Develop Reference

In-memory caching on repository level for Kotlin Flows on Android

Suppose you have a list of users downloaded from a remote data source in your Android application, and for some reason you do not have a local DB. This list of users is then used throughout your entire application in multiple ViewModels to make other network requests, so you would surely like to have it cached for as long as the app lives and re-fetch it only on demand. This necessarily means you want to cache it inside the Data Layer, which is a Repository in my case, to then get it from your ViewModels.
It is easy to do in a state holder like a ViewModel - just make a StateFlow or whatever. But what if we want a Flow of List<User> (that is cached in RAM after every API request) available inside a repository to then collect from it from the UI Layer? What is the most testable, stable and right way of achieving this?
My initial idea led to this:
class UsersRepository #Inject constructor(
private val usersApi: UsersApi,
private val handler: ResponseHandler
) {
private val _usersFlow = MutableStateFlow<Resource<List<UserResponse>>>(Resource.Empty)
val usersFlow = _usersFlow.asStateFlow()
suspend fun fetchUserList() = withContext(Dispatchers.IO) {
_usersFlow.emit(Resource.Loading)
_usersFlow.emit(
handler {
usersApi.getUsers()
}
)
}
}
Where ResponseHandler is:
class ResponseHandler {
suspend operator fun <T> invoke(block: suspend () -> T) = try {
Resource.Success(block())
} catch (e: Exception) {
Log.e(javaClass.name, e.toString())
val errorCode = when (e) {
is HttpException -> e.code()
is SocketTimeoutException -> ErrorCodes.SocketTimeOut.code
is UnknownHostException -> ErrorCodes.UnknownHost.code
else -> Int.MAX_VALUE
}
Resource.Error(getErrorMessage(errorCode))
}
}
But while researching I found random guy on the internet telling that it is wrong:
Currently StateFlow is hot in nature so it’s not recommended to use in repository. For cold and reactive stream, you can use flow, channelFlow or callbackFlow in repository.
Is he right? If he is, how exactly do cold flows help in this situation, and how do we properly manage them?
If it helps, my UI Layer is written solely with Jetpack Compose

In the official "Guide to app architecture" from Google for Android:
About the source of true: ✅ The repository can contain an in-memory-cache.
The source of truth can be a data source—for example, the database—or even an in-memory cache that the repository might contain. Repositories combine different data sources and solve any potential conflicts between the data sources to update the single source of truth regularly or due to a user input event.
About the lifecycle: ✅ You can scope an instance of your repository to the Application class (but take care).
If a class contains in-memory data—for example, a cache—you might want
to reuse the same instance of that class for a specific period of
time. This is also referred to as the lifecycle of the class instance.
If the class's responsibility is crucial for the whole application,
you can scope an instance of that class to the Application class. This
makes it so the instance follows the application's lifecycle.
About the implementation: I recommend you to check the link directly.
class NewsRepository(
private val newsRemoteDataSource: NewsRemoteDataSource
) {
// Mutex to make writes to cached values thread-safe.
private val latestNewsMutex = Mutex()
// Cache of the latest news got from the network.
private var latestNews: List<ArticleHeadline> = emptyList()
suspend fun getLatestNews(refresh: Boolean = false): List<ArticleHeadline> {
if (refresh || latestNews.isEmpty()) {
val networkResult = newsRemoteDataSource.fetchLatestNews()
// Thread-safe write to latestNews
latestNewsMutex.withLock {
this.latestNews = networkResult
}
}
return latestNewsMutex.withLock { this.latestNews }
}
}
You should read the following page, I think it will answer a lot of your questions : https://developer.android.com/topic/architecture/data-layer

To make this work as a cache you will have to use this repository as a singleton. This effectively create a huge memory leak since you have no control over this memory. You cannot free it, you cannot bypass cache if you want (i mean you can, but it requires additional code outside the flow), you don't have any control over eviction. It's very dumb cache which acts like a memory leak. Not worth it.
Cold flow don't "help" in caching per se. They just give you control over each request that comes from the client. There you can check some outside memory cache if the entry is cached. If yes - is it correct or should be evicted? If it is evicted you can just a normal request. And all this is a single flow that gets disposed right after, so no memory leaks. The only part that have to be singleton is the cache. Although you can implement it as disk cache, it will be faster than network anyway

Why does by repository Flow not update my viewModels livedata?

