
} catch (e: HttpException) {
emit(
Resource.Error<ProfileReport>(
message = "An unknown http exception occured"
)
)
}
if (report!= null) {
emit(Resource.Success<ProfileReport>(data = report))
}
}
Say I have 3 such flows to fetch data in my repository and they have different return types (ex: ProfileReport, ProfileInfo, ProfileStatus).
Now in my viewmodel I have a function to execute these flows and perform actions on the values emitted such as :
ViewModel.kt
fun getProfileData(profileId: Int) {
getReportData(profileId)
.onEach { result ->
when (result) {
is Resource.Loading -> {
_loading.value = true
}
is Resource.Error -> {
_loading.value = false
// UI event to display error snackbar
}
is Resource.Success -> {
_loading.value = false
if (result.data != null) {
_report.value = _report.value.copy(
// Use result here
)
}
}
}
}.launchIn(viewModelScope)
}
This works ok for one flow but how can I execute 3 flows one after the other.
That is, execute first one and if its successful, execute second one and so on, and if all of them are successful use the results.
I did it like this :
fun getProfileData(profileId: Int) {
getReportData(profileId)
.onEach { result1 ->
when (result1) {
is Resource.Loading -> {/*do stuff*/}
is Resource.Error -> {/*do stuff*/}
is Resource.Success -> {
getProfileStatus(profileId)
.onEach { result2 ->
is Resource.Loading -> {/*do stuff*/}
is Resource.Error -> {/*do stuff*/}
is Resource.Success -> {
getProfileInfo(profileId)
.onEach { result3 ->
is Resource.Loading -> {/*do stuff*/}
is Resource.Error -> {/*do stuff*/}
is Resource.Success -> {
/*
Finally update viewmodel state
using result1, result2 and result3
*/
}
}.launchIn(viewModelScope)
}
}.launchIn(viewModelScope)
}
}
}.launchIn(viewModelScope)
}
But, this feels too cumbersome and probably there is a better way to chain flows based on success condition and collect results at the end. I checked some ways that use combine() or flatMapMerge() but was unable to use them in this situation.
Is there a way to achieve this? Or is this approach itself wrong from a design perspective maybe?
I think this could be modeled much more cleanly using imperative coroutines than with flows. Since you're overriding functions, this depends on you being able to modify the supertype abstract function signatures.
This solution doesn't use Resource.Loading, so you should remove that to make smart casting easier.
suspend fun getReportData(profileId: Int): Resource<ProfileReport> =
try {
val report = api.getReport(profileId).toProfileReport()
Resource.Success<ProfileReport>(data = report)
} catch (e: HttpException) {
Resource.Error<ProfileReport>(
message = "An unknown http exception occured"
)
}
//.. similar for the other two functions that used to return flows.
fun getProfileData(profileId: Int) {
viewModelScope.launch {
// do stuff to indicate 1st loading state
when(val result = getReportData(profileId)) {
Resource.Error<ProfileReport> -> {
// do stuff for error state
return#launch
}
Resource.Success<ProfileReport> -> {
// do stuff with result
}
}
// Since we returned when there was error, we know first
// result was successful.
// do stuff to indicate 2nd loading state
when(val result = getProfileStatus(profileId)) {
Resource.Error<ProfileStatus> -> {
// do stuff for error state
return#launch
}
Resource.Success<ProfileStatus> -> {
// do stuff with result
}
}
// do stuff to indicate 3rd loading state
when(val result = getProfileInfo(profileId)) {
Resource.Error<ProfileInfo> -> {
// do stuff for error state
return#launch
}
Resource.Success<ProfileInfo> -> {
// do stuff with result
}
}
}
}
If you want to keep your current Flows, you could collect your flows this way to avoid the deep nesting. This works because your source flows are designed to be finite (they aren't repeatedly emitting new values indefinitely, but have only one final result).
