point-cluster

Point clustering for 2D spatial indexing. Incorporates optimized quad-tree data structure.

const cluster = require('point-cluster')

let ids = cluster(points)

// get point ids in the indicated range
let selectedIds = ids.range([10, 10, 20, 20])

// get levels of details: list of ids subranges for rendering purposes
let lod = ids.range([10, 10, 20, 20], { lod: true })

API

ids = cluster(points, options?)

Create index for the set of 2d points based on options.

• points is an array of [x,y, x,y, ...] or [[x,y], [x,y], ...] coordinates.
• ids is Uint32Array with point ids sorted by zoom levels, suitable for WebGL buffer, subranging or alike.
• options

Option	Default	Description
bounds	'auto'	Data range, if different from points bounds, eg. in case of subdata.
depth	256	Max number of levels. Points below the indicated level are grouped into single level.
output	'array'	Output data array or data format. For available formats see dtype.

---

result = ids.range(box?, options?)

Get point ids from the indicated range.

• box can be any rectangle object, eg. [l, t, r, b], see parse-rect.
• options

Option	Default	Description
lod	false	Makes result a list of level details instead of ids, useful for obtaining subranges to render.
px	0	Min pixel size in data dimension (number or [width, height] couple) to search for, to ignore lower levels.
level	null	Max level to limit search.

let levels = ids.range([0,0, 100, 100], { lod: true, d: dataRange / canvas.width })

levels.forEach([from, to] => {
    // offset and count point to range in `ids` array
    render( ids.subarray( from, to ) )
})

Related

• snap-points-2d − grouping points by pixels.
• kdgrass − minimal kd-tree implementation.
• regl-scatter2d − highly performant scatter2d plot.

License

© 2017 Dmitry Yv. MIT License

Development supported by plot.ly.