759 lines
26 KiB
JavaScript
759 lines
26 KiB
JavaScript
(function (global, factory) {
|
|
typeof exports === 'object' && typeof module !== 'undefined' ? module.exports = factory() :
|
|
typeof define === 'function' && define.amd ? define(factory) :
|
|
(global = typeof globalThis !== 'undefined' ? globalThis : global || self, global.Supercluster = factory());
|
|
})(this, (function () { 'use strict';
|
|
|
|
const ARRAY_TYPES = [
|
|
Int8Array, Uint8Array, Uint8ClampedArray, Int16Array, Uint16Array,
|
|
Int32Array, Uint32Array, Float32Array, Float64Array
|
|
];
|
|
|
|
/** @typedef {Int8ArrayConstructor | Uint8ArrayConstructor | Uint8ClampedArrayConstructor | Int16ArrayConstructor | Uint16ArrayConstructor | Int32ArrayConstructor | Uint32ArrayConstructor | Float32ArrayConstructor | Float64ArrayConstructor} TypedArrayConstructor */
|
|
|
|
const VERSION = 1; // serialized format version
|
|
const HEADER_SIZE = 8;
|
|
|
|
class KDBush {
|
|
|
|
/**
|
|
* Creates an index from raw `ArrayBuffer` data.
|
|
* @param {ArrayBuffer} data
|
|
*/
|
|
static from(data) {
|
|
if (!(data instanceof ArrayBuffer)) {
|
|
throw new Error('Data must be an instance of ArrayBuffer.');
|
|
}
|
|
const [magic, versionAndType] = new Uint8Array(data, 0, 2);
|
|
if (magic !== 0xdb) {
|
|
throw new Error('Data does not appear to be in a KDBush format.');
|
|
}
|
|
const version = versionAndType >> 4;
|
|
if (version !== VERSION) {
|
|
throw new Error(`Got v${version} data when expected v${VERSION}.`);
|
|
}
|
|
const ArrayType = ARRAY_TYPES[versionAndType & 0x0f];
|
|
if (!ArrayType) {
|
|
throw new Error('Unrecognized array type.');
|
|
}
|
|
const [nodeSize] = new Uint16Array(data, 2, 1);
|
|
const [numItems] = new Uint32Array(data, 4, 1);
|
|
|
|
return new KDBush(numItems, nodeSize, ArrayType, data);
|
|
}
|
|
|
|
/**
|
|
* Creates an index that will hold a given number of items.
|
|
* @param {number} numItems
|
|
* @param {number} [nodeSize=64] Size of the KD-tree node (64 by default).
|
|
* @param {TypedArrayConstructor} [ArrayType=Float64Array] The array type used for coordinates storage (`Float64Array` by default).
|
|
* @param {ArrayBuffer} [data] (For internal use only)
|
|
*/
|
|
constructor(numItems, nodeSize = 64, ArrayType = Float64Array, data) {
|
|
if (isNaN(numItems) || numItems < 0) throw new Error(`Unpexpected numItems value: ${numItems}.`);
|
|
|
|
this.numItems = +numItems;
|
|
this.nodeSize = Math.min(Math.max(+nodeSize, 2), 65535);
|
|
this.ArrayType = ArrayType;
|
|
this.IndexArrayType = numItems < 65536 ? Uint16Array : Uint32Array;
|
|
|
|
const arrayTypeIndex = ARRAY_TYPES.indexOf(this.ArrayType);
|
|
const coordsByteSize = numItems * 2 * this.ArrayType.BYTES_PER_ELEMENT;
|
|
const idsByteSize = numItems * this.IndexArrayType.BYTES_PER_ELEMENT;
|
|
const padCoords = (8 - idsByteSize % 8) % 8;
|
|
|
|
if (arrayTypeIndex < 0) {
|
|
throw new Error(`Unexpected typed array class: ${ArrayType}.`);
|
|
}
|
|
|
|
if (data && (data instanceof ArrayBuffer)) { // reconstruct an index from a buffer
|
|
this.data = data;
|
|
this.ids = new this.IndexArrayType(this.data, HEADER_SIZE, numItems);
|
|
this.coords = new this.ArrayType(this.data, HEADER_SIZE + idsByteSize + padCoords, numItems * 2);
|
|
this._pos = numItems * 2;
|
|
this._finished = true;
|
|
} else { // initialize a new index
|
|
this.data = new ArrayBuffer(HEADER_SIZE + coordsByteSize + idsByteSize + padCoords);
|
|
this.ids = new this.IndexArrayType(this.data, HEADER_SIZE, numItems);
|
|
this.coords = new this.ArrayType(this.data, HEADER_SIZE + idsByteSize + padCoords, numItems * 2);
|
|
this._pos = 0;
|
|
this._finished = false;
|
|
|
|
// set header
|
|
new Uint8Array(this.data, 0, 2).set([0xdb, (VERSION << 4) + arrayTypeIndex]);
|
|
new Uint16Array(this.data, 2, 1)[0] = nodeSize;
|
|
new Uint32Array(this.data, 4, 1)[0] = numItems;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* Add a point to the index.
