[core] Limit pitch based on edge insets. Fix max Z calculation in getProjMatrix.

[astojilj]

Patch partly fixes #15163 in a way that it doesn't allow loading tens of thousands of tiles and attempt to show area above horizon:

1. Limit pitch based on edge insets. It is not too bad - current limit of 60 degrees stays active until center of perspective is moved towards the bottom, to 84% of screen height. The plan is to split removal of 60 degrees limit to follow up patch.
2. Fix max Z calculation in getProjMatrix. TransformState::getProjMatrix calculation of farZ was complex with possibility to lead to negative z values. Replacing it with simpler, precise calculation:

furthestDistance = cameraToCenterDistance / (1 - tanFovAboveCenter * std::tan(getPitch()));

Mapping diagram to code:
fD: furthestDistance,
c: cameraToCenterDistance,
small phi: getPitch()

Addresses: #15163

[1ec5]

Note that the failsafe introduced here introduces a change in behavior that’ll be noticeable in the navigation SDK and navigation-related applications. It’s a sound move, but it does put more pressure on us to implement adaptive zoom levels (aka level of detail): #9037.

[astojilj]

Note that the failsafe introduced here introduces a change in behavior that’ll be noticeable in the navigation SDK and navigation-related applications. It’s a sound move, but it does put more pressure on us to implement adaptive zoom levels (aka level of detail): #9037.

@1ec5,

Edited the comment:

The only change in behavior I see if moving center of perspective towards the bottom, to 84% of screen height. Then app wouldn't try to load e.g. 20000+ tiles but e.g. 500. It would still be very slow but prevents crash and is improvement compared to current state.

About change in behavior:

• I don't think there is a change from API point of view: https://docs.mapbox.com/mapbox-gl-js/api/#cameraoptions describes it "desired pitch value". Didn't see the limit of 60 degrees in documentation.

The change here is done this way to prevent current visual change (it is still 60 degrees for smaller padding, like it was), and allow the largest possible pitch that doesn't show area above horizon.

Agree, there is a pressure to work on #9037 immediately, with or without changes in this PR. I to work on #9037

Did I understand the "change in behavior" correctly or there is something additional?

[astojilj]

@1ec5 , @alexshalamov
PTAL.
As mentioned, this is first part of fixing #15195.

@alexshalamov I'll split the tile cover optimization to other patch - here, contributing unit tests for verifying the upcoming patch against current state.

[astojilj]

@1ec5 ,
Before merging this, I'd like to verify that I've correctly understood and addressed your concerns here.

Work on this, and making the max pitch larger, continues in ##15230. But, I'd like to push this patch first as it is failsafe, preventing out of memory and block but not altering currently max allowed/possible pitch.

[1ec5]

Work on this, and making the max pitch larger, continues in ##15230. But, I'd like to push this patch first as it is failsafe, preventing out of memory and block but not altering currently max allowed/possible pitch.

This is a good plan, thank you!

[1ec5]

Suggested changelog entry:

• Fixed an issue where it was possible to set the map’s content insets then tilt the map enough to see the horizon, causing performance issues.

[astojilj]

Suggested changelog entry:

• Fixed an issue where it was possible to set the map’s content insets then tilt the map enough to see the horizon, causing performance issues.

Thanks, added.