How to merge 3 or more LiveData sources of different generics?

Question

I am trying to generically merge 3 or more LiveData as sources where each LiveData is of a different generic type than the others. For this my intention is to use a vararg which each LiveData receives with its respective type.

I have already solved this but with a "hardcode" for 3 data sources with a solution very similar to that of this answer but this is not scalable.

The intention of use is as follows (does not compile):

    fun getSourceR(): MutableLiveData<Object_R>() {}
    fun getSourceS(): MutableLiveData<Object_S>() {}
    fun getSourceU(): MutableLiveData<Object_U>() {}

    MergeLiveDataTest(getSourceR(), getSourceS(), getSourceU()) {
        sources ->
            Log.d("MergeLiveData", sources[0].value.methodOf_R)
            Log.d("MergeLiveData", sources[1].value.methodOf_S)
            Log.d("MergeLiveData", sources[2].value.methodOf_U)
    }

Problem: I can't access the methods of the source type object.

The implementation I have so far is as follows:

What I'm trying to do is make each iteration of the loop X to be R, S, or U.

class MergeLiveDataTest<T, X>(
        vararg sources: LiveData<X>,
        private val onChanged: (newData: MutableList<LiveData<X>>) -> T
) : MediatorLiveData<T>() {
    private lateinit var _sources: MutableList<LiveData<X>>
    init {
        for (i in sources.indices) {
            super.addSource(sources[i]) {
                _sources.add(i, sources[i])
                value = onChanged(_sources)
            }
        }
    }
}

Any suggestion? Thank you

Source

You can't create a data structure with an arbitrary number of different generic types. — Tenfour04, Apr 27 '20 at 19:53
@Tenfour04 but you can create data structures of given numbers of different generic types. — EpicPandaForce, Apr 27 '20 at 21:00

EpicPandaForce · Answer 1 · 2021-03-22T16:45:05.497

I have already solved this but with a "hardcode" for 3 data sources with a solution very similar to that of this answer but this is not scalable.

Sure is, haven't you seen the EDIT? ;)

You can use the library that I wrote for this specific purpose https://github.com/Zhuinden/livedata-combinetuple-kt which does the same thing.

Now link-only answers aren't a good thing, so I'll explain what it does.

First I have tuples from 4-16 arity (because Kotlin already has Pair and Triple) in tuples-kt (which I wrote so that you don't need to write so many tuples yourself)

They look like this:

data class Tuple4<A, B, C, D>(
    val first: A,
    val second: B,
    val third: C,
    val fourth: D
) : Serializable {
    override fun toString(): String {
        return "Tuple4[$first, $second, $third, $fourth]"
    }
}

data class Tuple5<A, B, C, D, E>(
    val first: A,
    val second: B,
    val third: C,
    val fourth: D,
    val fifth: E
) : Serializable {
    override fun toString(): String {
        return "Tuple5[$first, $second, $third, $fourth, $fifth]"
    }
}

Then in LiveData-CombineTuple-KT, there is code like this:

fun <T1, T2, T3, T4> combineTuple(f1: LiveData<T1>, f2: LiveData<T2>, f3: LiveData<T3>, f4: LiveData<T4>): LiveData<Tuple4<T1?, T2?, T3?, T4?>> = MediatorLiveData<Tuple4<T1?, T2?, T3?, T4?>>().also { mediator ->
    mediator.value = Tuple4(f1.value, f2.value, f3.value, f4.value)

    mediator.addSource(f1) { t1: T1? ->
        val (_, t2, t3, t4) = mediator.value!!
        mediator.value = Tuple4(t1, t2, t3, t4)
    }

    mediator.addSource(f2) { t2: T2? ->
        val (t1, _, t3, t4) = mediator.value!!
        mediator.value = Tuple4(t1, t2, t3, t4)
    }

    mediator.addSource(f3) { t3: T3? ->
        val (t1, t2, _, t4) = mediator.value!!
        mediator.value = Tuple4(t1, t2, t3, t4)
    }

    mediator.addSource(f4) { t4: T4? ->
        val (t1, t2, t3, _) = mediator.value!!
        mediator.value = Tuple4(t1, t2, t3, t4)
    }
}

fun <T1, T2, T3, T4, T5> combineTuple(f1: LiveData<T1>, f2: LiveData<T2>, f3: LiveData<T3>, f4: LiveData<T4>, f5: LiveData<T5>): LiveData<Tuple5<T1?, T2?, T3?, T4?, T5?>> = MediatorLiveData<Tuple5<T1?, T2?, T3?, T4?, T5?>>().also { mediator ->
    mediator.value = Tuple5(f1.value, f2.value, f3.value, f4.value, f5.value)

    mediator.addSource(f1) { t1: T1? ->
        val (_, t2, t3, t4, t5) = mediator.value!!
        mediator.value = Tuple5(t1, t2, t3, t4, t5)
    }

    mediator.addSource(f2) { t2: T2? ->
        val (t1, _, t3, t4, t5) = mediator.value!!
        mediator.value = Tuple5(t1, t2, t3, t4, t5)
    }

    mediator.addSource(f3) { t3: T3? ->
        val (t1, t2, _, t4, t5) = mediator.value!!
        mediator.value = Tuple5(t1, t2, t3, t4, t5)
    }

    mediator.addSource(f4) { t4: T4? ->
        val (t1, t2, t3, _, t5) = mediator.value!!
        mediator.value = Tuple5(t1, t2, t3, t4, t5)
    }

    mediator.addSource(f5) { t5: T5? ->
        val (t1, t2, t3, t4, _) = mediator.value!!
        mediator.value = Tuple5(t1, t2, t3, t4, t5)
    }
}

Written all the way from 3 up to 16. Although it is a Kotlin library, so it assumes you have the Kotlin-stdlib in your project for Pair and Triple.

Anyways, I'd assume the ability to combine 16 LiveData into a tuple should be sufficient for most scenarios.

If Kotlin isn't allowed, then I'm sure the relevant Kotlin logic can be translated into Java, but it'd be a lot more verbose, so I never did.

In Kotlin, you can now easily do this:

val liveData = combineTuple(liveData1, liveData2, liveData3).map { (value1, value2, value3) ->
    // do something with nullable values
}

liveData.observe(this) { mappedValue ->
    // do something with mapped value
}

Your example would change as follows

fun getSourceR(): MutableLiveData<Object_R>() {}
fun getSourceS(): MutableLiveData<Object_S>() {}
fun getSourceU(): MutableLiveData<Object_U>() {}

combineTuple(getSourceR(), getSourceS(), getSourceU()).map { (r, s, u) ->
    sources ->
        Log.d("MergeLiveData", r?.methodOf_R)
        Log.d("MergeLiveData", s?.methodOf_S)
        Log.d("MergeLiveData", u?.methodOf_U)
}

I advise you should try it out.

This is great even though the code is verbose, it repeats itself and every time you need a larger number of fonts you have to modify the class. Anyway, I understand that this is a limitation of the language itself and also it is rare that you need to merge so many sources into one variable. So this response to my question. Thank you — Braian Coronel, May 02 '20 at 21:55

How to merge 3 or more LiveData sources of different generics?

The intention of use is as follows (does not compile):

The implementation I have so far is as follows:

1 Answers1