|
|
* @param {number} x
|
|
* @param {number} y
|
|
* @returns {number} An incremental index associated with the added item (starting from `0`).
|
|
*/
|
|
add(x, y) {
|
|
const index = this._pos >> 1;
|
|
this.ids[index] = index;
|
|
this.coords[this._pos++] = x;
|
|
this.coords[this._pos++] = y;
|
|
return index;
|
|
}
|
|
|
|
/**
|
|
* Perform indexing of the added points.
|
|
*/
|
|
finish() {
|
|
const numAdded = this._pos >> 1;
|
|
if (numAdded !== this.numItems) {
|
|
throw new Error(`Added ${numAdded} items when expected ${this.numItems}.`);
|
|
}
|
|
// kd-sort both arrays for efficient search
|
|
sort(this.ids, this.coords, this.nodeSize, 0, this.numItems - 1, 0);
|
|
|
|
this._finished = true;
|
|
return this;
|
|
}
|
|
|
|
/**
|
|
* Search the index for items within a given bounding box.
|
|
* @param {number} minX
|
|
* @param {number} minY
|
|
* @param {number} maxX
|
|
* @param {number} maxY
|
|
* @returns {number[]} An array of indices correponding to the found items.
|
|
*/
|
|
range(minX, minY, maxX, maxY) {
|
|
if (!this._finished) throw new Error('Data not yet indexed - call index.finish().');
|
|
|
|
const {ids, coords, nodeSize} = this;
|
|
const stack = [0, ids.length - 1, 0];
|
|
const result = [];
|
|
|
|
// recursively search for items in range in the kd-sorted arrays
|
|
while (stack.length) {
|
|
const axis = stack.pop() || 0;
|
|
const right = stack.pop() || 0;
|
|
const left = stack.pop() || 0;
|
|
|
|
// if we reached "tree node", search linearly
|
|
if (right - left <= nodeSize) {
|
|
for (let i = left; i <= right; i++) {
|
|
const x = coords[2 * i];
|
|
const y = coords[2 * i + 1];
|
|
if (x >= minX && x <= maxX && y >= minY && y <= maxY) result.push(ids[i]);
|
|
}
|
|
continue;
|
|
}
|
|
|
|
// otherwise find the middle index
|
|
const m = (left + right) >> 1;
|
|
|
|
// include the middle item if it's in range
|
|
const x = coords[2 * m];
|
|
const y = coords[2 * m + 1];
|
|
if (x >= minX && x <= maxX && y >= minY && y <= maxY) result.push(ids[m]);
|
|
|
|
// queue search in halves that intersect the query
|
|
if (axis === 0 ? minX <= x : minY <= y) {
|
|
stack.push(left);
|
|
stack.push(m - 1);
|
|
stack.push(1 - axis);
|
|
}
|
|
if (axis === 0 ? maxX >= x : maxY >= y) {
|
|
stack.push(m + 1);
|
|
stack.push(right);
|
|
stack.push(1 - axis);
|
|
}
|
|
}
|
|
|
|
return result;
|
|
}
|
|
|
|
/**
|
|
* Search the index for items within a given radius.
