API / geotoolkit3d / picking / RendererPicking / RendererPicking

picking.RendererPicking.RendererPicking

Default implementation for the picking renderer mechanism.<br> <br> This class uses rendering as a way to do picking.<br> The idea is similar to the one used here: THREEJS tutorial<br> Instead of rendering color in each pixel, it will render more information coded as a rgba pixel.<br> Then it will read the pixel and extract the information encoded in it.<br> <br> To do so, it will replace the current materials by custom shader-materials which will encode the required information into the pixel.<br> The picking might require several steps to retrieve all the information it needs, for each pass one rgba component will be read from the graphic card.<br> Typically it will execute the picking-rendering 4 passes:<br> <ul> <li>shape-identifier: First pass to find which shapes is at the given position using its identifier</li> <li>x-coordinate: Another pass to retrieve the world x coordinate at the pick position</li> <li>y-coordinate: Another pass to retrieve the world y coordinate at the pick position</li> <li>z-coordinate: Another pass to retrieve the world z coordinate at the pick position</li> </ul> For each pass, the scene will be reconfigured to write the corresponding attribute in place of the regular pixels.<br> One could add more passes to the picking by declaring them in the #getExtraPickingModes() of the picking material.<br> <br> Unlike the Ray-casting approach, the picking occurs on the GPU, which means that:<br> <ul> <li>Vertex&Fragment shader are supported</li> <li>No distance based criteria, the shape rendered at the given pixel coordinates is the one picked</li> <li>Performance is based on frame rendering performance only, however it may require several frames for a single picking operation</li> <li>Returns only the closest object(s) to the camera (the one(s) visible on screen), transparency is ignored. (see filtering)</li> </ul> <br> Note that this picking implementation accepts a width&height parameters to select shapes in a rectangle centered on the given coordinates.<br> A given shape will be returned only once even if it has several pixels in the picked area.<br> The results entry for such case would correspond to the pixel closest to the given picking coordinate.<br> For example if the picked object fills the whole area being picked.<br> Then the picking coordinates returned will match the center of the picked rectangle.<br> <br> This picking also accepts a 'filter' function that can hide objects that are not pickable.<br> As this filtering occurs prior to said GPU picking it may make some objects pickable even if those were initially hidden.<br> For example, if a cube is behind a plane, it's not pickable because it's hidden.<br> But if a filter that discards the plane is given, then the cube can be picked 'through' the plane.<br> This feature is intended to implement picking through non opaque object not as a convenience filter to ignore some shapes.<br> <br>

• AbstractPicking

  ↳ RendererPicking

import {RendererPicking} from '@int/geotoolkit3d/picking/RendererPicking';
import {Line} from 'three';
// Pick the objects in the rectangle of size 5,5 centered at (12,13)
const selection = RendererPicking.pick(plot, 12, 13, 2, 2, (object, pickable) => {
if (!pickable) return false; // If object is already filtered natively, keep filtering it
return !(object instanceof Line); // Prevent line picking, letting user pick through a line
});