Reputation: 108
This is the ViewModel where the Rx transactions are called
RealmHelperRepo is the implementation of HelperRepo interface
@PerActivity
public class RoleSelectionViewModel extends BaseViewModel<RoleSelectionMvvm.View> implements RoleSelectionMvvm.ViewModel {
private Disposable roleGroupSubscription;
@Inject
public RoleSelectionViewModel(@AppContext Context context, HelperRepo helperRepo, ApiOAuth2 ApiOAuth2) {
this.mContext = context;
this.mUserRepo = userRepo;
this.mHelperRepo = helperRepo;
ApiOAuth2.initialize();
this.mApiOAuth2 = ApiOAuth2;
this.mCurrentUser = mUserRepo.getByField("isLoggedIn", true, true);
if (mCurrentUser != null) {
this.mCurrentUserId = mCurrentUser.getId();
this.mHelper = mHelperRepo.getByField("user.id", mCurrentUserId, true);
}
Observable<Response<ResponseHelper>> postHelperObservable = mApiOAuth2.postHelperRX(new Helper());
Observable<Response<ResponseHelper>> getHelperObservable = mApiOAuth2.getHelperRX(mCurrentUserId);
roleGroupSubscription = postRoleGroupsObservable
.subscribeOn(Schedulers.io())
.observeOn(AndroidSchedulers.mainThread())
.flatMap((response) -> {
if (response.isSuccessful()) {
ResponseHelper responseHelper = response.body();
mHelper = responseHelper.getHelper();
return Observable.just(mHelper);
} else if (response.code() == 409) {
// handle POST conflict (i.e. helper already exists)
return getHelperObservable;
}
})
.subscribe((data) -> {
if (data instanceof Response<?>) {
// data came from getHelperObservable
Response response = (Response) data;
if (!response.isSuccessful()) {
ResponseHelper responseHelper = (ResponseHelper) response.body();
mHelper = responseHelper.getHelper();
else {
// data came from Observable.just(helper)
mApiOAuth2.getHelperRX(mCurrentUserId).subscribe(
responseHelperResponse -> {
if (responseHelperResponse.isSuccessful()) {
String helperID = responseHelperResponse.body().getHelper().getId();
Log.d("RealmCount", "save: " + Realm.getLocalInstanceCount(realmProvider.get().getConfiguration()));
mHelper.setId(helperID);
mHelper.setUser(mCurrentUser);
--------> // when mHelperRepo.save(mHelper) is called, it goes to RealmHelperRepo to save and
--------> // thus triggering mRealm.executeTransaction causing Realm threading
mHelperRepo.save(mHelper);
}
saveAndBegin();
},
Throwable::printStackTrace);
});
This is the RealmRepo class where realm calls are made.
@PerApplication
public class RealmHelperRepo implements HelperRepo {
private final Provider<Realm> mRealmProvider;
private Realm mRealm;
@Inject
public RealmHelperRepo(Provider<Realm> realmProvider) {
this.mRealmProvider = realmProvider;
this.mRealm = mRealmProvider.get();
}
@Override
public void save(Helper helper) {
if (mRealm != null) {
---------> // code runs into threading issue here when a realmThread executeTransaction is called
mRealm.executeTransaction(r -> r.copyToRealmOrUpdate(helper));
}
}
Is there something I'm missing here? Other Rx functions I should be using instead of flatmap? Are there other ways to save my observable data without running into threading issue? help!
Upvotes: 2
Views: 106
Reputation: 81539
Is there something I'm missing here?
A Realm
instance represents a reference-counted, thread-local instance. It's not a global thing, it's a "local instance" that is opened by getInstance()
and then closed by close()
.
So you can't just initialize a Realm instance as a singleton, because it won't be accessible from background threads.
You could for example provide a singleton Realm manager class that is able to open thread-local Realm instances.
/**
* The RealmManager allows creating a singleton Realm manager which can open thread-local instances.
*
* It also allows obtaining the open thread-local instance without incrementing the reference count.
*/
@PerApplication
public class RealmManager {
private final ThreadLocal<Realm> localRealms = new ThreadLocal<>();
@Inject
RealmManager() {
}
/**
* Opens a reference-counted local Realm instance.
*
* @return the open Realm instance
*/
public Realm openLocalInstance() {
checkDefaultConfiguration();
Realm realm = Realm.getDefaultInstance(); // <-- maybe configuration should be constructor parameter
if(localRealms.get() == null) {
localRealms.set(realm);
}
return realm;
}
/**
* Returns the local Realm instance without adding to the reference count.
*
* @return the local Realm instance
* @throws IllegalStateException when no Realm is open
*/
public Realm getLocalInstance() {
Realm realm = localRealms.get();
if(realm == null) {
throw new IllegalStateException(
"No open Realms were found on this thread.");
}
return realm;
}
/**
* Closes local Realm instance, decrementing the reference count.
*
* @throws IllegalStateException if there is no open Realm.
*/
public void closeLocalInstance() {
checkDefaultConfiguration();
Realm realm = localRealms.get();
if(realm == null) {
throw new IllegalStateException(
"Cannot close a Realm that is not open.");
}
realm.close();
// noinspection ConstantConditions
if(Realm.getLocalInstanceCount(Realm.getDefaultConfiguration()) <= 0) {
localRealms.set(null);
}
}
private void checkDefaultConfiguration() {
if(Realm.getDefaultConfiguration() == null) {
throw new IllegalStateException("No default configuration is set.");
}
}
}
You can use this like so
@PerApplication
public class RealmHelperRepo implements HelperRepo {
private final RealmManager realmManager;
@Inject
public RealmHelperRepo(RealmManager realmManager) {
this.realmManager = realmManager;
}
@Override
public void save(Helper helper) {
try(Realm realm = realmManager.openLocalInstance()) {
realm.executeTransaction(r -> r.copyToRealmOrUpdate(helper));
}
}
Technically it just hides the Realm.getDefaultInstance()
call and allows you to obtain your thread-local instance even without incrementing the internal RealmCache reference count, so not much real magic there.
Just open a Realm instance for a thread, and don't forget to close it when it is no longer needed.
Upvotes: 2