Connect deck.gl-healpix with the DGGS Zarr convention

[maxrjones]

It'd be nice to find time this PI (i.e., before July 11) to collaborate with @danielfdsilva to update developmentseed/deck.gl-healpix#7 to use the zarr dggs convention (fyi @keewis)

[keewis]

I've refactored healpix-geo into a python module and a rust crate and I'm planning to publish js bindings on node.js, so maybe deck.gl-healpix could make use of that for the ellipsoid support? Also it might be interesting to integrate with deck.gl-raster (not sure if deck.gl-healpix already does that).

[danielfdsilva]

@maxrjones I'd be very interested in doing that. Were can I see the progress on the convention? Did you have a chance to look at https://github.com/developmentseed/deck.gl-healpix/blob/29a1c532c34ee0f3dc317fbe9202731efee416d6/docs/specs/healpix-pyramid-zarr.md for the proposed structure needs? (Happy to talk through this if there's a different option)

@keewis I'd be keen to learn more about this ellipsoid support. Not very familiar with this part. Would also be interested in knowing what sort of integration you mean with deck-gl.raster. The libraries have a similar output, but they work in different ways given that the data is also different. HEALPix not being raster, I'm not sure how an integration would look like.

[keewis]

The DGGS zarr convention is at https://github.com/zarr-conventions/dggs. I was thinking it would be nice to use the multiscales convention for image pyramids (see zarr-conventions/dggs#23).

Indeed, individual healpix cells are closer to vector than raster data. However, if you have a contiguous region of cells, it kinda becomes raster data (depending on how strict your definition of "raster" is). In that case it might be nice to not have to render each cell individually. If we were to have a efficient way to convert the healpix cell region into a mesh, we might get similar speed ups as deck.gl-raster got for the more traditional raster data.

[keewis]

oh, and ellipsoid support comes down to adding an intermediate layer between the spherical healpix implementation and any function that interacts with geographic coordinates (I believe this is also how map projections are commonly extended to the ellipsoid). For example, for converting ellipsoidal-geographic coordinates to cell ids, that layer converts the latitude to authalic latitude such that we have spherical-geographic coordinates and can use the standard spherical healpix functions to compute to the cell ids.

See https://healpix-geo.readthedocs.io/en/latest/healpix/ellipsoidal-extension.html for a schematic of the process, which follows The rHEALPix Discrete Global Grid System, Appendix A.

[danielfdsilva]

@keewis The deck.gl-healpix does not convert the cells to polygons for them to be rendered. The conversion is done directly to the triangle positions in the GPU, having pretty good speed. Check this example: https://deck.gl-healpix.ds.io/examples/zarr-tiles/

Thank you for the links. I'll check them out. If needed it shouldn't be a problem to add!

[keewis]

that looks interesting, thanks, I'll have a look.

---

Not really related, but accessing https://deck.gl-healpix.ds.io/ gives me infinite ~and~ added to the url... something like

https://deck.gl-healpix.ds.io///?/&/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/~and~/