|
|
* @param {number} qx
|
|
* @param {number} qy
|
|
* @param {number} r Query radius.
|
|
* @returns {number[]} An array of indices correponding to the found items.
|
|
*/
|
|
within(qx, qy, r) {
|
|
if (!this._finished) throw new Error('Data not yet indexed - call index.finish().');
|
|
|
|
const {ids, coords, nodeSize} = this;
|
|
const stack = [0, ids.length - 1, 0];
|
|
const result = [];
|
|
const r2 = r * r;
|
|
|
|
// recursively search for items within radius in the kd-sorted arrays
|
|
while (stack.length) {
|
|
const axis = stack.pop() || 0;
|
|
const right = stack.pop() || 0;
|
|
const left = stack.pop() || 0;
|
|
|
|
// if we reached "tree node", search linearly
|
|
if (right - left <= nodeSize) {
|
|
for (let i = left; i <= right; i++) {
|
|
if (sqDist(coords[2 * i], coords[2 * i + 1], qx, qy) <= r2) result.push(ids[i]);
|
|
}
|
|
continue;
|
|
}
|
|
|
|
// otherwise find the middle index
|
|
const m = (left + right) >> 1;
|
|
|
|
// include the middle item if it's in range
|
|
const x = coords[2 * m];
|
|
const y = coords[2 * m + 1];
|
|
if (sqDist(x, y, qx, qy) <= r2) result.push(ids[m]);
|
|
|
|
// queue search in halves that intersect the query
|
|
if (axis === 0 ? qx - r <= x : qy - r <= y) {
|
|
stack.push(left);
|
|
stack.push(m - 1);
|
|
stack.push(1 - axis);
|
|
}
|
|
if (axis === 0 ? qx + r >= x : qy + r >= y) {
|
|
stack.push(m + 1);
|
|
stack.push(right);
|
|
stack.push(1 - axis);
|
|
}
|
|
}
|
|
|
|
return result;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* @param {Uint16Array | Uint32Array} ids
|
|
* @param {InstanceType<TypedArrayConstructor>} coords
|
|
* @param {number} nodeSize
|
|
* @param {number} left
|
|
* @param {number} right
|
|
* @param {number} axis
|
|
*/
|
|
function sort(ids, coords, nodeSize, left, right, axis) {
|
|
if (right - left <= nodeSize) return;
|
|
|
|
const m = (left + right) >> 1; // middle index
|
|
|
|
// sort ids and coords around the middle index so that the halves lie
|
|
// either left/right or top/bottom correspondingly (taking turns)
|
|
select(ids, coords, m, left, right, axis);
|
|
|
|
// recursively kd-sort first half and second half on the opposite axis
|
|
sort(ids, coords, nodeSize, left, m - 1, 1 - axis);
|
|
sort(ids, coords, nodeSize, m + 1, right, 1 - axis);
|
|
}
|
|
|
|
/**
|
|
* Custom Floyd-Rivest selection algorithm: sort ids and coords so that
|
|
* [left..k-1] items are smaller than k-th item (on either x or y axis)
|
|
* @param {Uint16Array | Uint32Array} ids
|
|
* @param {InstanceType<TypedArrayConstructor>} coords
|
|
* @param {number} k
|
|
* @param {number} left
|
|
* @param {number} right
|
|
* @param {number} axis
|
|
*/
|
|
function select(ids, coords, k, left, right, axis) {
|
|
|
|
while (right > left) {
|
|
if (right - left > 600) {
|
|
const n = right - left + 1;
|
|
const m = k - left + 1;
|
|
const z = Math.log(n);
|
|
const s = 0.5 * Math.exp(2 * z / 3);
|
|
const sd = 0.5 * Math.sqrt(z * s * (n - s) / n) * (m - n / 2 < 0 ? -1 : 1);
