Hello everybody, not sure if that may be a MockK issue, but let’s start posting here and let’s see :...

# compose-android

[Davide Giuseppe Farella]

Hello everybody, not sure if that may be a MockK issue, but let’s start posting here and let’s see 🙂

I have a mapper that maps from LazyPagingItems to my model/state (I do that in the presenter, using Molecule, if you’re wondering)
In order to test that, I created this convenience mock (since LPI has internal constructor)

fun <T : Any> mockLazyPagingItems(isEmpty: Boolean, refresh: LoadState): LazyPagingItems<T> = mockk {
    every { loadState } returns mockk loadState@{
        every { [e-mail] } returns refresh
        every { [e-mail] } returns LoadState.Loading
        every { [e-mail] } returns LoadState.Loading
    }
    every { itemCount } returns if (isEmpty) 0 else 1
}

Updating to alpha19 I now have this exception (full stacktrace in thread)

java.lang.NullPointerException: Cannot invoke “androidx.compose.runtime.State.getValue()” because “$this$getValue$iv” is null

How would I fix that?

[Davide Giuseppe Farella]

full stack trace

full stack trace

[shikasd]

It seems like LazyPagingItems are reading some state that wasn't initialized because of mockk

[Davide Giuseppe Farella]

I would get a mock exception "no answer found blabla" 🤔
My guess is that that $iv is something from the IR, no? 🤔

[Davide Giuseppe Farella]

I didn't have time to dive yet.
By the way I think it would be nice having a test artifact that provides a way to create an instance of a LPI, but I'm not sure in which scenarios that would be useful besides mine, as it would be impossible to get a LPI outside of the UI without Molecule (as collectAsLazyPagingItems is the only way)

[Davide Giuseppe Farella]

I dig a bit more, and I assume the problem is that loadState is now a delegate.
I instantiated a LPI using reflection, but I’ve not been able to run my tests correctly, as I’ve not been able to update the snapshot list under test 😞

[Davide Giuseppe Farella]

I think I’ll create a surrogate of LPI and create a surrogate mapper in between, so I can test my mapper (m2) using the surrogate

LazyPagingItems (m1)-> LazyPagingItemsSurrogate (m2)-> PagingItemsState

[Ian Lake]

I have a mapper that maps from LazyPagingItems to my model/state (I do that in the presenter, using Molecule, if you're wondering)

This is the wrong thing to do in the first place. Map your Flow<PagingData> using the operators provided by Paging at that level

[Davide Giuseppe Farella]

I'm not trying to map the models, I want a declarative state for the UI, which can be:
• Empty
• Items (also loading or not)
• Loading
• Error (also loading or not)
While LPI provides a set of about 50 combinations or more 😁

[Ian Lake]

I'm a bit confused, since Loading is completely covered by Empty and Items? The number of items and load state are independent properties

[Ian Lake]

Sounds like you just need a Kotlin extension that takes CombinedLoadState and make it into the simplified view of the world you want?

[Ian Lake]

E.g., Loading, Loaded, or Error?

[Davide Giuseppe Farella]

Yes, they're independent, but in my UI logic one depends on the other (eg, when loading I'd show something different regarding whether I have items or not).

That was my purpose, but encapsulated in a (generic) mapper class with tests 😊

[Ian Lake]

And I'm still not entirely sure where mocking this comes into the picture? You can create a fake Flow<PagingData> easily enough and just use the actual collectAsLazyPagingItems()?

[Davide Giuseppe Farella]

I did solve with the surrogate tho, a bit verbose, but it does its job

[Davide Giuseppe Farella]

Using the collectAsLazyPagingItems didn't work as expected under JUnit test, at least with my setup

[Ian Lake]

E.g., flowOf(PagingData.from(...)).collectAsLazyPagingItems()

[Davide Giuseppe Farella]

The snapshot list was never updated in my case. I also tried using Dispatchers.setMain, but didn't help

[Ian Lake]

Seems like you should have led with "how do I get collectAsLazyPagingItems working in my test" rather than bringing up a problem a few steps down the wrong path - that way we'd avoid that XY Problem: https://en.wikipedia.org/wiki/XY_problem

[Davide Giuseppe Farella]

True 😁 the reason is because the test was working as expected, but the update to alpha19 changed that, so the history conditioned my question

[Davide Giuseppe Farella]

But I was also thinking now that using collectAsLPI would led to another problem: I'd need to test using the Flow coming from the library, which emission is subject to changes: e.g. the first emitted LPI has changed in some way on alpha17

[Ian Lake]

You'll really probably want to wait one more build, since it only then that PagingData.from instantly provide the static state to the first composition if you use it with StateFlow (e.g. MutableStateFlow(PagingData.from...)))

[Ian Lake]

Since that'll be the build that this change is released in

[Davide Giuseppe Farella]

Thank you for the info, I'm also super excited about the previews 🥳