So far, we didn't properly disable layers that are outside the zoom range. This means that we rendered layers that should not have been rendered, albeit we didn't make any attempt to load tiles for those layers. However, when zooming in/out, existing tiles might already have been loaded in the source which continued to be rendered. In most cases they weren't actually visible because either the matrices weren't updated, or the clip IDs weren't set so that they would be "rendered" off-screen and clipped completely. In any case, we did way too much work.
* [android] - do not update marker while creating, add activity to prevent regression
* [core] - add asserts to incorrect usage of update and remove annotations api
* [android] - fix marker tests
SymbolLayout lives on the worker thread and contains the persistent data needed for repeated placement. SymbolBucket contains the data generated during placement, and is transferred to the main thread for rendering.
This eliminates the risky sharing of GeometryTile::buckets between the main thread and worker thread during TileWorker::redoPlacement.
While here, rationalize the names of states a SymbolLayout may be in:
* Pending: Waiting for the necessary glyphs or icons to be available.
* Prepared: The potential positions of text and icons have been determined.
* Placed: The final positions have been determined, taking into account prior layers.
In TileWorker, all SymbolLayouts are stored in a single vector. Each SymbolLayout knows what state it is in, and TileWorker can easily determine how much progress it can make toward a final result.
Until Style::recalculate() is called to check that there are no visible layers using the source, we have to assume there are. Otherwise, Style::isLoaded() can return a false positive.
This paves the way for updates to filter and layout properties to trigger a source reload, without each SDK having to participate in the implementation.
* Renamed {Source,Tile}Observer::onNeedsRepaint to onTileUpdated. Messages should be in terms of what happened to the observed object, not in terms of what the observer needs to do. This also removes a confusing overlap of virtual methods on StyleObserver.
* Added style::Observer::onUpdate(Update). This is also a violation of the above rule, but I'm hopeful that it will disappear when update batching is implemented.