|
|
const newLeft = Math.max(left, Math.floor(k - m * s / n + sd));
|
|
const newRight = Math.min(right, Math.floor(k + (n - m) * s / n + sd));
|
|
select(ids, coords, k, newLeft, newRight, axis);
|
|
}
|
|
|
|
const t = coords[2 * k + axis];
|
|
let i = left;
|
|
let j = right;
|
|
|
|
swapItem(ids, coords, left, k);
|
|
if (coords[2 * right + axis] > t) swapItem(ids, coords, left, right);
|
|
|
|
while (i < j) {
|
|
swapItem(ids, coords, i, j);
|
|
i++;
|
|
j--;
|
|
while (coords[2 * i + axis] < t) i++;
|
|
while (coords[2 * j + axis] > t) j--;
|
|
}
|
|
|
|
if (coords[2 * left + axis] === t) swapItem(ids, coords, left, j);
|
|
else {
|
|
j++;
|
|
swapItem(ids, coords, j, right);
|
|
}
|
|
|
|
if (j <= k) left = j + 1;
|
|
if (k <= j) right = j - 1;
|
|
}
|
|
}
|
|
|
|
/**
|
|
* @param {Uint16Array | Uint32Array} ids
|
|
* @param {InstanceType<TypedArrayConstructor>} coords
|
|
* @param {number} i
|
|
* @param {number} j
|
|
*/
|
|
function swapItem(ids, coords, i, j) {
|
|
swap(ids, i, j);
|
|
swap(coords, 2 * i, 2 * j);
|
|
swap(coords, 2 * i + 1, 2 * j + 1);
|
|
}
|
|
|
|
/**
|
|
* @param {InstanceType<TypedArrayConstructor>} arr
|
|
* @param {number} i
|
|
* @param {number} j
|
|
*/
|
|
function swap(arr, i, j) {
|
|
const tmp = arr[i];
|
|
arr[i] = arr[j];
|
|
arr[j] = tmp;
|
|
}
|
|
|
|
/**
|
|
* @param {number} ax
|
|
* @param {number} ay
|
|
* @param {number} bx
|
|
* @param {number} by
|
|
*/
|
|
function sqDist(ax, ay, bx, by) {
|
|
const dx = ax - bx;
|
|
const dy = ay - by;
|
|
return dx * dx + dy * dy;
|
|
}
|
|
|
|
const defaultOptions = {
|
|
minZoom: 0, // min zoom to generate clusters on
|
|
maxZoom: 16, // max zoom level to cluster the points on
|
|
minPoints: 2, // minimum points to form a cluster
|
|
radius: 40, // cluster radius in pixels
|
|
extent: 512, // tile extent (radius is calculated relative to it)
|
|
nodeSize: 64, // size of the KD-tree leaf node, affects performance
|
|
log: false, // whether to log timing info
|
|
|
|
// whether to generate numeric ids for input features (in vector tiles)
|
|
generateId: false,
|
|
|
|
// a reduce function for calculating custom cluster properties
|
|
reduce: null, // (accumulated, props) => { accumulated.sum += props.sum; }
|
|
|
|
// properties to use for individual points when running the reducer
|
|
map: props => props // props => ({sum: props.my_value})
|
|
};
|
|
|
|
const fround = Math.fround || (tmp => ((x) => { tmp[0] = +x; return tmp[0]; }))(new Float32Array(1));
|
|
|
|
const OFFSET_ZOOM = 2;
|
|
const OFFSET_ID = 3;
|
|
const OFFSET_PARENT = 4;
|
|
const OFFSET_NUM = 5;
|
|
const OFFSET_PROP = 6;
|
|
|
|
class Supercluster {
|
|
constructor(options) {
|
|
this.options = Object.assign(Object.create(defaultOptions), options);
|
|
this.trees = new Array(this.options.maxZoom + 1);
|
|
this.stride = this.options.reduce ? 7 : 6;
|
|
this.clusterProps = [];
|
|
}
|
|
|
|
load(points) {
|
|
const {log, minZoom, maxZoom} = this.options;
