IssueTracker

Re #3 - no, just use the code in comment#1:

val navHostFragment = supportFragmentManager.findFragmentById(R.id.my_nav_host_fragment) as NavHostFragment
val navController = navHostFragment.navController

Re #5 - navigation-runtime does not depend on fragment. You can certainly write an extension method to do the code in comment#4 if you wish as that code will work at any point.

Re #7 - please don't follow that bad advice. onPostCreate() runs after onStart(), which is most certainly later than you should be setting up your NavController. Just use the approach in comment#1 and comment#4:

NavHostFragment navHostFragment = (NavHostFragment) getSupportFragmentManager()
    .findFragmentById(R.id.my_nav_host_fragment);
NavController navController = navHostFragment.getNavController();

Maybe you can change implementation of extension function from navigation-ui-ktx? In place of

fun Activity.findNavController(@IdRes viewId: Int): NavController =
        Navigation.findNavController(this, viewId)

to

fun FragmentActivity.findNavController(@IdRes viewId: Int) =
        (supportFragmentManager.findFragmentById(viewId) as? NavHostFragment)?.navController
                ?: Navigation.findNavController(this, viewId)

I was wondering if it is actually a good idea to inflate NavHostFragment with a FragmentContainerView because if we do that what happen from my understanding is a NavHostFragment instance would be created with its view attached to that FragmentContainerView (named it FCV1).

The view of the NavHostFragment is a FragmentContainerView too (FCV2) according to the NavHostFragment source code:

    @Nullable
    @Override
    public View onCreateView(@NonNull LayoutInflater inflater, @Nullable ViewGroup container,
                             @Nullable Bundle savedInstanceState) {
        FragmentContainerView containerView = new FragmentContainerView(inflater.getContext());
        containerView.setId(getContainerId());
        return containerView;
    }

Given NavHostFragment is generally used to "host" other fragments which are destinations in its associated NavGraph, those fragments' views would be attached to FCV2, leaving FCV1 not having clear value, is it a waste?

On the other hand, if we use <fragment> tag to attach NavHostFragment to an Activity, FCV1 would not be created and FCV2 is directly attached to the activity's view.

I'm general fine with using <fragment> for NavHostFragment.

But the real problem here is Android Studio showing warning for this setup suggesting to replace <fragment> with FragmentContainerView.

In other words, Android Studio suggestion is breaking the code, leaving developer in dark. :-(

Yes, I mean more like, if my understand above is correct, then Android Studio actually gives an incorrect warning because using FragmentContainerView instead of <fragment> would not make the app better, and can say even worse considering the extra nested layout.

It's not only about performance. Using FragmentContainerView is breaking the app.

Check the title. When you use FragmentContainerView findNavController() is not working anymore!

Yeah sorry for confusion, I should have been clearer.

I mean I understand that it breaks the app if we address the warning. And what I wanted to say was even if it didn't break the app, I don't think it's a valid warning since using replacing <fragment> with FragmentContainerView` would create extra nested layout while there is no benefit gained.

Re #10 - yes, you should always use FragmentContainerView. There are absolutely other fixes around window insets and layout issues that occur when a fragment's root layout is directly within other layouts such as ConstraintLayout, besides the underlying issues with the <fragment> tag where fragments added via that tag go through lifecycle states entirely differently from the other Fragments added to the FragmentManager. The Lint check is there exactly because you absolutely should switch over to FragmentContainerView in all cases. It is not a no-op or negative even for a single static fragment.

FragmentContainerView uses a FragmentTransaction to add the initial fragment. If you were to use that approach manually, you'd note that findNavController() would not work in that case either. The whole point of FragmentContainerView is to provide a consistent path and lifecycle transition points for all fragments - as mentioned in comment #2 - the Activity's onCreate() is not where you should be accessing the Fragment's views, which is why getting the NavController from the NavHostFragment itself is the recommended solution in comment #4.

Re #15 - that can be all well and true, however in the state it's now it's pure madness.

Look at this stack overflow question https://stackoverflow.com/questions/50502269/illegalstateexception-link-does-not-have-a-navcontroller-set. What do you think will most developers facing this issue do? Top answers are to revert back to <fragment> and I guarantee you most developers won't think twice about it. At the and all the good intensions are useless.

So until there is a lint check for findNavController() in onCreate() telling developer to call it later (in onPostCreate) or at least documentation stating this fact, you can't expect people will actually switch to FragmentContainerView.

Another option would be apply suggestion from #5 and incorporate searching fragment manager in findNavController before throwing exception that NavController is not present.

Long story short, as it is right now is bad and you shouldn't be turning a blind eye to it.

Re #15 - thanks for the clarification. I wasn't aware of the issue with using <fragment> in xml. Is it documented somewhere or is there any related bug report?

While I generally agree with #16, I think the original issue was the default generated code in Android Studio uses <fragment> in XML and also having findNavController in onCreate(). If it could be updated by Android team to properly use FragmentContainerView (for example can use what's recommended in #4), a lot of confusion can be avoided. At the end of the day, Android Studio's example code should reflect the most correct and recommended way of doing things.

I'll stick to #4.

RG

working in onCreate()

val navHostFragment by lazy { supportFragmentManager.findFragmentById(R.id.navHostFragment) as NavHostFragment }
val navController get() = navHostFragment.navController
override fun onCreate(savedInstanceState: Bundle?) {
    navHostFragment.childFragmentManager...
}

I'm curious about the use cases that lead to this exception. If the user has defined the nav graph and set it as the nav graph of the FragmentContainerView, what other functionality do they require during onCreate()?

Ad #20 for example one may want to parametrize start destination depending on some condition.

For an activity, onPostCreate works. For example in the activity:

   override fun onPostCreate(savedInstanceState: Bundle?) {
        super.onPostCreate(savedInstanceState)
        // Needs to be in the post create
        // Refer issue https://issuetracker.google.com/issues/142847973
        _binding.navHostFragment.findNavController().addOnDestinationChangedListener { _, destination, _ ->
                // code goes here :)
            }
        }
    }

Just ran into this today and I am sad that this is marked as "infeasible" when a good response to this problem, at minimum, is to update the boilerplate code for new projects so that it doesn't generate code that automatically gets complaints from lint and then auto-fixes to broken code.