search for k nearest neighbors on a sphere


sphere-knn is a Node.JS module that provides fast nearest-neighbor lookups on a sphere. This is useful if, for example, you have a database of geographic points (latitude, longitude) and want to swiftly look up which of those points are near a given latitude, longitude pair. It came out of a need to do such lookups in the Dark Sky API, but all existing libraries either threw scary runtime flags, were too slow, broke at the International Date Line, or didn't have tests.

So anyway, this one is well-tested and works correctly regardless of where on the earth things are located. It's been in production use at Dark Sky since Oct 2012.

To install:

npm install sphere-knn

To use:

var sphereKnn = require("sphere-knn"),
    lookup    = sphereKnn([
      /* This array needs to be full of objects that have latitudes and
       * longitudes. Accepted property names are "lat", "latitude", "lon",
       * "lng", "long", "longitude". */
      {lat: ..., lon: ...},

var points = lookup(someLatitude, someLongitude, maxResults, maxDistance)

The points array consists of objects that were in the array passed to sphereKnn(), ordered from nearest to furthest. The maxResults value is the maximum size of the returned array, and is mandatory. (Often, you'll just want it set to 1, but there are use-cases for more points.) The maxDistance value is the maximum distance away we should look up for in meters. This is handy if you want to find, say, any points within 200 kilometers of a given point. maxDistance is optional and defaults to Infinity. (Please note that since our backing earth model is a sphere, all distances given to this function are approximate. If it matters that much to you, add a few meters of padding and check the results with some very accurate (and complex) distance function.)