Remove deprecated MarkerView api

[tobrun]

Closes #13169

[LukasPaczos]

Looks great! There are some leftover drawables that were used with MarkerViews, or even MyLocationView, maybe we can clean those up as a part of this PR as well?

[LukasPaczos]

We can also remove this layout.

[LukasPaczos]

We can also remove this layout.

[LukasPaczos]

Actually, it might be beneficial to keep transform.invalidateCameraPosition() call here to always have a non-null camera position once the control is handled to the user. I believe we refactored Transform in the past to always, even internally, use Transform#getCameraPostion instead of the cameraPosition field and avoid NPEs (unless core returns a null camera position), but just to future proof, we can invalidate that position here as well.

[tobrun]

The use-cases are handled correctly by transform.

  @UiThread
  public final CameraPosition getCameraPosition() {
    if (cameraPosition == null) {
      cameraPosition = invalidateCameraPosition();
    }
    return cameraPosition;
  }

but just to future proof, we can invalidate that position here as well.

hardening and thus future proofing is handled in Transform.java, if there is no concrete indication to keep this use-case around. I would prefer not doing any obsolete work.

[tobrun]

alternatively, I'm ok with keeping that onPreMapReady hook but let's remove the hardened check in Transform#getCameraPosition() instead

[LukasPaczos]

I'm referring to a scenario when we would use Transform#cameraPosition field instead of the Transform#getCameraPosition method to perform operations on the camera object inside the Transform class. All of those usages have been refactored to use Transform#getCameraPosition, but I'm worried they might pop up again.

However, on the second thought, I'm not really sure whether we should perform obsolete calls to harden that, tests should be the way to go. Let's remove the call now, as it's not needed, and try to expnad our test suite to cover for that.

[tobrun]

I think you concerns are valid. I did a find usages on the Transfrom#cameraPosition: 3 usages were found accessing this field directly and all where handling the null case. The more I think about it, the more I would prefer rewriting this part of the SDK..

[LukasPaczos]

Not sure if this should be the way to go. Since the initial place where we fetch the camera position from the core is on frame fully rendered, the cameraPosition is still null when we deliver the onMapReady, as the frame is typically not fully render yet. I can reproduce this by calling mapboxMap#getCameraPosition inside of the onMapReady.

We should either keep this guard, or invalidate the camera before delivering the onMapReady.

[LukasPaczos]

Capturing from the chat that this is only a test setup, to let the CI catch possible regressions.

[tobrun]

This PR needs to be retargeted at master

[tobrun]

PR has been retargeted at master and is ready for review.

[LukasPaczos]

🚀