So currently I have a Dao with a function that emits a Flow<>
#Query("SELECT * FROM ${Constants.Redacted}")
fun loadAllContacts(): Flow<List<Redacted>>
I am calling this from a repository like so
val loadAllContacts: Flow<List<Redacted>> = contactDao.loadAllContacts()
I am injecting the repository into the viewModel's constructor, and then at the top of my viewModel I have a val like so
val contacts: LiveData<List<Redacted>> = contactRepository.loadAllContacts.asLiveData()
Which is being observed in my Activity like so
viewModel.contacts.observe(this) { contacts ->
viewModel.onContactsChange(contacts)
}
My thinking is that the Flow is converted to a LiveData, and then I can observe this LiveData from my activity and kick off this function to actually update the viewModel upon the data being updated.
For now onContactsChange just looks like
fun onContactsChange(list: List<Redacted>) {
Timber.i("VIEW UPDATE")
}
The problem is that I only see this Timber log upon opening the activity, and never again. I verified that data IS going into my database, and I verified that an insert occurred successfully while the activity & viewModel are open. But I never see the log from onContactsChange again. When I close the activity, and reopen it, I do see my new data, so that is another reason I know my insert is working correctly.
I would like to add that I am using a single instance (singleton) of my repository, and I think I can verify this by the fact that I can see my data at all, at least when the view is first made.

Figured it out:
Note: If your app runs in a single process, you should follow the singleton design pattern when instantiating an AppDatabase object. Each RoomDatabase instance is fairly expensive, and you rarely need access to multiple instances within a single process.
If your app runs in multiple processes, include enableMultiInstanceInvalidation() in your database builder invocation. That way, when you have an instance of AppDatabase in each process, you can invalidate the shared database file in one process, and this invalidation automatically propagates to the instances of AppDatabase within other processes.

It's a little bit hard to follow your question, but I think I see the overall problem with your Flow object not updating the way you want it too.
Following this quick tutorial, it seems that first you should declare your Flow object inside your Repository the same way you're already doing
val loadAllContacts: Flow<List<Redacted>> = contactDao.loadAllContacts()
and have your VM 'subscribe' to it by using the collect coroutine which would then allow you to dump all this data into a MutableLiveData State
data class YourState(..)
val state = MutableLiveData<YourState>()
init {
contactRepository.loadAllContacts().collect {
if (it.isNotEmpty()) {
state.postValue(YourState(
...
)
}
}
}
that your Activity/Fragment could then observe for changes
viewModel.state.observe(.. { state ->
// DO SOMETHING
})
P.S. The tutorial also mentions that because of how Dao's work, you might be getting updates for even the slightest of changes, but that you can use the distinctUntilChanged() Flow extension function to get more specific results.

Why Realm doesn't find object which I inserted in transaction before?

I'm trying to implement a simple chat application on web sockets in Clean Architecture. I had to choose a db for caching all information, so I decided to use Realm, because I heard it was pretty good database for any kind of mobile applications. But when I actually faced the Realm, it turned out to be really painful experience for me to implement caching logic with it.
All problems come from applying transaction to database which then must be synced on all threads working with Realm. There seems to some kind of synchronization problem with my code. For example, I want to save my object to Realm and then query it out of.
Here I have two simple functions to save and to get chat:
fun getBackgroundLooper(): Looper {
val handlerThread = HandlerThread("backgroundThread")
if (!handlerThread.isAlive)
handlerThread.start()
return handlerThread.looper
}
fun saveChat(chat: Chat): Completable {
val realmChat = ChatMapper.domainToCache(chat)
return Completable.create { e ->
val realm = Realm.getDefaultInstance()
realm.executeTransactionAsync({
it.insertOrUpdate(realmChat)
}, {
realm.close()
e.onComplete()
}, {
realm.close()
e.onError(it)
})
// Subscribe on background looper thread
// to be able to execute async transaction
}.subscribeOn(AndroidSchedulers.from(getBackgroundLooper()))
}
fun getSingleChat(chatId: String): Single<Chat> {
return Single.defer {
val realm = Realm.getDefaultInstance()
realm.isAutoRefresh = true
val realmChat = realm.where(RealmChat::class.java)
.equalTo("id", chatId).findFirstAsync()
if (realmChat.isValid) {
realmChat.load()
val chat = ChatMapper.cacheToDomain(realmChat)
realm.close()
Single.just(chat)
}
realm.close()
Single.error<Chat>(ChatNotExistException())
// Subscribe on background looper thread
// to be able to execute auto refreshing
}.subscribeOn(AndroidSchedulers.from(getBackgroundLooper()))
}
So, when I try to run simple code like this
remote.getChat().flatMap {
cache.saveChat(it) //save chat to realm
.andThen(cache.getSingleChat(it.id)) //then query it by id
}
I always get no matter of what ChatNotExistException, but if I try to run query again in another attempt or after restarting the application, then the chat object gets found
I also tried many different approaches to execute this code:
I tried to use realm.refresh() in getSingleChat or not use it at all.
I tried to query chat synchronously with findFirst() and findAll() instead of findFirstAsync().
I tried querying chat on current thread without .subscribeOn().
I tried to use realm.executeTransaction() instead of async transactions.
I tried to add thread sleep between saving and querying, so that transaction may take some time to get applied and I need to wait before attempting to query the chat
I'm begging anybody to explain me what am I doing wrong and how to make this code working. I can't change the architecture of my application and use Realm objects as my view models, I need to find solution in these conditions.

