Point Clustering - Kinetica Docs

Example

An example of point clustering is available below:

Point Cluster

You will need a current Mapbox API key and some minor modifications will be required in order to run each example. At a minimum, you will need a Kinetica table with latitude/longitude values.

Visualizing Point Clusters

This example demonstrates the power of Kinetica to aggregate millions of records into granular clusters (thousands as a result) and send them to the browser. It then leverages the power of Mapbox and Supercluster to further aggregate these clusters based on zoom scale and visualize them on the map. The result is a high-performance and impressive display of clustering on a massive scale, which allows for the visualization of millions of points in a digestible way. Furthermore, each cluster can be clicked-on to reveal statistics about the cluster itself, including how many individual points are contained and the smallest & largest clusters in the current grouping.

Initializing the Map

First, ensure Kinetica Kickbox.js and its stylesheet are included in the HTML page. If you’re using npm to install Kickbox, add this JavaScript include between the final DOM element and the closing </body> tag:

<script src="/node_modules/kinetica-kickbox-mapbox-gl/dist/kinetica-kickbox-mapbox-gl.min.js"></script>

The style sheet should be included in the <head> of the HTML:

<link rel="stylesheet" href="/node_modules/kinetica-kickbox-mapbox-gl/dist/kinetica-kickbox-mapbox-gl.min.css" />

The map will have to be initialized before a Kinetica WMS layer can be added to it. Map initialization via API Documentation is promisified, so that initialization is guaranteed to complete before any layers are added, avoiding startup errors. Since this initialization function sets up basic authentication for Kinetica and connects to the Mapbox API all at once, it is intended to replace the existing map initialization function provided by Mapbox.

kickbox.initMap({
    mapboxgl: <mapboxgl_object>,
    kineticaUrl: 'http://<db.host>:9191',
    wmsUrl: 'http://<db.host>:9191/wms',
    mapboxKey: '<MapboxKey>',
    mapDiv: '<map_html_tag_name>',
    username: '<kinetica_username>',
    password: '<kinetica_password>'
}).then(function(map){...}

The code above initializes the Mapbox map, sets the parameters for basic authentication inside of Mapbox, and also sets the basic authentication parameters for any XmlHttpRequests made to Kinetica itself. The function returns the standard Mapbox map. For an example call, see API Documentation.

Adding a Point Cluster Layer

After initialization, a point cluster layer can be added to the map. Performing this manually would require creating & adding a source, layer, & supercluster object; requesting & processing the aggregations from Kinetica; then piecing together the WMS URL with all of the configuration & rendering parameters, and finally, binding every zoomend and moveend event to reconstruct the WMS URL during interaction with the map. Fortunately, this process can be simplified with the API Documentation function:

kickbox.addClusterLayer(
    map,
    {
        layerId: '<layer_id>',
        mapboxgl: window.mapboxgl,
        kineticaUrl: 'http://<db.host>:9191',
        tableName: '<table_name>',
        xAttr: '<x_column_name>',
        yAttr: '<y_column_name>',
        geohashAttr: '<geohash_column_name>',
        precision: <geohash_precision>,
        clusterRadius: <cluster_radius_in_pixels>,
        useBbox: false,
        renderingOptions:
        {
            // Customize output here
        }
    }
);

That is all the code required to visualize data in Kinetica on a Mapbox map, using point clustering! By customizing the renderingOptions parameters, the way the points are rendered can be altered, including size, color, and shape. See the WMS endpoint documentation for the full set of rendering options to customize output.

A WKT column is not currently allowed when creating clusters—x/y columns must be used.

For an example call, see API Documentation.

Full Code Example

The kbConfig object should be updated with the settings for the target Kinetica environment for this example to run. Note that the Kickbox files, kickbox.min.js & kickbox.css are assumed to be local to this HTML file.

