Hello maintainers. Thanks a lot for this library! It really simplifies work a lot, and I am replacing navigation-compose with it. However, I want to know if there is a way to replicate what startDestinationRoute.popUpto() in navigation compose does. That is, when navigating to 3 different screens sequentially, on tapping the back button on the 3rd screen, I should come back to the first screen. Thanks again!

Currently all branches are merged into main without requiring being up to date with main.

This could mean we merge an old branch that has a breaking change, and the main branch needs to be fixed.

Also, potentially if we introduce stricter static code analysis, potentially this older branch will increase our tech debt

We should consider using a merge queue to avoid this problem: https://docs.github.com/en/repositories/configuring-branches-and-merges-in-your-repository/configuring-pull-request-merges/managing-a-merge-queue

Otherwise we could also enforce that all PRs need to be up to date with main, but I believe that will be far more annoying.

When I quit my app using the back button on a phone running SDK 29, Leak Canary reports a memory leak.

If I use the fix reported here: https://issuetracker.google.com/issues/139738913, the leak is fixed.

override fun onBackPressed() {
        // Fix for Memory Leak: https://issuetracker.google.com/issues/139738913
        if (Build.VERSION.SDK_INT == 29 && !onBackPressedDispatcher.hasEnabledCallbacks()) {
            finishAfterTransition()
        } else {
            super.onBackPressed()
        }
    }

However, if I use Appyx NodeActivity with a RootNode, I get the memory leak again.

Description

Basic gesture API + Swipe implementation. Implemented as a Modifier which is applied to container composable.

If we agree on the API more gestures to come:

• draggable
• multi touching
• ..

Fixes https://github.com/bumble-tech/appyx/issues/97

https://user-images.githubusercontent.com/5773436/186649066-32f5b004-6ce2-4587-bca2-84000399fff7.mp4

Check list

• [ ] I have updated CHANGELOG.md if required.
• [ ] I have updated documentation if required.

The issue seems to be happening when doing consecutive operations (push or pop), best explained with video:

https://user-images.githubusercontent.com/20369236/177795492-c806aeaa-2046-467a-ad29-7ff85852a7c6.mp4

Full sample code:

class RootActivity : NodeActivity() {
  override fun onCreate(savedInstanceState: Bundle?) {
    super.onCreate(savedInstanceState)
    WindowCompat.setDecorFitsSystemWindows(window, false)

    setContent {
      MaterialTheme {
        NodeHost(
          integrationPoint = integrationPoint,
          factory = { buildContext ->
            RootNode(
              buildContext = buildContext,
              backstack = BackStack(
                initialElement = RootNode.Routing.Child("Root"),
                savedStateMap = buildContext.savedStateMap,
              )
            )
          },
        )
      }
    }
  }
}

class RootNode(
  buildContext: BuildContext,
  private val backstack: BackStack<Routing>,
) : ParentNode<RootNode.Routing>(
  routingSource = backstack,
  buildContext = buildContext,
) {
  sealed class Routing : Parcelable {
    @Parcelize
    data class Child(val id: String) : Routing()
  }

  @Composable
  override fun View(modifier: Modifier) {
    Surface(
      modifier = modifier.fillMaxSize(),
      color = Color.White
    ) {
      Children(
        routingSource = backstack,
        transitionHandler = rememberBackstackSlider(
          transitionSpec = { tween(2000, easing = LinearEasing) }
        ),
      )
    }
  }

  override fun resolve(routing: Routing, buildContext: BuildContext): Node {
    return ChildNode(
      name = when (routing) {
        is Routing.Child -> routing.id
      },
      buildContext = buildContext,
      pushNew = { backstack.push(Routing.Child(UUID.randomUUID().toString())) },
    )
  }
}

class ChildNode(
  private val name: String,
  private val pushNew: () -> Unit,
  buildContext: BuildContext
) : Node(buildContext = buildContext) {

  private val colors = listOf(
    manatee,
    sizzling_red,
    atomic_tangerine,
    silver_sand,
    md_pink_500,
    md_indigo_500,
    md_blue_500,
    md_light_blue_500,
    md_cyan_500,
    md_teal_500,
    md_light_green_500,
    md_lime_500,
    md_amber_500,
    md_grey_500,
    md_blue_grey_500
  )

  private val colorIndex =
    buildContext.savedStateMap?.get(KEY_COLOR_INDEX) as? Int ?: Random.nextInt(colors.size)
  private val color = colors[colorIndex]

  override fun onSaveInstanceState(state: MutableSavedStateMap) {
    super.onSaveInstanceState(state)
    state[KEY_COLOR_INDEX] = colorIndex
  }