fun getProfileData(profileId: Int) = viewModelScope.launch {
var shouldBreak = false
getReportData(profileId).collect { result ->
when (result) {
is Resource.Loading -> { /*do stuff*/ }
is Resource.Error -> {
/*do stuff*/
shouldBreak = true
}
is Resource.Success -> { /*do stuff*/ }
}
}
if (shouldBreak) return#launch
getProfileStatus(profileId).collect { result ->
when (result) {
is Resource.Loading -> { /*do stuff*/ }
is Resource.Error -> {
/*do stuff*/
shouldBreak = true
}
is Resource.Success -> { /*do stuff*/ }
}
}
if (shouldBreak) return#launch
getProfileInfo(profileId).collect { result ->
when (result) {
is Resource.Loading -> { /*do stuff*/ }
is Resource.Error -> { /*do stuff*/ }
is Resource.Success -> { /*do stuff*/ }
}
}
}
I have two quite similar functions and I'm trying to avoid duplication in my code by the use of generics. The functions have both a try catch block and notify its observers with two MutableLiveData of two different types:
val noWasteRecipesPosts: MutableLiveData<List<Recipe>> = MutableLiveData()
val lastArticlesPosts: MutableLiveData<List<Article>> = MutableLiveData()
fun getNoWasteRecipesPosts() {
makeCall(service.getRecipes(), noWasteRecipesPosts)
scope.launch {
try {
val response = service.getRecipes().await()
when (response.isSuccessful) {
true -> {
response.body()?.let {
noWasteRecipesPosts.postValue(ArrayList(response.body()))
} ?: run {
errorLiveData.postValue(response.message())
}
}
false -> errorLiveData.postValue(response.message())
}
} catch (e: Exception) {
noConnectionLiveData.postValue(true)
}
}
}
fun getLastArticlesPosts(excludeRecipes: Boolean) {
scope.launch {
try {
val response = when (excludeRecipes) {
true -> service.getLastArticles(categoriesToExclude = arrayListOf(BlogCategories.NO_WASTE_RECIPES.id))
.await()
false -> service.getLastArticles()
.await()
}
when (response.isSuccessful) {
true -> {
response.body()?.let {
lastArticlesPosts.postValue(ArrayList(response.body()))
} ?: run {
errorLiveData.postValue(response.message())
}
}
false -> errorLiveData.postValue(response.message())
}
} catch (e: Exception) {
noConnectionLiveData.postValue(true)
}
}
}
To avoid code repeating I'm trying to use generics, but probably in the wrong way. I've defined a function that takes the Deferred api response as first parameter and I would like to pass a MutableLiveData to notify observers as the second parameter:
fun makeCall(function: Deferred<Response<*>>, successLiveData: MutableLiveData<*>) {
scope.launch {
try {
val response = function.await()
when (response.isSuccessful) {
true -> {
response.body()?.let {
successLiveData.postValue(it) // Compile error here
} ?: run {
errorLiveData.postValue(response.message())
}
}
false -> errorLiveData.postValue(response.message())
}
} catch (e: Exception) {
noConnectionLiveData.postValue(true)
}
}
}
Unfortunately I'm missing something and the IDE is giving me a Type mismatch error trying to post the LiveData value:
Type mismatch: Required : Nothing! Found: Any.
I'm quite confused, do you have any suggestion to make about MutableLiveData and Generics in kotlin?
The response.body() type and the MutableLiveData type must match. The function signature should be something like this:
fun <T> makeCall(function: Deferred<Response<T>>, successLiveData: MutableLiveData<T>)
I've got this function getting documents from Cloud Firestore:
fun getBasicItems(callback: (MutableList<FireStoreBasicItem>) -> Unit) {
fireStore.collection("BasicItems")
.get()
.addOnCompleteListener { task ->
if (task.isSuccessful) {
val basicItems = mutableListOf<FireStoreBasicItem>()
for (document in task.result!!) {
val fireStoreBasicItem = document.toObject(FireStoreBasicItem::class.java)
basicItems.add(fireStoreBasicItem)
callback(basicItems)
}
}
}
}
In my ViewModel I want to transform this to an Observable an then to a ViewState:
private fun loadDataTransformer(): ObservableTransformer<ItemEvent.LoadDataEvent, ItemsViewState> {
return ObservableTransformer { event ->
event.map {
itemRepository.getBasicItems(){myBasicItemList -> Observable.just(myBasicItemList)}
}
}
I tried it also with Observable.fromCallable. What am I doing wrong?
EDIT: My Solution
private fun loadDataTransformer(): ObservableTransformer<ItemEvent.LoadDataEvent, ItemsViewState> {
return ObservableTransformer { event ->
event.flatMap {
Single.create<MutableList<FireStoreBasicItem>> {
itemRepository.getBasicItems { myBasicItemList ->
it.onSuccess(myBasicItemList)
}
}.toObservable()
.map {
ItemsViewState.ItemDataState(it)
}
}
}
}
I would like to suggest you to use Single instead of Observable if you are expecting only one list of items. Then you can use Single.create:
private fun loadDataTransformer(): Single<ItemsViewState> =
Single.create { emitter ->
itemRepository.getBasicItems() { myBasicItemList ->
val viewState = // do some transformations
emitter.onSuccess(viewState)
}
}
Is there anything similar in Kotlin that provides same ability as the Swift keyword 'defer' ?
