perf: ST_Buffer implementation using geo
#233
Conversation
| } | ||
|
|
||
| #[test] | ||
| fn test_empty_geometry() { |
There was a problem hiding this comment.
The test suite here is the same as the geos buffer test suite plus this new function, which I also copied over to geos st_buffer to be sure it works the same.
| // PostGIS returns POLYGON EMPTY for all empty geometries | ||
| let is_empty = is_geometry_empty(wkb).map_err(|e| DataFusionError::External(Box::new(e)))?; | ||
| if is_empty { |
There was a problem hiding this comment.
I was running into an error here with POINT EMPTY since geo apparently doesn't support it.
sedona-db/rust/sedona-geo/src/to_geo.rs
Lines 62 to 66 in 3d75664
My current workaround is to use WKBExecutor here instead of GeoTypesExecutor and use our native is_geometry_empty check.
Wondering if there's a better way we can handle empty points in our item_to_geometry() method. The docstring for geo's try_to_point() function we are in there says returning None represents an empty point. Though returning None is not a safe option, so I can't think of anything better than this workaround atm.
|
Are there any convenient existing ways to benchmark this against the |
This should work: # on main
cargo bench -- st_buffer
# switch to your branch
cargo bench -- st_buffer |
2 participants
closes #232