|
|
|
|
if (log) console.time('total time');
|
|
|
|
const timerId = `prepare ${ points.length } points`;
|
|
if (log) console.time(timerId);
|
|
|
|
this.points = points;
|
|
|
|
// generate a cluster object for each point and index input points into a KD-tree
|
|
const data = [];
|
|
|
|
for (let i = 0; i < points.length; i++) {
|
|
const p = points[i];
|
|
if (!p.geometry) continue;
|
|
|
|
const [lng, lat] = p.geometry.coordinates;
|
|
const x = fround(lngX(lng));
|
|
const y = fround(latY(lat));
|
|
// store internal point/cluster data in flat numeric arrays for performance
|
|
data.push(
|
|
x, y, // projected point coordinates
|
|
Infinity, // the last zoom the point was processed at
|
|
i, // index of the source feature in the original input array
|
|
-1, // parent cluster id
|
|
1 // number of points in a cluster
|
|
);
|
|
if (this.options.reduce) data.push(0); // noop
|
|
}
|
|
let tree = this.trees[maxZoom + 1] = this._createTree(data);
|
|
|
|
if (log) console.timeEnd(timerId);
|
|
|
|
// cluster points on max zoom, then cluster the results on previous zoom, etc.;
|
|
// results in a cluster hierarchy across zoom levels
|
|
for (let z = maxZoom; z >= minZoom; z--) {
|
|
const now = +Date.now();
|
|
|
|
// create a new set of clusters for the zoom and index them with a KD-tree
|
|
tree = this.trees[z] = this._createTree(this._cluster(tree, z));
|
|
|
|
if (log) console.log('z%d: %d clusters in %dms', z, tree.numItems, +Date.now() - now);
|
|
}
|
|
|
|
if (log) console.timeEnd('total time');
|
|
|
|
return this;
|
|
}
|
|
|
|
getClusters(bbox, zoom) {
|
|
let minLng = ((bbox[0] + 180) % 360 + 360) % 360 - 180;
|
|
const minLat = Math.max(-90, Math.min(90, bbox[1]));
|
|
let maxLng = bbox[2] === 180 ? 180 : ((bbox[2] + 180) % 360 + 360) % 360 - 180;
|
|
const maxLat = Math.max(-90, Math.min(90, bbox[3]));
|
|
|
|
if (bbox[2] - bbox[0] >= 360) {
|
|
minLng = -180;
|
|
maxLng = 180;
|
|
} else if (minLng > maxLng) {
|
|
const easternHem = this.getClusters([minLng, minLat, 180, maxLat], zoom);
|
|
const westernHem = this.getClusters([-180, minLat, maxLng, maxLat], zoom);
|
|
return easternHem.concat(westernHem);
|
|
}
|
|
|
|
const tree = this.trees[this._limitZoom(zoom)];
|
|
const ids = tree.range(lngX(minLng), latY(maxLat), lngX(maxLng), latY(minLat));
|
|
const data = tree.data;
|
|
const clusters = [];
|
|
for (const id of ids) {
|
|
const k = this.stride * id;
|
|
clusters.push(data[k + OFFSET_NUM] > 1 ? getClusterJSON(data, k, this.clusterProps) : this.points[data[k + OFFSET_ID]]);
|
|
}
|
|
return clusters;
|
|
}
|
|
|
|
getChildren(clusterId) {
|
|
const originId = this._getOriginId(clusterId);
|
|
const originZoom = this._getOriginZoom(clusterId);
|
|
const errorMsg = 'No cluster with the specified id.';
|
|
|
|
const tree = this.trees[originZoom];
|
|
if (!tree) throw new Error(errorMsg);
|
|
|
|
const data = tree.data;
|
|
if (originId * this.stride >= data.length) throw new Error(errorMsg);
