Fly across the antimeridan if shorter

[1ec5]

If flying across the antimeridian yields a shorter flight path than flying across the Prime Meridian, cross the antimeridian. This PR also adds a full complement of unit tests based on the easeTo() unit tests and fixes a couple corner cases that differ between easeTo() and flyTo(): now, if you fly to the current location but with a different zoom level, bearing, or pitch, we ease to the new camera instead of bailing. We also avoid offsetting if the options don’t call for a new center point.

Fixes #1853. Ported from mapbox/mapbox-gl-native@97c143a for mapbox/mapbox-gl-native#3355.

/cc @mourner @lucaswoj

[jfirebaugh]

Can you add a test please?

[1ec5]

We don’t have any flyTo tests yet. I’ve been trying to copy the easeTo tests for use as flyTo tests, but I’m getting tripped up by what may be a bug in the existing implementation: the final camera values are always ever-so-slightly off:

        Expected: {"lng":100,"lat":0}
        Actual:   {"lng":99.99999999999994,"lat":0}

In the absence of #1854, I’m making the transition synchronous by setting animate: false in the options object. I don’t know if that has anything to do with the failures.

ref #1773

[jfirebaugh]

Does using fixedLngLat like in the other camera tests help?

[mourner]

@jfirebaugh I think this is a valid bug, we should set a final position here https://github.com/mapbox/mapbox-gl-js/blob/master/js/ui/camera.js#L632

[jfirebaugh]

@mourner I don't think that should be necessary. The first function passed to this._ease will be called for the final time with a k value corresponding to t=1, and should calculate the correct end point based on that. 99.99999999999994 looks like normal floating point imprecision to me.

[1ec5]

Tests added and edge cases fixed. See the original comment, which I’ve edited, for details.

[jfirebaugh]

👍 Awesome