  @Composable
  override fun View(modifier: Modifier) {
    Box(
      modifier = Modifier
        .fillMaxSize()
        .background(
          color = color,
          shape = RoundedCornerShape(6.dp)
        )
    ) {
      Column(
        modifier = Modifier.padding(24.dp),
        verticalArrangement = Arrangement.spacedBy(8.dp),
      ) {
        Text("Child ($name).")
        Row {
          // Local UI state should be saved too (both in backstack and onSaveInstanceState)
          var counter by rememberSaveable { mutableStateOf(0) }
          Text(text = "Counter $counter", modifier = Modifier.align(CenterVertically))
          Spacer(modifier = Modifier.width(16.dp))
          Button(onClick = { counter++ }, content = { Text("Increment") })
        }
        Row {
          Button(onClick = { navigateUp() }, content = { Text("Go up") })
        }
        Row {
          Button(onClick = pushNew, content = { Text("Push new") })
        }
      }
    }
  }

  companion object {
    private const val KEY_COLOR_INDEX = "ColorIndex"
  }
}

More info here

Similarly, can be reproduced following these steps:

1. Open sample sandbox
2. Click explicit navigation button
3. Click Push two
4. In Grandchild one click Navigate to child one
5. Observe lifecycle events in the logs. On destroy is not received

Hi, I am newbie to this and I went to your website and tried to find a page where I can get started like calling libraries inside Gradle:app and so forth but I can't find it. It would be nice if you guys could update better from the web page to here. Thanks

Description

Adds the ability to see dependency changes within PRs.

You can see the diff in action within this PR (https://github.com/LachlanMcKee/appyx/pull/2) - It needs to be raised in my fork as this needs to be the base branch.

Check list

• [x] I have updated CHANGELOG.md if required.
• [x] I have updated documentation if required.

Description

ActivityIntegrationPoint leaking Activity instances since ActivityIntegrationPoint holds a strong reference to the Activity. WeakHashMap is unable to clear the entry since the same activity instance is being used as key and also in the value object

┬───
│ GC Root: Thread object
│
├─ android.net.ConnectivityThread instance
│    Leaking: NO (PathClassLoader↓ is not leaking)
│    Thread name: 'ConnectivityThread'
│    ↓ Thread.contextClassLoader
├─ dalvik.system.PathClassLoader instance
│    Leaking: NO (ActivityIntegrationPoint↓ is not leaking and A ClassLoader is
│    never leaking)
│    ↓ ClassLoader.runtimeInternalObjects
├─ java.lang.Object[] array
│    Leaking: NO (ActivityIntegrationPoint↓ is not leaking)
│    ↓ Object[884]
├─ com.bumble.appyx.core.integrationpoint.ActivityIntegrationPoint class
│    Leaking: NO (a class is never leaking)
│    ↓ static ActivityIntegrationPoint.integrationPoints
│                                      ~~~~~~~~~~~~~~~~~
├─ java.util.WeakHashMap instance
│    Leaking: UNKNOWN
│    Retaining 169.6 kB in 3954 objects
│    ↓ WeakHashMap.table
│                  ~~~~~
├─ java.util.WeakHashMap$Entry[] array
│    Leaking: UNKNOWN
│    Retaining 169.5 kB in 3951 objects
│    ↓ WeakHashMap$Entry[0]
│                       ~~~
├─ java.util.WeakHashMap$Entry instance
│    Leaking: UNKNOWN
│    Retaining 83.2 kB in 1926 objects
│    referent instance of co.zuper.android.ui.host.HostActivity with mDestroyed
│    = true
│    ↓ WeakHashMap$Entry.value
│                        ~~~~~
├─ com.bumble.appyx.core.integrationpoint.ActivityIntegrationPoint instance
│    Leaking: UNKNOWN
│    Retaining 83.2 kB in 1925 objects
│    activity instance of co.zuper.android.ui.host.HostActivity with mDestroyed
│    = true
│    ↓ ActivityIntegrationPoint.activity
│                               ~~~~~~~~
╰→ co.zuper.android.ui.host.HostActivity instance
​     Leaking: YES (ObjectWatcher was watching this because co.zuper.android.ui.
​     host.HostActivity received Activity#onDestroy() callback and
​     Activity#mDestroyed is true)
​     Retaining 82.9 kB in 1916 objects
​     key = 005ec729-daaa-43df-a6d2-2f97825f9d7a
​     watchDurationMillis = 18279
​     retainedDurationMillis = 13277
​     mApplication instance of co.zuper.android.ZuperApp
​     mBase instance of androidx.appcompat.view.ContextThemeWrapper

