Is it possible to enforce non-nullability of LiveData values?

Question

Is there any way to enforce non-nullability of LiveData values? Default Observer implementation seems to have @Nullable annotation which forces an IDE to suggest that the value might be null and should be checked manually:

public interface Observer<T> {
    /**
     * Called when the data is changed.
     * @param t  The new data
     */
    void onChanged(@Nullable T t);
}

Solution 1

While there a few things you can do, it is your responsibility to make sure you don't pass null to the LiveData. In addition to that, every 'solution' is more a suppression of the warning, which can be dangerous (if you do get a null value, you might not handle it and Android Studio will not warn you).

Assert

You can add assert t != null;. The assert will not be executed on Android, but Android Studio understands it.

class PrintObserver implements Observer<Integer> {

    @Override
    public void onChanged(@Nullable Integer integer) {
        assert integer != null;
        Log.d("Example", integer.toString());
    }
}

Suppress the warning

Add an annotation to suppress the warning.

class PrintObserver implements Observer<Integer> {

    @Override
    @SuppressWarnings("ConstantConditions")
    public void onChanged(@Nullable Integer integer) {
        Log.d("Example", integer.toString());
    }
}

Remove the annotation

This also works in my installation of Android Studio, but it might not work for you, but you could try to just remove the @Nullable annotation from the implementation:

class PrintObserver implements Observer<Integer> {

    @Override
    public void onChanged(Integer integer) {
        Log.d("Example", integer.toString());
    }
}

Default methods

It's unlikely you can use this on Android, but purely from a Java perspective, you could define a new interface and add a null check in a default method:

interface NonNullObserver<V> extends Observer<V> {

    @Override
    default void onChanged(@Nullable V v) {
        Objects.requireNonNull(v);
        onNonNullChanged(v);
        // Alternatively, you could add an if check here.
    }

    void onNonNullChanged(@NonNull V value);
}

Solution 2

If you use Kotlin, you can create much nicer non-null observe function with extension. There is an article about it. https://medium.com/@henrytao/nonnull-livedata-with-kotlin-extension-26963ffd0333

Solution 3

It's possible to do it safely only if you are in control of the code which sets the data because you'll also have to wrap the LiveData class. This way the data setting methods will be protected with @NonNull and you can be sure that the data has already been checked before reaching the Observer.

Wrap the LiveData class:

public class NonNullMutableLiveData<T> extends MutableLiveData<T> implements NonNullLiveData<T> {

    private final @NonNull T initialValue;

    public NonNullMutableLiveData(@NonNull T initialValue) {
        this.initialValue = initialValue;
    }

    @Override
    public void postValue(@NonNull T value) {
        super.postValue(value);
    }

    @Override
    public void setValue(@NonNull T value) {
        super.setValue(value);
    }

    @NonNull
    @Override
    public T getValue() {
        //the only way value can be null is if the value hasn't been set yet.
        //for the other cases the set and post methods perform nullability checks.
        T value = super.getValue();
        return value != null ? value : initialValue;
    }

    //convenience method
    //call this method if T is a collection and you modify it's content
    public void notifyContentChanged() {
        postValue(getValue());
    }

    public void observe(@NonNull LifecycleOwner owner, @NonNull NonNullObserver<T> observer) {
        super.observe(owner, observer.getObserver());
    }
}

Create an interface for exposing as immutable:

public interface NonNullLiveData<T> {

    @NonNull T getValue();

    void observe(@NonNull LifecycleOwner owner, @NonNull NonNullObserver<T> observer);
}

Finally, wrap the Observer:

//not implementing Observer<T> to make sure this class isn't passed to
//any class other than NonNullMutableLiveData.
public abstract class NonNullObserver<T> {

    public Observer<T> getObserver() {
        return new ActualObserver();
    }

    public abstract void onValueChanged(@NonNull T t);

    private class ActualObserver implements Observer<T> {

        @Override
        public void onChanged(@Nullable T t) {
            //only called through NonNullMutableLiveData so nullability check has already been performed.
            //noinspection ConstantConditions
            onValueChanged(t);
        }
    }
}

Now you can create your data like this:

class DataSource {
    private NonNullMutableLiveData<Integer> data = new NonNullMutableLiveData<>(0);

    public NonNullLiveData<Integer> getData() {
        return data;
    }
}

And use it like this:

dataSource.getData().observe(this, new NonNullObserver<Integer>() {
            @Override
            public void onValueChanged(@NonNull Integer integer) {

            }
        });

Completely null safe.

Solution 4

A new option is available if you use Kotlin. You can replace LiveData with StateFlow. It is more suitable for Kotlin code and provides built-in null safety.

Instead of using:

class MyViewModel {
    val data: LiveData<String> = MutableLiveData(null) // the compiler will allow null here!
}

class MyFragment: Fragment() {
    model.data.observe(viewLifecycleOwner) {
        // ...
    }
}

You can use:

class MyViewModel {
    val data: StateFlow<String> = MutableStateFlow(null) // compilation error!
}

class MyFragment: Fragment() {
    lifecycleScope.launch {
        model.data.collect {
          // ...
        }
    }
}

StateFlow is part of coroutines and to use the lifecycleScope you need to add the lifecycle-extensions dependency:

implementation "androidx.lifecycle:lifecycle-extensions:$lifecycle_version"

Note that this API has been experimental before coroutines 1.4.0.

Here's some additional reading about replacing LiveData with StateFlow.

As Igor Bubelov pointed out, another advantage of this approach is that it's not Android specific so it can be used in shared code in multiplatform projects.

Solution 5

fun <T> LiveData<T>.observeNonNull(owner: LifecycleOwner, observer: (t: T) -> Unit) {
    this.observe(owner, Observer {
        it?.let(observer)
    })
}

View more solutions