Which is a better way to combine multiple LiveData(s): using MediatorLiveData or switchMap?

Which approach is more recommended to combine multiple LiveData(s): using MediatorLiveData or switchMap?

// MediatorLiveData approach
fun <A, B, C> combine(
    liveData1: LiveData<A>,
    liveData2: LiveData<B>,
    onChanged: (A?, B?) -> C
): MediatorLiveData<C> {
    return MediatorLiveData<C>().apply {
        addSource(liveData1) {
            value = onChanged(liveData1.value, liveData2.value)
        }
        addSource(liveData2) {
            value = onChanged(liveData1.value, liveData2.value)
        }
    }
}

// switchMap approach
fun <A, B, C> combine(
    liveData1: LiveData<A>,
    liveData2: LiveData<B>,
    onChanged: (A, B) -> C
): LiveData<C> {
    return liveData1.switchMap { a ->
        liveData2.map { b ->
            onChanged(a, b)
        }
    }
}

The MediatorLiveData approach may be more common but the switchMap approach looks simpler. Both approaches seem to serve the same purpose in many cases.

You should be using MediatorLiveData directly.

All Transformations. are actually MediatorLiveDatas - switchMap is just a mediator that dynamically swaps source depending on return value of provided switchMapFunction.

Second approach just generates useless overhead - you allocate switchMap MediatorLiveData that keeps creating and swapping in new map MediatorLiveData objects whenever liveData1 changes.