update Transform geo/screen conversions for tilt

[incanus]

In #2116 we have a PR for tilt/pitch. However we have yet to update Transform::pixelForLatLng & Transform::latLngForPixel to actually use the matrix instead of just doing trig for center, zoom, and rotation, so that pitch can be factored in. As a result, we get things like the floating user dot on iOS and these conversions will be off for e.g. long-press marker drop gestures, too.

Note that we do already account for world wrap and the date line in the current routines.

/cc @ansis @friedbunny @1ec5

[1ec5]

pixelForLatLng() and latLngForPixel() need to work with pitch, or things like fitToBounds() will yield unexpected results.

[incanus]

Right, this is not just cosmetic.

[incanus]

Per chat, it's unclear if -[MGLMapView metersPerPixelAtLatitude:] is affected by this. Need to investigate MapKit.

[ansis]

These should be implemented in 1551424 and fixed in 1081992.

I'm still working on:

• making the same location stay under your finger during panning, zooming, etc
• making fitBounds work better

[ansis]

Gesture improvements are finished in https://github.com/mapbox/mapbox-gl-native/commits/perspective-continued

• when you pan the same location stays under your finger
• when you zoom the location at the center of the pinch stays the same
• when you rotate the location at the center of the pinch stays the same

[incanus]

making fitBounds work better

Is above and/or anything else here still actionable @ansis?

[1ec5]

This is still actionable: fitBounds does not account for the trapezoidal shape of the bounds when the map is tilted. As a result, the map’s camera fails to round-trip correctly (tending to zoom out) when the map starts out tilted.

[incanus]

Just a clarification question here — does the fact that MapKit doesn’t use perspective, but just simple tilt, behave differently in this regard? I wonder if MKMapRect etc. automatically account for tilt.

[mb12]

@incanus is it possible to share what's the difference between Mapbox perspective and Map kit tilt and why one picks over the other?

[1ec5]

#2259 covers the fitBounds work that needs to take place. Let’s see if we can get @ansis’ work on the perspective-continued branch into master.

[1ec5]

The perspective-continued branch appears to have been deleted. @ansis, can you re-push your work towards this issue? Or is that now covered by the perspective-improved-gestures branch? Thanks.

[friedbunny]

I have re-pushed perspective-continued from my local copy, feel free to delete if unnecessary.

[1ec5]

I don’t think that was the tip of perspective-continued. Hopefully @ansis has the fixes described in #2139 (comment) around locally still.

[stale]

This issue has been automatically detected as stale because it has not had recent activity and will be archived. Thank you for your contributions.