rust: Eliminate dependency on forked geo-index crate

[paleolimbot]

We use the geo-index crate ( https://github.com/kylebarron/geo-index ) heavily in sedona-spatial join. In the lead up to the release, we discovered a few features that we needed to iterate on in conjunction with our at the time not-yet-public code ( https://github.com/wherobots/geo-index ). Now that we have a release, we should upstream the required changes or expose hooks in geo-index such that we can separate the code we need for the KNN join with released geo-index!

Note that many of the changes in the fork have already been upstreamed but I think the one involving spherical distance hasn't quite made it there yet.

cc @zhangfengcdt

[jiayuasu]

I agree with Kyle here: kylebarron/geo-index#135

Why did we use heavy sphere distance calculation in place in geo-index?

CC @zhangfengcdt @Kontinuation

[Kontinuation]

This is a valid concern. geo crate already has Haversine and Geodesic for point <-> point distance calculation.

Sedona-db's KNN join requires yet another patch to be merged into upstream: wherobots/geo-index#5. I have also suggested defining DistanceMetric more flexibly for that patch: #53 (review).

[zhangfengcdt]

I agree with Kyle here: kylebarron/geo-index#135

Why did we use heavy sphere distance calculation in place in geo-index?

CC @zhangfengcdt @Kontinuation

@jiayuasu The PR Kyle created seems using old code from our geo-index fork. The latest one already uses the geo crate's distance calculation:
wherobots/geo-index#5

[zhangfengcdt]

This is a valid concern. geo crate already has Haversine and Geodesic for point <-> point distance calculation.

Sedona-db's KNN join requires yet another patch to be merged into upstream: wherobots/geo-index#5. I have also suggested defining DistanceMetric more flexibly for that patch: #53 (review).

Our geo-index was already refactored to use the geo crate distance function.

w.r.t. the refactoring to neighbors_geometry to be indices based parameters, I will try to implement it (maybe coexist with the current geometries approach) before merging to upstream.

[jiayuasu]

Thanks. Let's get it done by 0.2.0. We will eliminate all forks in 0.2.0 and release to crates.io