|
|
|
|
const r = this.options.radius / (this.options.extent * Math.pow(2, originZoom - 1));
|
|
const x = data[originId * this.stride];
|
|
const y = data[originId * this.stride + 1];
|
|
const ids = tree.within(x, y, r);
|
|
const children = [];
|
|
for (const id of ids) {
|
|
const k = id * this.stride;
|
|
if (data[k + OFFSET_PARENT] === clusterId) {
|
|
children.push(data[k + OFFSET_NUM] > 1 ? getClusterJSON(data, k, this.clusterProps) : this.points[data[k + OFFSET_ID]]);
|
|
}
|
|
}
|
|
|
|
if (children.length === 0) throw new Error(errorMsg);
|
|
|
|
return children;
|
|
}
|
|
|
|
getLeaves(clusterId, limit, offset) {
|
|
limit = limit || 10;
|
|
offset = offset || 0;
|
|
|
|
const leaves = [];
|
|
this._appendLeaves(leaves, clusterId, limit, offset, 0);
|
|
|
|
return leaves;
|
|
}
|
|
|
|
getTile(z, x, y) {
|
|
const tree = this.trees[this._limitZoom(z)];
|
|
const z2 = Math.pow(2, z);
|
|
const {extent, radius} = this.options;
|
|
const p = radius / extent;
|
|
const top = (y - p) / z2;
|
|
const bottom = (y + 1 + p) / z2;
|
|
|
|
const tile = {
|
|
features: []
|
|
};
|
|
|
|
this._addTileFeatures(
|
|
tree.range((x - p) / z2, top, (x + 1 + p) / z2, bottom),
|
|
tree.data, x, y, z2, tile);
|
|
|
|
if (x === 0) {
|
|
this._addTileFeatures(
|
|
tree.range(1 - p / z2, top, 1, bottom),
|
|
tree.data, z2, y, z2, tile);
|
|
}
|
|
if (x === z2 - 1) {
|
|
this._addTileFeatures(
|
|
tree.range(0, top, p / z2, bottom),
|
|
tree.data, -1, y, z2, tile);
|
|
}
|
|
|
|
return tile.features.length ? tile : null;
|
|
}
|
|
|
|
getClusterExpansionZoom(clusterId) {
|
|
let expansionZoom = this._getOriginZoom(clusterId) - 1;
|
|
while (expansionZoom <= this.options.maxZoom) {
|
|
const children = this.getChildren(clusterId);
|
|
expansionZoom++;
|
|
if (children.length !== 1) break;
|
|
clusterId = children[0].properties.cluster_id;
|
|
}
|
|
return expansionZoom;
|
|
}
|
|
|
|
_appendLeaves(result, clusterId, limit, offset, skipped) {
|
|
const children = this.getChildren(clusterId);
|
|
|
|
for (const child of children) {
|
|
const props = child.properties;
|
|
|
|
if (props && props.cluster) {
|
|
if (skipped + props.point_count <= offset) {
|
|
// skip the whole cluster
|
|
skipped += props.point_count;
|
|
} else {
|
|
// enter the cluster
|
|
skipped = this._appendLeaves(result, props.cluster_id, limit, offset, skipped);
|
|
// exit the cluster
|
|
}
|
|
} else if (skipped < offset) {
|
|
// skip a single point
|
|
skipped++;
|
|
} else {
|
|
// add a single point
|
|
result.push(child);
|
|
}
|
|
if (result.length === limit) break;
|
|
}
|
|
|
|
return skipped;
|
|
}
|
|
|
|
_createTree(data) {
|
|
const tree = new KDBush(data.length / this.stride | 0, this.options.nodeSize, Float32Array);
|
|
for (let i = 0; i < data.length; i += this.stride) tree.add(data[i], data[i + 1]);
|
|
tree.finish();
|
|
tree.data = data;
|
|
return tree;
|
|
}
|
|
|
|
_addTileFeatures(ids, data, x, y, z2, tile) {
|
|
for (const i of ids) {
|
