DataFilterExtension

The DataFilterExtension adds GPU-based data filtering functionalities to layers. It allows the layer to show/hide objects based on user-defined properties. This extension provides a significantly more performant alternative to filtering the data array on the CPU.

Note: This extension does not work with all deck.gl layers. See "limitations" below.

import {GeoJsonLayer} from '@deck.gl/layers';
import {DataFilterExtension} from '@deck.gl/extensions';

const layer = new GeoJsonLayer({
  id: 'geojson-layer',
  data: GEOJSON,

  // props from GeoJsonLayer
  getFillColor: [160, 160, 180],
  getLineColor: [0, 0, 0],
  getLineWidth: 10,

  // props added by DataFilterExtension
  getFilterValue: f => f.properties.timeOfDay,  // in seconds
  filterRange: [43200, 46800],  // 12:00 - 13:00

  // Define extensions
  extensions: [new DataFilterExtension({filterSize: 1})]
});

Installation

To install the dependencies from NPM:

npm install deck.gl
# or
npm install @deck.gl/core @deck.gl/layers @deck.gl/extensions

import {DataFilterExtension} from '@deck.gl/extensions';
new DataFilterExtension({});

To use pre-bundled scripts:

<script src="https://unpkg.com/deck.gl@^9.0.0/dist.min.js"></script>
<!-- or -->
<script src="https://unpkg.com/@deck.gl/core@^9.0.0/dist.min.js"></script>
<script src="https://unpkg.com/@deck.gl/layers@^9.0.0/dist.min.js"></script>
<script src="https://unpkg.com/@deck.gl/extensions@^9.0.0/dist.min.js"></script>

new deck.DataFilterExtension({});

Constructor

new DataFilterExtension({filterSize, fp64});

• filterSize (number) - the size of the filter (number of columns to filter by). The data filter can show/hide data based on 1-4 numeric properties of each object. Set to 0 to disable numeric filtering. Default 1.
• categorySize (number) - the size of the category filter (number of columns to filter by). The category filter can show/hide data based on 1-4 properties of each object. Set to 0 to disable category filtering. Default 0.
• fp64 (boolean) - if true, use 64-bit precision instead of 32-bit. Default false. See the "remarks" section below for use cases and limitations.
• countItems (boolean) - if true, reports the number of filtered objects with the onFilteredItemsChange callback. Default false.

Layer Properties

When added to a layer via the extensions prop, the DataFilterExtension adds the following properties to the layer:

getFilterValue (Accessor<number>|Accessor<number[]>)

Called to retrieve the value for each object that it will be filtered by. Returns either a number (if filterSize: 1) or an array.

For example, consider data in the following format:

[
  {"timestamp": 0.1, "coordinates": [-122.45, 37.78], "speed": 13.3},
  ...
]

To filter by timestamp:

new ScatterplotLayer({
  data,
  getPosition: d => d.coordinates,
  getFilterValue: d => d.timestamp,
  filterRange: [0, 1],
  extensions: [new DataFilterExtension({filterSize: 1})]
})

To filter by both timestamp and speed:

new ScatterplotLayer({
  data,
  getPosition: d => d.coordinates,
  getFilterValue: d => [d.timestamp, d.speed],
  filterRange: [[0, 1], [10, 20]],
  extensions: [new DataFilterExtension({filterSize: 2})]
})

Note that all filtered values are uploaded as 32-bit floating numbers, so certain values e.g. raw unix epoch time may not be accurately represented. You may test the validity of a timestamp by calling Math.fround(t) to check if there would be any loss of precision.

filterRange (number[2] | number[2][])

The bounds which defines whether an object should be rendered. If an object's filtered value is within the bounds, the object will be rendered; otherwise it will be hidden. This prop can be updated on user input or animation with very little cost.