But when I actually faced the Realm, it turned out to be really painful experience for me to implement caching logic with it.
Reading the docs regarding best practices help. For example, the default idea is that you define a RealmResults using an async query on the UI thread, add a change listener to it, and observe the latest emission of the database.
There is no "caching" involved in that beyond saving to the database and observing the database. Any additional complexity is added by you and is completely optional.
All problems come from applying transaction to database which then must be synced on all threads working with Realm.
All looper threads automatically make the Realm auto-refresh, therefore if addChangeListener is used as intended in the docs, then there is no need for trickery, Realm will manage the synchronization between threads.
I want to save my object to Realm and then query it out of.
realm.executeTransactionAsync({
No reason to use executeTransactionAsync when you are already on a background thread.
try(Realm realm = Realm.getDefaultInstance()) {
realm.executeTransaction((r) -> {
// do write here
});
}
realm.where(RealmChat::class.java)
If you do import io.realm.kotlin.where, then you can do realm.where<RealmChat>().
.findFirstAsync()
No reason to use findFirstAsync() instead of findFirst() when you are already on a background thread. Also no reason to use load() when you're on a background thread, because you should be using findFirst() in the first place anyway.
You are also most likely missing a return#defer Single.just(chat) to actually return the chat if it's found. That is most likely what your original problem is.
With the handler thread things you're doing though, you might want to consider taking a look at this project called "Monarchy", as it intends to set up the ability to run queries on a background looper thread while still observing the results. It is labelled stagnant but the ideas are sound.