METADATA

Build.VERSION.SDK_INT: 33
Build.MANUFACTURER: Google
LeakCanary version: 2.9.1
App process name: co.zuper.android.dev
Class count: 30112
Instance count: 243224
Primitive array count: 143091
Object array count: 38493
Thread count: 31
Heap total bytes: 31331321
Bitmap count: 28
Bitmap total bytes: 900521
Large bitmap count: 0
Large bitmap total bytes: 0
Stats: LruCache[maxSize=3000,hits=117858,misses=248001,hitRate=32%]
RandomAccess[bytes=12538085,reads=248001,travel=81765454152,range=37319569,size=
46458866]
Analysis duration: 4738 ms

Check list

• [x] I have updated CHANGELOG.md if required.
• [x] I have updated documentation if required.

Description

Ensure that unit tests and instrumentation tests are run as part of the PR

Fixes #34

Check list

• [x] I do not need to update CHANGELOG.md.
• [x] I do not need to update the documentation.

Description

This pull request adds a unit testing API. New helpers: InteractorTestHelper NodeTestHelper ParentNodeTestHelper

To make testing easier you can use: FeatureStub NodeViewStub instead of relying on mocks.

Check list

• [x] I have updated CHANGELOG.md if required.
• [x] I have updated documentation if required.

I've a question, given the exemple here: https://bumble-tech.github.io/appyx/navigation/implicit-navigation/#use-case-2

We listen to the UserRepository in the RootNode to then branch the right LoggedInNode or NotLoggedInNode. We make the listening lifecycle aware so we stop listening to UserRepository when RootNode is destroyed.

init {
        userRepository.isLoggedIn()
            .distinctUntilChanged()
            .onEach { isLoggedIn ->
                if (isLoggedIn) {
                        backstack.replace(NavTarget.LoggedInNode)
                    } else {
                        backstack.replace(NavTarget.NotLoggedInNode)
                    }
                }
            .launchIn(lifecycleScope)
    }

So, if we provoke a configuration change, the activity and all the nodes will be destroyed and recreated. The right node will be restored, but we'll also start listening to the UserRepository again, so we'll replace the restored Node by a new one.

Not sure how we can manage this case, any idea?

Description

This PR demonstrates how VM can be used and scoped to a Node using reflection API.

By design VM is stored in a ViewModelStore class and lives as long as activity lives.

The desired behaviour is the following:

1. VM is created when Node is created
2. VM is persistent when Node is moved in DESTROY state and Activity is changing configuration
3. VM is destroyed when Node is moved in DESTROY state and Activity IS NOT changing configuration

This behaviour can not be achieved using public API as it's not possible to clear only one VM from ViewModelStore. It has only one public method –ViewModelStore.clear()which clears allVMs.

I the meantime I'm exploring alternative solutions.

https://user-images.githubusercontent.com/5773436/207447633-65f79414-2313-4d7b-b453-464c4e51de30.mp4

Check list

• [ ] I have updated CHANGELOG.md if required.
• [ ] I have updated documentation if required.

It seems that not disposing NodeConnector.intake when a Node is destroyed is a bug.

Please let me know if this is intended. It could also potentially be a breaking change, so perhaps we need to throw an error if it happens

The FAQs callout an example with ViewModels was coming soon, but I couldn't find an associated tracking issue 😄

I imagine it might look something like HiltSupport from the Hilt support FAQ?