Format:

• If filterSize is 1: [min, max]
• If filterSize is 2 to 4: [[min0, max0], [min1, max1], ...] for each filtered property, respectively.

filterSoftRange (number[2] | number[2][], optional)

• Default: null

If specified, objects will be faded in/out instead of abruptly shown/hidden. When the filtered value is outside of the bounds defined by filterSoftRange but still within the bounds defined by filterRange, the object will be rendered as "faded." See filterTransformSize and filterTransformColor for additional control over this behavior.

new ScatterplotLayer({
  data,
  getPosition: d => d.coordinates,
  getFilterValue: d => d.timestamp,
  filterRange: [0, 1],
  filterSoftRange: [0.2, 0.8],
  filterTransformSize: true,
  filterTransformColor: true,
  extensions: [new DataFilterExtension({filterSize: 1})]
})

Format:

• If filterSize is 1: [softMin, softMax]
• If filterSize is 2 to 4: [[softMin0, softMax0], [softMin1, softMax1], ...] for each filtered property, respectively.

getFilterCategory (Accessor<number | string>|Accessor<(number | string)[]>, optional)

• Default: 0

Called to retrieve the category for each object that it will be filtered by. Returns either a category as a number or string (if categorySize: 1) or an array.

For example, consider data in the following format:

[
  {"industry": "retail", "coordinates": [-122.45, 37.78], "size": 10},
  ...
]

To filter by industry:

new ScatterplotLayer({
  data,
  getPosition: d => d.coordinates,
  getFilterCategory: d => d.industry,
  filterCategories: ['retail', 'health'],
  extensions: [new DataFilterExtension({categorySize: 1})]
})

To filter by both industry and size:

new ScatterplotLayer({
  data,
  getPosition: d => d.coordinates,
  getFilterCategory: d => [d.industry, d.size],
  filterCategories: [['retail', 'health'], [10, 20, 50]],
  extensions: [new DataFilterExtension({categorySize: 2})]
})

filterCategories (string[] | string[][], optional)

• Default: [0]

The list of categories that should be rendered. If an object's filtered category is in the list, the object will be rendered; otherwise it will be hidden. This prop can be updated on user input or animation with very little cost.

Format:

• If categorySize is 1: ['category1', 'category2']
• If categorySize is 2 to 4: [['category1', 'category2', ...], ['category3', ...], ...] for each filtered property, respectively.

The maximum number of supported is determined by the categorySize:

• If categorySize is 1: 128 categories
• If categorySize is 2: 64 categories per dimension
• If categorySize is 3 or 4: 32 categories per dimension

If this value is exceeded any categories beyond the limit will be ignored.

filterTransformSize (boolean, optional)

• Default: true

When an object is "faded", manipulate its size so that it appears smaller or thinner. Only works if filterSoftRange is specified.

filterTransformColor (boolean, optional)

• Default: true

When an object is "faded", manipulate its opacity so that it appears more translucent. Only works if filterSoftRange is specified.

filterEnabled (boolean, optional)

• Default: true

Enable/disable the data filter. If the data filter is disabled, all objects are rendered.

onFilteredItemsChange (Function, optional)

Only used if the countItems option is enabled. Called with the following arguments when the filter changes:

• event (object)
  • id (string) - the id of the source layer. Note when this prop is specified on a CompositeLayer, such as GeoJsonLayer, the callback is called once by each sub layer.
  • count (number) - the number of data objects that pass the filter.

Remarks

Filter precision

By default, both the filter values and the filter range are uploaded to the GPU as 32-bit floats. When using very large filter values, most commonly Epoch timestamps, 32-bit float representation could lead to an error margin of >1 minute. Enabling 64-bit precision by setting fp64: true would allow the filter range to be evaluated more accurately. However, 64-bit support requires one extra attribute slot, which increases the risk of exceeding the hardware limit on vertex attributes. Depending on the layer that the DataFilterExtension is used with, it may interfere with the layer's ability to use other extensions.

If this becomes an issue, an alternative technique is to transform each filter value by subtracting a fixed "origin" value, thus making the numbers smaller:

getFilterValue: d => d.timestamp - ORIGIN_TS,
filterRange: [rangeStart - ORIGIN_TS, rangeEnd - ORIGIN_TS]

32-bit floats can accurately represent each second within ~190 days (2^24). Unless the filter values require both a large span and fine granularity, 32-bit floats should be sufficient.

Limitations

Given data filtering happens on GPU, not all layers of @deck.gl/aggregation-layers module, support this feature.

Always supported:

• HeatMapLayer
• GPUGridLayer

Supported only when aggregation is performed on GPU:

• ScreenGridlayer
• ContourLayer
• GridLayer

Not supported:

• CPUGridLayer
• HexagonLayer

Source

modules/extensions/src/data-filter