RealmError: Realm Out of memory size

I am using Realm 3.0.0 as the DB of my Android app. It's like a questionnaire application, in which the user navigates inside the app a lot. When I use the app (go back and forth) continuously, I get the following error:
Fatal Exception: io.realm.exceptions.RealmError: Unrecoverable error. mmap() failed: Out of memory size: 1073741824 offset: 0 in /Users/cm/Realm/realm-java/realm/realm-library/src/main/cpp/io_realm_internal_SharedRealm.cpp line 109
at io.realm.internal.SharedRealm.nativeGetSharedRealm(SharedRealm.java)
at io.realm.internal.SharedRealm.(SharedRealm.java:187)
at io.realm.internal.SharedRealm.getInstance(SharedRealm.java:229)
at io.realm.internal.SharedRealm.getInstance(SharedRealm.java:204)
at io.realm.RealmCache.createRealmOrGetFromCache(RealmCache.java:124)
at io.realm.Realm.getDefaultInstance(Realm.java:210)
Now I know the main cause of this is not closing Realm instances. But I've already checked for that multiple times. And I am positive that I close every instance I open.
The app has many activities and fragments that all get a Realm instance on their onCreate and close it on their onDestroy. There are also other background network jobs that run to upload data that get Realm instances. These jobs close their Realm instances when they've finished running or when they cancel.
All of the above get their Realm instance thru injection via Dagger 2:
#Provides
#Nullable
static Realm realm(#Nullable RealmConfiguration configuration) {
if (configuration != null) {
Realm.setDefaultConfiguration(configuration);
return Realm.getDefaultInstance();
}
return null;
}
Configuration is also provided in the same Dagger Module.
To be more specific, a Questionnaire consists of many Question Fragments displayed in a ViewPager. Each Fragment gets injected with a realm. Many interactions in a given Question Fragment write data to the DB (some async, some blocking). These Fragments also query the database on onResume to get their updated Data. Some of this data is also copied out of Realm via realm.copyFromRealm(). Now at any given time of these happening, an upload job is most likely running and reading data from the DB and uploading it to a server. When an upload job finishes, it then writes to the DB.
I think I can have up to 7-12 fragment/activities holding a realm reference on the UI thread at a given moment. And 0-6 other references on 0-3 other threads (Background Jobs).
Moreover, I compact my realm DB via Realm.compactRealm(realmConfiguration) on every app launch (perhaps as a separate problem, this doesn't seem to do it's job consistently).
Above I've tried to describe my Realm usage descriptively without going into details. Now my problem is, when a user excessively uses the app (going back and forth between activities/fragments (realm injection + DB read query), uploading data (realm injection + DB read&write query)), I get the above posted Out of Memory Error.
I am also using Leak Canary, and it hasn't detected any leaks. (Not sure if it can anyway)
Am I using Realm in a way it's not supposed to be used? Should I close Realm instances onPause instead of onDestroy? Should I have only one realm instance in an activity and have all it's fragmetns (up to 5 in my case) use this instance? What kind of changes can I make in my app, and perhaps my app architecture to solve this problem?
I appreciate any help in trying to solve this problem.
EDIT: I'm sharing the realm open-close logic in my background threads.
All my jobs share the same realm usage, which is the following:
Realm is injected lazily via:
#Inject protected transient Lazy<Realm> lazyRealm;
The realm object reference is held at the private transient Realm realm; field. I am using Android Priority Job Queue. When the job is added:
#Override
public void onAdded() {
realm = lazyRealm.get();
realm.executeTransaction(realm1 -> {
//write some stuff into realm
});
realm.close();
}
And when the job is run realm is retreived once, and every possible ending of this method has a call to realm.close()
#Override public void onRun() throws Throwable {
synchronized (syncAdapterLock) {
realm = lazyRealm.get();
Answer answer = realm.where(Answer.class).equalTo(AnswerQuery.ID, answerId).findFirst();
if (answer == null) {
realm.close();
throw new RealmException("File not found");
}
final File photoFile = new File(answer.getFilePath());
final Response response = answerService.uploadPhotoAnswer(answerId, RequestBody.create(MediaType.parse("multipart/form-data"), photoFile)).execute();
if (!response.isSuccessful()) {
realm.close();
throw new HttpError(statusCode);
}
realm.executeTransaction(realm1 -> {
answer.setSyncStatus(SyncStatus.SYNCED.getCode());
});
}
realm.close();
}
}
As you can see, these background threads do close their realm instances properly as far as I'm concerned.

While it was true that all my background tasks did call realm.close(), one of them called it too late in it's lifecycle. That was my GPSService, which is a background service. The problem was that GPS service is initialized at the launch of the App as an Android Service, which is rarely destroyed. I was injecting a realm instance onCreate and closing it onDestroy. After the comments of #EpicPandaForce and reading his articles about using realm properly. I realized that this was the cause of the leak. A non-looper thread was keeping an open realm reference for an extremely long time, thus, the mmap was bloating every time a write transaction occures. Now that I moved the realm get/close to happen every time the service runs, my problem is fixed.
My take away is that one needs to treat background thread realm access very delicately. Thank you both for your quick responses and help!

RxJava and Cached Data

I'm still fairly new to RxJava and I'm using it in an Android application. I've read a metric ton on the subject but still feel like I'm missing something.
I have the following scenario:
I have data stored in the system which is accessed via various service connections (AIDL) and I need to retrieve data from this system (1-n number of async calls can happen). Rx has helped me a ton in simplifying this code. However, this entire process tends to take a few seconds (upwards of 5 seconds+) therefore I need to cache this data to speed up the native app.
The requirements at this point are:
Initial subscription, the cache will be empty, therefore we have to wait the required time to load. No big deal. After that the data should be cached.
Subsequent loads should pull the data from cache, but then the data should be reloaded and the disk cache should be behind the scenes.
The Problem: I have two Observables - A and B. A contains the nested Observables that pull data from the local services (tons going on here). B is much simpler. B simply contains the code to pull the data from disk cache.
Need to solve:
a) Return a cached item (if cached) and continue to re-load the disk cache.
b) Cache is empty, load the data from system, cache it and return it. Subsequent calls go back to "a".
I've had a few folks recommend a few operations such as flatmap, merge and even subjects but for some reason I'm having trouble connecting the dots.
How can I do this?

