jump back to world zero when crossing the antimeridian in flyTo and other camera operations. Fixes #2071

[andrewharvey]

Launch Checklist

• briefly describe the changes in this PR
  Some background in By default, snap back to "w=0" world after panning operation (i.e. "worldCopyJump") #2071. When flyTo decides to cross the anti-meridian from world 0, it will end up in either world -1 or 1. Because Marker DOM elements are only drawn in one place and only on world 0, they don't show up at the moment.

This PR aims to ensure that when flyTo is about to step across the anti-meridian into world -1 or 1, it teleports to ensure it lands back in world 0, and so Marker elements appear.

It also ensure's that other camera operations like jumpTo, easeTo don't leave the map in a non world 0 state.

• write tests for all new functionality

• document any changes to public APIs
  I feel this should be transparent to the public API, it's a private behaviour.

• post benchmark scores

• manually test the debug page

I wrote a new debug page to test this, but haven't committed it.

• the world sometimes flashes white when you cross the anti-meridian, I'm not sure how to fix this.

• only jump back to world 0 when renderWorldCopies is true

[andrewharvey]

@1ec5 Sorry for making a further change here after you approved it. But I decided to tackle the wider issue of #2071. I've added another commit which ensures other camera operations don't leave world 0 (not just flyTo).

[jfirebaugh]

Thanks for working on this!

[jfirebaugh]

Any particular reason to change this behavior? We'd want to at least update the API documentation above, which says that wrap always returns a new object.

[andrewharvey]

Since my changes meant that LngLat.wrap is called a lot more (every camera change due to an interaction), I wanted to ensure that creating the new LngLat object wouldn't impact performance.

Although on further testing, map interaction seems to have no degradation without this check. So I'm not sure whether to leave it in and update the API docs or just remove this change, and leave it as is.

Thoughts?

[jfirebaugh]

If there's no performance impact, let's leave it as is.

[andrewharvey]

Agreed, fixed, and squashed commits.

[jfirebaugh]

the world sometimes flashes white when you cross the anti-meridian

I didn't observe this in my testing. If you're still seeing it, can you open an issue with steps to replicate?

[andrewharvey]

I didn't observe this in my testing. If you're still seeing it, can you open an issue with steps to replicate?

I only noticed it occasionally, what happens is the map is there and after the jump it redraws part of the map (a vector tile) so the map flashes the background colour while this happens.

I agree that we should merge this and follow up with a new issue to address that.