What the defer key word does is, it ensure that the code inside a defer block get executed before returning from a function.
Below is an example imagining that defer keyword existed in Kotlin.
class MyClass {
var timeStamp = 0L
fun isEdible(fruit: Fruit): Boolean {
defer {
timeStamp = System.currentTimeMillis()
}
if (fruit.isExpired) {
return false
}
if (fruit.isRipe) {
return true
}
return false
}
}
In the case above, regardless of at what point the function returns, the block inside defer will get executed and timestamp's value will get updated, just before the function ends.
I know Java there is the finally {} keyword used along with try{} catch{}, but it's is not exactly what defer offers.
There's no such keyword in Kotlin, but you can make a construct yourself that will work quite similarly. Something like this (note that this does not handle exceptions in the deferred blocks):
class Deferrable {
private val actions: MutableList<() -> Unit> = mutableListOf()
fun defer(f: () -> Unit) {
actions.add(f)
}
fun execute() {
actions.forEach { it() }
}
}
fun <T> defer(f: (Deferrable) -> T): T {
val deferrable = Deferrable()
try {
return f(deferrable)
} finally {
deferrable.execute()
}
}
Then you can use it like this:
class MyClass {
var timeStamp = 0L
fun isEdible(fruit: Fruit): Boolean = defer { d ->
d.defer {
timeStamp = System.currentTimeMillis()
}
if (fruit.isExpired) {
return false
}
if (fruit.isRipe) {
return true
}
return false
}
}
The closest equivalent is try/finally. catch is not necessary if there's no exceptions thrown.
try {
println("do something")
// ... the rest of your method body here
}
finally {
println("Don't forget about me!");
}
In Swift, defer is usually used to ensure you don't forget to clean up some kind of resource or another (file handle, database connection, shared memory map, etc.). For this purpose, Kotlin use with, which takes a closure, to which the resource is passed as an argument. The resource is valid for the lifetime of the closure, and is automatically closed at the end.
FileWriter("test.txt")
.use { it.write("something") }
// File is closed by now
Solution with exception handling:
class DeferContext {
private val list = mutableListOf<() -> Unit>()
fun defer(payload: () -> Unit) {
list += payload
}
/** lombok `#Cleanup` analog */
fun AutoCloseable.deferClose() = apply {
defer { close() }
}
fun executeDeferred(blockError: Throwable?) {
var error: Throwable? = blockError
for (element in list.reversed()) {
try {
element()
} catch (e: Throwable) {
if (error == null) {
error = e
} else {
error.addSuppressed(e)
}
}
}
error?.let { throw it }
}
}
inline fun <T> deferBlock(payload: DeferContext.() -> T): T {
val context = DeferContext()
var error: Throwable? = null
var result: T? = null
try {
result = context.payload()
} catch (e: Throwable) {
error = e
} finally {
context.executeDeferred(error)
}
return result as T
}
IMHO, main point of defer functionality is execution of deferred actions regardless of previously thrown exceptions.
usage:
deferBlock {
defer { println("block exited") }
val stream = FileInputStream("/tmp/a").deferClose()
}
I came across the same question today.
While I think the answer provided by marstran is good, I decided to refactor it a little bit.
fun <T> deferred(f: ((() -> Unit) -> Unit) -> T): T {
val actions: MutableList<() -> Unit> = mutableListOf()
try {
return f(actions::add)
} finally {
actions.asReversed().forEach { it() }
}
}
I got rid of the Deferrable class by using the list directly in the deffered function. This also solves the fact that the whole Deferrable object was passed to the calling code needing to call it.defer/d.defer. In this version the add method of the mutable list is directly passed into the lambda allowing to have a code that is closer to its go/swift version.
To address the suggestion given by mvndaai to use Stack I decided to call .asReversed() on the list. Maybe there is a LI-FO type in kotlin that is also available in non JVM variants, but if not I think this is a good solution.
the given sample would look like:
class MyClass {
var timeStamp = 0L
fun isEdible(fruit: Fruit): Boolean = deferred { defer ->
defer {
timeStamp = System.currentTimeMillis()
}
if (fruit.isExpired) {
return false
}
if (fruit.isRipe) {
return true
}
return false
}
}
If the class is Closeable you can use use block:
class MyClass : Closeable {
var timeStamp = 0L
override fun close() {
timeStamp = System.currentTimeMillis()
}
fun test(): Boolean {
this.use {
if (fruit.isExpired) {
return false
}
if (fruit.isRipe) {
return true
}
return false
}
}
}