Here are a couple options on how to do this. I'll try to explain them as best I can as I go along. This is napkin-code, and I'm using Java8-style lambda syntax because I'm lazy and it's prettier. :)
A subject, like AsyncSubject, would be perfect if you could keep these as instance states in memory, although it sounds like you need to store these to disk. However, I think this approach is worth mentioning just in case you are able to. Also, it's just a nifty technique to know.
AsyncSubject is an Observable that only emits the LAST value published to it (A Subject is both an Observer and an Observable), and will only start emitting after onCompleted has been called. Thus, anything that subscribes after that complete will receive the next value.
In this case, you could have (in an application class or other singleton instance at the app level):
public class MyApplication extends Application {
private final AsyncSubject<Foo> foo = AsyncSubject.create();
/** Asynchronously gets foo and stores it in the subject. */
public void fetchFooAsync() {
// Gets the observable that does all the heavy lifting.
// It should emit one item and then complete.
FooHelper.getTheFooObservable().subscribe(foo);
}
/** Provides the foo for any consumers who need a foo. */
public Observable<Foo> getFoo() {
return foo;
}
}
Deferring the Observable. Observable.defer lets you wait to create an Observable until it is subscribed to. You can use this to allow the disk cache fetch to run in the background, and then return the cached version or, if not in cache, make the real deal.
This version assumes that your getter code, both cache fetch and non- catch creation, are blocking calls, not observables, and the defer does work in the background. For example:
public Observable<Foo> getFoo() {
Observable.defer(() -> {
if (FooHelper.isFooCached()) {
return Observable.just(FooHelper.getFooFromCacheBlocking());
}
return Observable.just(FooHelper.createNewFooBlocking());
}).subscribeOn(Schedulers.io());
}
Use concatWith and take. Here we assume our method to get the Foo from the disk cache either emits a single item and completes or else just completes without emitting, if empty.
public Observable<Foo> getFoo() {
return FooHelper.getCachedFooObservable()
.concatWith(FooHelper.getRealFooObservable())
.take(1);
}
That method should only attempt to fetch the real deal if the cached observable finished empty.
Use amb or ambWith. This is probably one the craziest solutions, but fun to point out. amb basically takes a couple (or more with the overloads) observables and waits until one of them emits an item, then it completely discards the other observable and just takes the one that won the race. The only way this would be useful is if it's possible for the computation step of creating a new Foo to be faster than fetching it from disk. In that case, you could do something like this:
public Observable<Foo> getFoo() {
return Observable.amb(
FooHelper.getCachedFooObservable(),
FooHelper.getRealFooObservable());
}
I kinda prefer Option 3. As far as actually caching it, you could have something like this at one of the entry points (preferably before we're gonna need the Foo, since as you said this is a long-running operation) Later consumers should get the cached version as long as it has finished writing. Using an AsyncSubject here may help as well, to make sure we don't trigger the work multiple times while waiting for it to be written. The consumers would only get the completed result, but again, that only works if it can be reasonably kept around in memory.
if (!FooHelper.isFooCached()) {
getFoo()
.subscribeOn(Schedulers.io())
.subscribe((foo) -> FooHelper.cacheTheFoo(foo));
}
Note that, you should either keep around a single thread scheduler meant for disk writing (and reading) and use .observeOn(foo) after .subscribeOn(...), or otherwise synchronize access to the disk cache to prevent concurrency issues.

I’ve recently published a library on Github for Android and Java, called RxCache, which meets your needs about caching data using observables.
RxCache implements two caching layers -memory and disk, and it counts with several annotations in order to configure the behaviour of every provider.
It is highly recommended to use with Retrofit for data retrieved from http calls. Using lambda expression, you can formulate expression as follows:
rxCache.getUser(retrofit.getUser(id), () -> true).flatmap(user -> user);
I hope you will find it interesting :)

Take a look at the project below. This is my personal take on things and I have used this pattern in a number of apps.
https://github.com/zsiegel/rxandroid-architecture-sample
Take a look at the PersistenceService. Rather than hitting the database (or MockService in the example project) you could simply have a local list of users that are updated with the save() method and just return that in the get().
Let me know if you have any questions.