|
const k = i * this.stride;
|
|
const isCluster = data[k + OFFSET_NUM] > 1;
|
|
|
|
let tags, px, py;
|
|
if (isCluster) {
|
|
tags = getClusterProperties(data, k, this.clusterProps);
|
|
px = data[k];
|
|
py = data[k + 1];
|
|
} else {
|
|
const p = this.points[data[k + OFFSET_ID]];
|
|
tags = p.properties;
|
|
const [lng, lat] = p.geometry.coordinates;
|
|
px = lngX(lng);
|
|
py = latY(lat);
|
|
}
|
|
|
|
const f = {
|
|
type: 1,
|
|
geometry: [[
|
|
Math.round(this.options.extent * (px * z2 - x)),
|
|
Math.round(this.options.extent * (py * z2 - y))
|
|
]],
|
|
tags
|
|
};
|
|
|
|
// assign id
|
|
let id;
|
|
if (isCluster || this.options.generateId) {
|
|
// optionally generate id for points
|
|
id = data[k + OFFSET_ID];
|
|
} else {
|
|
// keep id if already assigned
|
|
id = this.points[data[k + OFFSET_ID]].id;
|
|
}
|
|
|
|
if (id !== undefined) f.id = id;
|
|
|
|
tile.features.push(f);
|
|
}
|
|
}
|
|
|
|
_limitZoom(z) {
|
|
return Math.max(this.options.minZoom, Math.min(Math.floor(+z), this.options.maxZoom + 1));
|
|
}
|
|
|
|
_cluster(tree, zoom) {
|
|
const {radius, extent, reduce, minPoints} = this.options;
|
|
const r = radius / (extent * Math.pow(2, zoom));
|
|
const data = tree.data;
|
|
const nextData = [];
|
|
const stride = this.stride;
|
|
|
|
// loop through each point
|
|
for (let i = 0; i < data.length; i += stride) {
|
|
// if we've already visited the point at this zoom level, skip it
|
|
if (data[i + OFFSET_ZOOM] <= zoom) continue;
|
|
data[i + OFFSET_ZOOM] = zoom;
|
|
|
|
// find all nearby points
|
|
const x = data[i];
|
|
const y = data[i + 1];
|
|
const neighborIds = tree.within(data[i], data[i + 1], r);
|
|
|
|
const numPointsOrigin = data[i + OFFSET_NUM];
|
|
let numPoints = numPointsOrigin;
|
|
|
|
// count the number of points in a potential cluster
|
|
for (const neighborId of neighborIds) {
|
|
const k = neighborId * stride;
|
|
// filter out neighbors that are already processed
|
|
if (data[k + OFFSET_ZOOM] > zoom) numPoints += data[k + OFFSET_NUM];
|
|
}
|
|
|
|
// if there were neighbors to merge, and there are enough points to form a cluster
|
|
if (numPoints > numPointsOrigin && numPoints >= minPoints) {
|
|
let wx = x * numPointsOrigin;
|
|
let wy = y * numPointsOrigin;
|
|
|
|
let clusterProperties;
|
|
let clusterPropIndex = -1;
|
|
|
|
// encode both zoom and point index on which the cluster originated -- offset by total length of features
|
|
const id = ((i / stride | 0) << 5) + (zoom + 1) + this.points.length;
|
|
|
|
for (const neighborId of neighborIds) {
|
|
const k = neighborId * stride;
|
|
|
|
if (data[k + OFFSET_ZOOM] <= zoom) continue;
|
|
data[k + OFFSET_ZOOM] = zoom; // save the zoom (so it doesn't get processed twice)
|
|
|
|
const numPoints2 = data[k + OFFSET_NUM];
|
|
wx += data[k] * numPoints2; // accumulate coordinates for calculating weighted center
|
|
wy += data[k + 1] * numPoints2;
|
|
|
|
data[k + OFFSET_PARENT] = id;
|
|
|
|
if (reduce) {
|
|
