Support for pmtiles

[aemonm]

Any plans to support pmtiles in a future release? I believe they are similar to MVTs but are queryable via HTTP range headers rather than the some-tile-server.com/z/x/y method.

[ibgreen]

• This would be more of a question for loaders.gl. deck.gl leverages loaders.gl for data loading and parsing.
• We don't currently have any plans for pmtiles, but it fits well into the general loaders.gl roadmap.
• Generally, Use of HTTP range requests has become very fashionable lately and there are a range of new formats adopting them (COGs, COPC, FlatGeobuf, etc) which we'd all love to provide support for at some point.

[ibgreen]

The TileSourceLayer that powers the PMTiles demo in loaders.gl is available in the @deck.gl-community/layers module.

[cboettig]

@ibgreen really nice, thanks! any pointers for getting TileSourceLayer to render in the corresponding pydeck code? I'm not able to find any docs or examples on the pydeck side with this unfortunately.

[ibgreen]

Hmm, no sorry, we haven't yet discussed making pydeck bindings for the deck.gl-community modules. I will try to bring it some attention, but It is a fair bit of work so can't guarantee that someone will pick it up.

[cboettig]

Thanks, understood. I tried experimenting based on https://github.com/agressin/pydeck_myTileLayer without success. I don't suppose there's any roadmap to make pmtiles a core tile layer to sidestep the challenges of extending generic community modules to pydeck? The performance difference with pmtiles (e.g. rendered with maplibre) is really remarkable.

[ibgreen]

Well, the TIleSource system in loaders.gl is still experimental and the idea was to let this take its time, and we'd publish in the community repo in the mean time to let early adopters start using it.

Cleaning up and reviewing the Source APIs and the TIleSourceLayer is a fair bit of work.

It could perhaps be added as an experimental, underscore, will-break-in-future-release export. I'll add it to agenda for next bi-weekly community meeting.