Add useful H3 functions

[bziobrowski]

Currently Apache Pinot makes use of H3 only in certain scenarios, e.g. to speed up predicates using st_distance function against h3-indexed points.
To fully utilize it, it should expose more functions available in com.uber.h3core.H3Core class, especially:

• gridDistance()
• gridDisk()
• geoToH3()

Example:

Assuming there's stores table with h3_pos long column (with inverted index) storing h3 index value (at resolution 8) of store's position, we could query for stores in closest 4 rings to given coordinates with:

select * 
from stores
where h3_pos in gridDisk( geoToH3(10.1, 20.3, 8), 4 )

or

select * 
from stores
where h3Distance(h3_pos, geoToH3(10.1, 20.3, 8) ) <= 4

There are many more functions that could be useful for expressing logic in terms of hexagons and hexagon distance that really suit H3.

See:
https://h3geo.org/docs/api/traversal
https://github.com/uber/h3-java/blob/master/src/main/java/com/uber/h3core/H3Core.java

[meshari343]

Hi, I would like to take on this issue, can someone please assign it to me, and if possible guide me towards the modules I need to work on.

[Jackie-Jiang]

@bziobrowski I guess we should submit this issue to h3 repo because we are asking for these functionalities from their end

[meshari343]

@Jackie-Jiang those functionalities are already implemented in h3 repo, we just need to integrated within Pinot.

[Jackie-Jiang]

@meshari343 I see. Assigned it to you. Let us know if you need more help, thanks!

[ashishjayamohan]

@meshari343 Are you still working on this? If not, can I take a crack at it?

[meshari343]

@ashishjayamohan , I still haven’t fully understood the design of the system, but since you’re interested on this issue then sure you’re more than welcome.