polygonToCellsExperimental Compact Short-Circuit

[ben-burwood]

We use H3 heavily for a geospatial querying engine and have developed a use case which means we often want to compact a polyfill. Currently we do these as seperate steps:

polygonToCellsExperimental(polygon, r)      →  tile the whole polygon at resolution r
compactCells(cells)

For a large area at a high resolution, this materialises many many times the final cell count, where the vast majority of them immediately get compacted. So we see a high peak memory spike as the polyfill works.

It seems for this use case the high-res interior fill is wasted, and if we can prove containment of a coarse resolution cell then we don't need the high res children at all...
Considering a top-down (coarse to fine cell) approach is how the polygonToCellsExperimental is already implemented, is there any appetite to build in a compact flag which could short-circuit before recursing down into the children cells and return the compacted set?

[ben-burwood]

I've implemented a similar version of polygonToCellsExperimental including this 'compact short-circuit' to assess the feasibility.

https://github.com/ben-burwood/h3-compactfill

I've written this in Rust with the help of h3o and early result's show a large improvement in memory usage and runtime vs a serial polyfill + compact.

Maybe there's no desire for this but just offering it up incase anyone has the same usecase