Chevron DownAPI Reference


This is the high-level roadmap. For detailed information see

  • RFCs - technical writeups that describe proposed features.
  • Github Tracker Issues - "Tracker" issues contained detailed technical roadmaps for specific features/modules.

Core Feature Roadmap

Many ideas are in tracker tasks in github, but here are some ideas:

Worker Improvements

  • Worker Warmup: core should have an option to pre-load workers so that loader thread pool is primed and ready to start off-thread parsing as soon as data arrives on the wire. This can avoid a typical 1-2 second lag to load and parse the worker script on the initial load using that worker.
  • Node.js Workers: Supported since Node 10.6 / 12.3, would increase the appeal of in cloud/backend applications

Progress Tracking

  • loaders can provide progress callbacks and a ProgressTracker class to track the progress of a set of parallel loads.

Automatic Timing

  • could integrate with Stats.
  • loaders could export a global stats object.
  • load options could accept a stats parameter.
  • objects returned from loaders could contain a stats object with timing stats.
  • setDefaultOptions({stats: true}) to enable stats collection, etc.

MIME types

  • Allow MIME types (e.g. from response headers) to be provided to assist in loader auto-selection.

Writer Roadmap

Note that Writer support is currently minimal in, so far mainly due to lack of strong internal use cases.

  • Worker support for writers
  • Enable Writers to return recommended MIME type(s).

Loader Roadmap

Data loaders

  • Develop table category helper classes
  • Develop table to Arrow mapping support
  • Develop GeoJSON to Arrow mapping

Streaming tabular loaders


  • Improve perf of CSV Loader to match d3.dsv


  • Good example that shows various (compressed) formats loading
  • Decide on ImageWorkerLoader
  • Decide on type: 'data'
  • Basis image decoder, finalize API
  • Compressed image decoder, finalize API
  • Finalize documentation

Geospatial loaders

Focus on loading of large, complex geospatial data.

  • KML
  • Shapefile


  • MTL - Implement MTL format support, we should have full OBJ/MTL support.
  • Determine how materials fit into the Mesh category, or if adding MTL returns a Scenegraph category object.
  • OBJ can also support face groups => simple scenegraph?

Note: Given industry convergence on glTF, we do not envision supporting other mesh formats beyond OBJ/MTL.

Point Clouds

Focus on support formats for large point clouds.

  • Better example for point cloud loading
    • auto discover extents
    • support streaming loads (display points as they stream in).
    • better selector for data sets with preview images


  • Implement streaming load
  • load emscripten-lib from unpkg CDN
  • compile C++ to WASM instead of JS?
  • investigate support for LAS 1.4?

3D Tiles

  • Support for alternative/non-geospatial coordinate systems (potree)
  • Support unlit materials
  • Finalize i3s support
  • Finalize potree support
  • Better example
    • better selector for data sets with preview images

Scenegraph Formats

Focus on glTF/GLB - should to have a very solid implementation.

  • The glTF loaders should handle (e.g. preprocess) any glTF extensions that can be handled during the load phase (such as Draco, Basis - but many can only be handled during rendering).

What will NOT support: Given the emergence of glTF as a major Khronos standard, and availability of good glTF conversion tools and exporters, loaders will most likely not supprt any other large scene/mesh description formats such as COLLADA.

Other loaders

Finally, some "unusual" loaders may be included just for fun, e.g. SVG tesselation.