<!DOCTYPE html>
<html>
    <head>
        <meta charset='utf-8' />
        <title>Kinetica Kickbox: Point Cluster Example</title>
        <meta name='viewport' content='initial-scale=1,maximum-scale=1,user-scalable=no' />

        <!-- Include Mapbox stylesheet -->
        <link href='https://api.tiles.mapbox.com/mapbox-gl-js/v0.41.0/mapbox-gl.css' rel='stylesheet' />

        <!-- Include Kickbox CSS -->
        <link rel="stylesheet" href='kickbox.css' />

        <!-- Configure Your Kinetica Config Here. Exposes a variable called "kbConfig" -->
        <script src="./config.js"></script>

        <!-- Basic styling for the map -->
        <style>
            body { margin:0; padding:0;}
            #map { position:absolute; top:0; bottom:0; width:100%; transition: all 0.3s; }
        </style>
    </head>
    <body>
        <div id='map'></div>
        <!-- Include Mapbox and Kickbox -->
        <script src='https://api.tiles.mapbox.com/mapbox-gl-js/v0.41.0/mapbox-gl.js'></script>
        <script src='kickbox.min.js'></script>

        <script>
            (function(mapboxgl) {

                // Initialize the map
                var kbConfig = {
                    mapboxKey: 'your_mapbox_api_key',
                    kineticaUrl: 'http://<db.host>:9191',
                    wmsUrl: 'http://<db.host>:9191/wms',
                    // If using basic authentication
                    // username: '',
                    // password: '',
                    tableName: 'your_table_name',
                    xColumnName: 'your_x_column_name',
                    yColumnName: 'your_y_column_name',
                    geohashColumnName: 'your_geohash_column_name',
                    center: [-74.2598555, 40.6971494],
                    zoom: 8
                };

                var layerId = tableName + '-cluster';

                kickbox.initMap({
                    mapDiv: 'map',
                    mapboxgl: window.mapboxgl,
                    mapboxKey: kbConfig.mapboxKey,
                    kineticaUrl: kbConfig.kineticaUrl,
                    wmsUrl: kbConfig.wmsUrl,
                    // If using basic authentication
                    // username: kbConfig.username,
                    // password: kbConfig.password,
                    zoom: kbConfig.zoom
                }).then(function(map) {

                    // Add a cluster layer to the map
                    kickbox.addClusterLayer(map, {
                        layerId: layerId,
                        mapboxgl: window.mapboxgl,
                        kineticaUrl: kbConfig.kineticaUrl,
                        tableName: kbConfig.tableName,
                        xAttr: kbConfig.xColumnName,
                        yAttr: kbConfig.yColumnName,
                        geohashAttr: kbConfig.geohashColumnName,
                        precision: 9,
                        clusterRadius: 40,
                        clusterMaxZoom: 24,
                        useBbox: false
                    });
                });

            })(window.mapboxgl);
        </script>
    </body>
</html>

A Note About Data Design

In order to visualize point clusters on a map, a geohashed column of coordinates in the table is required. A detailed discussion of geohashing is outside the scope of this documentation, but briefly, it is the encoding of coordinate data into strings of a predetermined precision. These strings can then be run through a GROUP BY query using a substring with a passed precision to generate the atomic clusters used in the example. They are referred to as atomic clusters, because the example has no atomic records underlying the visualization (in the browser, that is). That would produce far too many results for the browser to handle. Instead, the Supercluster is given the set of aggregated-by-geohash atomic clusters, which it then further aggregates as the user zooms in or out. In essence, this provides a way to visualize clusters of clusters in real-time, where the basic level of clustering delivered by Kinetica is chosen based on experimentation and performance requirements. Kinetica does have functions to encode & decode geohashes, under the Geometry Functions section of the Geospatial Function Reference. To generate a geohash column, in SQL, from a given source table containing x & y columns:

CREATE OR REPLACE TABLE table_with_coordinates_and_geohashes AS
SELECT
    lon,
    lat,
    GEOHASH_ENCODE(lon, lat, <precision>) geohash,
    ...
FROM table_with_coordinates

Replace <precision> with the number of geohash characters to generate for each string representation of the coordinate pairs.

This table can also be created natively with the /create/projection endpoint

​Example

​Visualizing Point Clusters

​Initializing the Map

​Adding a Point Cluster Layer

​Full Code Example

​A Note About Data Design

Example

Visualizing Point Clusters

Initializing the Map

Adding a Point Cluster Layer

Full Code Example

A Note About Data Design