Picking
Make sure you have read Interactivity before reading this section.
How It Works
The Color Picking Technique
Rather than doing traditional ray-casting or building octrees etc in JavaScript, deck.gl implements picking on the GPU using a technique we refer to as "color picking". When deck.gl needs to determine what is under the mouse (e.g. when the user moves or clicks the pointer over the deck.gl canvas), all pickable layers are rendered into an off-screen buffer, but in a special mode activated by a GLSL uniform. In this mode, the shaders of the core layers render picking colors instead of their normal visual colors.
Each object in each layer gets its own picking color assigned. The picking color is determined using layer.encodePickingColor()
that converts the index of a object of a given layer into a 3-byte color array (the color buffer allows us to distinguish between 16M unique colors per layer, and between 256 different layers).
After the picking buffer is rendered, deck.gl looks at the color of the pixel under the pointer, and decodes it back to the index number using layer.decodePickingColor()
.
Event Propagation
Once an object is picked, deck.gl creates a picking info object that describes what is under the pointer.
The layer.getPickingInfo()
method is called first on the layer that directly rendered the picked object, to modify or add additional fields to the info.
The info object is then passed to the getPickingInfo()
of its parent layer, and then its grandparent, and so on. This is so that composite layers can further augment the info
object after it is processed by the picked sublayer. This allows the composite layer to hide implementation details and expose only user-friendly information.
When the processing chain is over, the event is invoked on the top-level layer. This means that only the top-level layer's on<Event>
callbacks are invoked, and the final picking info's layer
field will point to the top-level layer. The idea is that the user should only interface with the layer that they created, and not having to know the underlying implementation.
When an event fires, the callback functions are executed in the following order:
layer.on<Event>
(default implementation invokesthis.props.on<Event>
)layer.props.on<Event>
(only if the layer method is not defined)deck.props.on<Event>
If any of the callback functions return true
, the event is marked handled and the rest of the callbacks will be skipped.
Default Handling of Pointer Events
Whenever the pointer moves over the canvas, deck.gl performs a new picking pass, yielding a picking info object describing the result. This object is used for multiple purposes:
- The
onHover
callbacks are called with it - To update the picked layer if autoHighlight is enabled
- Saved for later use
When other gestures (click, drag, etc.) are detected, deck.gl does not repeat picking. Instead, their callbacks are called with the last picked info from hover
.
Implementing Custom Picking
While deck.gl allows applications to implement picking however they want, special support is provided for the built-in "picking color" based picking system, which most layers use.
To take full control of picking, a layer need to take the following steps:
Creating A Picking Color Attribute
Add an attribute for each vertex using the layer's AttributeManager:
import GL from '@luma.gl/constants';
class MyLayer extends Layer {
initializeState() {
this.state.attributeManager.add({
customPickingColors: {
size: 3,
type: GL.UNSIGNED_BYTE,
update: this.calculatePickingColors
}
});
}
}
Populate the attribute by providing a different picking color for every object that you need to differentiate. The default implementation of layer.encodePickingColor()
and layer.decodePickingColor()
is likely sufficient, but you may need to implement your own pair.
class MyLayer extends Layer {
...
calculatePickingColors(attribute) {
const {data} = this.props;
const {value} = attribute;
let i = 0;
for (const object of data) {
const pickingColor = this.encodePickingColor(i);
value[i * 3] = pickingColor[0];
value[i * 3 + 1] = pickingColor[1];
value[i * 3 + 2] = pickingColor[2];
i++;
}
}
}
By default, the object
field of the picking info
object is indexed from the layer's data
prop. Custom layers often need to define on their own terms what constitutes meaningful information to the user's callbacks. A layer can achieve this by overriding layer.getPickingInfo()
to add or modify fields to the info
object.
Model object creation
If your layer creates its own Model object, add picking module to modules
array.
import {Model} from '@luma.gl/core';
import {picking} from '@deck.gl/core';
new Model(gl, {
...
vs: CUSTOM_VS,
fs: CUSTOM_FS,
modules: [picking]
});
Implementing Picking in Custom Shaders
All core layers (including composite layers) support picking using luma.gl's picking module
. If you are using custom shaders with any of the core layers or building custom layers with your own shaders, the following steps are needed to enable picking:
Vertex Shader
Vertex shader should set current picking color using picking_setPickingColor
method provided by picking shader module.
attribute vec3 customPickingColors;
void main(void) {
...
picking_setPickingColor(customPickingColors);
}
Fragment Shader
Fragment shader should use picking_filterPickingColor
to update gl_FragColor
, which outputs picking color if it is the picking pass.
void main(void) {
...
// Should be the last Fragment shader instruction that updates gl_FragColor
gl_FragColor = picking_filterPickingColor(gl_FragColor);
}
For more details refer to luma.gl's Picking Module
.