Awesome
geokdbush
A geographic extension for KDBush, the fastest static spatial index for points in JavaScript.
It implements fast nearest neighbors queries for locations on Earth, taking Earth curvature and date line wrapping into account. Inspired by sphere-knn, but uses a different algorithm.
Example
import KDBush from 'kdbush';
import * as geokdbush from 'geokdbush';
const index = new KDBush(points.length);
for (const {lon, lat} of points) index.add(lon, lat);
index.finish();
const nearestIds = geokdbush.around(index, -119.7051, 34.4363, 1000);
const nearest = nearestIds.map(id => points[id]);
API
geokdbush.around(index, longitude, latitude[, maxResults, maxDistance, filterFn])
Returns an array of the closest points from a given location in order of increasing distance.
index
: kdbush index.longitude
: query point longitude.latitude
: query point latitude.maxResults
: (optional) maximum number of points to return (Infinity
by default).maxDistance
: (optional) maximum distance in kilometers to search within (Infinity
by default).filterFn
: (optional) a function to filter the results with.
geokdbush.distance(longitude1, latitude1, longitude2, latitude2)
Returns great circle distance between two locations in kilometers.
Performance
This library is incredibly fast.
The results below were obtained with npm run bench
(Node v20, Macbook Pro 2020 M1 Pro).
benchmark | geokdbush | sphere-knn | naive |
---|---|---|---|
index 138398 points | 57.6ms | 408ms | n/a |
query 1000 closest | 1.6ms | 1.8ms | 72ms |
query 50000 closest | 14.7ms | 91.5ms | 72ms |
query all 138398 | 33.7ms | 500ms | 72ms |
1000 queries of 1 | 24.7ms | 27.5ms | 11.1s |