Global feature IDs

[bagnell]

• The batch ids are limited to [0, featuresLength). The batch id should be allowed to be any arbitrary number.
• For vector tiles, batch ids are required for all features or they are all generated.Batch ids should be generated if they are not provided.
• For vector tiles, a feature with e the same batch id should be styled the same across tiles.

[lilleyse]

I'm running into the same sorts of issue with classified point clouds. We may need a concept of global batch ids.

[pjcozzi]

@lilleyse can you give an example for global batch ids

@bagnell would Cesium still have a fast path that basically avoids the indirect texture lookup when batch ids are in range? We could make that an Implementation Note in the 3D Tiles spec; we do this often for glTF.

[lilleyse]

@lilleyse can you give an example for global batch ids

There could be a case where you have a feature whose points are split between 8 tiles and you want to change the feature's color or edit its properties. A common case is changing its color on pick. In Cesium you would only be able to pick an eighth of the feature at a time. With a global id you could at least identify which features across all tiles belong to the same thing.

This isn't really a problem though if your tileset is read-only and all you want to do is style it.

[pjcozzi]

Ah yes, we should scope out what it will take to address this and decide if it fits into 1.0.

[pjcozzi]

@lilleyse realistically can this be scoped to be small and quick for 1.0? Or hold for the future?

[lilleyse]

This one can hold - it will require a bit of thinking to do properly and it's possible for an app to work around it.

[lilleyse]

This is still very relevant for replacement-refinement tilesets that have the same features across tile boundaries or across LODs - e.g. photogrammetry and vector data.

Global IDs can be stored in a property table indexed by local feature IDs. These global IDs can be used to access metadata from an external database, or maybe from an external property table (CC #136).

If this is a common enough use case there could be a GLOBAL_ID metadata semantic.

[lilleyse]

Related discussion in #407 about querying features in a tileset, possibly based on some global ID.

[donmccurdy]

Perhaps the existing ID semantic? We don't define "unique" — which may differ for 3D Tiles and glTF — but I think the idea is the same.

[ptrgags]

@donmccurdy note that ID is a STRING (so more like an identifier in a programming language than an index). But certainly we could have a separate semantic

[donmccurdy]

It sounded like the idea of a Global ID was to provide an ID that is not constrained to be an integer [0, count -1], have I misunderstood that? We could also loosen the type requirement of ID, if the expected meaning is otherwise consistent.