if (!clusterProperties) {
|
|
clusterProperties = this._map(data, i, true);
|
|
clusterPropIndex = this.clusterProps.length;
|
|
this.clusterProps.push(clusterProperties);
|
|
}
|
|
reduce(clusterProperties, this._map(data, k));
|
|
}
|
|
}
|
|
|
|
data[i + OFFSET_PARENT] = id;
|
|
nextData.push(wx / numPoints, wy / numPoints, Infinity, id, -1, numPoints);
|
|
if (reduce) nextData.push(clusterPropIndex);
|
|
|
|
} else { // left points as unclustered
|
|
for (let j = 0; j < stride; j++) nextData.push(data[i + j]);
|
|
|
|
if (numPoints > 1) {
|
|
for (const neighborId of neighborIds) {
|
|
const k = neighborId * stride;
|
|
if (data[k + OFFSET_ZOOM] <= zoom) continue;
|
|
data[k + OFFSET_ZOOM] = zoom;
|
|
for (let j = 0; j < stride; j++) nextData.push(data[k + j]);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
return nextData;
|
|
}
|
|
|
|
// get index of the point from which the cluster originated
|
|
_getOriginId(clusterId) {
|
|
return (clusterId - this.points.length) >> 5;
|
|
}
|
|
|
|
// get zoom of the point from which the cluster originated
|
|
_getOriginZoom(clusterId) {
|
|
return (clusterId - this.points.length) % 32;
|
|
}
|
|
|
|
_map(data, i, clone) {
|
|
if (data[i + OFFSET_NUM] > 1) {
|
|
const props = this.clusterProps[data[i + OFFSET_PROP]];
|
|
return clone ? Object.assign({}, props) : props;
|
|
}
|
|
const original = this.points[data[i + OFFSET_ID]].properties;
|
|
const result = this.options.map(original);
|
|
return clone && result === original ? Object.assign({}, result) : result;
|
|
}
|
|
}
|
|
|
|
function getClusterJSON(data, i, clusterProps) {
|
|
return {
|
|
type: 'Feature',
|
|
id: data[i + OFFSET_ID],
|
|
properties: getClusterProperties(data, i, clusterProps),
|
|
geometry: {
|
|
type: 'Point',
|
|
coordinates: [xLng(data[i]), yLat(data[i + 1])]
|
|
}
|
|
};
|
|
}
|
|
|
|
function getClusterProperties(data, i, clusterProps) {
|
|
const count = data[i + OFFSET_NUM];
|
|
const abbrev =
|
|
count >= 10000 ? `${Math.round(count / 1000) }k` :
|
|
count >= 1000 ? `${Math.round(count / 100) / 10 }k` : count;
|
|
const propIndex = data[i + OFFSET_PROP];
|
|
const properties = propIndex === -1 ? {} : Object.assign({}, clusterProps[propIndex]);
|
|
return Object.assign(properties, {
|
|
cluster: true,
|
|
cluster_id: data[i + OFFSET_ID],
|
|
point_count: count,
|
|
point_count_abbreviated: abbrev
|
|
});
|
|
}
|
|
|
|
// longitude/latitude to spherical mercator in [0..1] range
|
|
function lngX(lng) {
|
|
return lng / 360 + 0.5;
|
|
}
|
|
function latY(lat) {
|
|
const sin = Math.sin(lat * Math.PI / 180);
|
|
const y = (0.5 - 0.25 * Math.log((1 + sin) / (1 - sin)) / Math.PI);
|
|
return y < 0 ? 0 : y > 1 ? 1 : y;
|
|
}
|
|
|
|
// spherical mercator to longitude/latitude
|
|
function xLng(x) {
|
|
return (x - 0.5) * 360;
|
|
}
|
|
function yLat(y) {
|
|
const y2 = (180 - y * 360) * Math.PI / 180;
|
|
return 360 * Math.atan(Math.exp(y2)) / Math.PI - 90;
|
|
}
|
|
|
|
return Supercluster;
|
|
|
|
}));
|