I have a complex web application, with several sub-views / widgets that add objects to the main scene.
Most simple THREE.js examples have a continous render loop, that re-renders the scene periodically.
This seems like a waste of ressources to me.
Therfore, I only want to re-render the scene if something has changed.
However, the sub-views / widgets do change things (like position of objects) in the scene on their own.
Is there something like a "scene.hasChanged" flag, that I can use to deceide if the scene needs to be rerendered? Or do I have to implement a parallel flag or notification structure myself?
Or is there another common solution to this?
Based on what you have written so far, it seems you have to implement a custom solution. Certain three.js examples like webgl_loader_gltf do not use an animation loop but utilize change events from the controls. The relevant line of code is
controls.addEventListener( 'change', render );
However, if more factors than just the user interaction trigger renderings, it's necessary to build some sort of notification/event system by yourself.
three.js R113
Related
I'm trying to run an animation based on scroll in the background of a React app.
I use Threejs with React Three Fiber. In R3F, there is a hook called useScroll which I want to use.
Problem is that my Threejs scene (a Canvas) is in the background with zIndex: -100, so the scroll handler doesn't get the user input.
If I remove the zIndex property, everything works, but It's not what I want.
The scene look like this demo, the only difference is that mine has element in front.
I don't know if the solution is to use a different css style or something more complexe to allow the handler to do his job. I've been looking at React Portal but I don't think it will work.
Any ideas ?
Thanks in advance.
I am having a human body object and what I want is that my code should open a popup whenever I click on any of the specified body parts such as eyes, nose, arms, etc. The object is a single compiled .obj file and I'm unable to figure out as to how should I attach an event listener to the multiple body parts. Any kind of help is much appreciated!
PS: - I am also using orbit controls to handle the zoom and rotation of object.
Thanks in advance..
To solve this issue, it's necessary to design your model in a way such that individual components like arms, eyes etc. are independent 3D objects grouped together to build a more complex asset. This is something you have to ensure during the design phase with a tool like Blender.
In the next step export to glTF instead of OBJ, load the model with THREE.GLTFLoader into your app and then perform a recursive raycasting with the entire asset via:
raycaster.intersectObject( gltf.scene, true, intersects );
By executing this code in a pointermove or pointerdown event listener you can find out whether the user interacts with the model or not.
Notice that three.js has no system to directly assign event listeners to 3D objects. You have to use a different solution like raycasting for this.
I have a scene that is mostly static, i.e. without user interaction it doesn't change (animations).
The user interacts with the scene. For example:
when the user clicks on certain location in a map, another image opens in another pane
the user may move the camera by moving the mouse
etc...
Originally I implemented the rendering via animate(), which calls:
function animate() {
requestAnimationFrame( animate );
render();
controls.update();
}
This works fine (in terms of response) except that
the rendering happens constantly, which is not necessary.
for one of the admin pages, which does not show the scene, I'm getting runtime errors when the scene does not exist, but the rendering still goes on...
So I re-implemented the rendering by calling, explicitly, after user action which change the scene
render();
controls.update();
This requires more work, because I have to cover all the cases where the scene changes.
In one of the cases, the behaviour is not as expected (the rendered scene is black and explicit call to render does not fix the problem).
The link here says that if your scene is static, there is no reason for an animation loop.
The link here suggests to use animate().
My questions:
Is explicit render, a good pattern to use in three js for static scenes?
Are there pros to render only when needed, instead of calling animate which renders constantly, when this is not needed?
In terms of resource load in general
The link here says that if your scene is static, there is no reason for an animation loop.
The answer of the respective stackoverflow question is also the correct one. The second link is unrelated to question whether to use on-demand rendering or not.
In general, having an animation loop requires more resources since you constantly render no matter if the scene changes or not. However, certain application like 3D viewers (without animation playback) do not need to do this. It's sufficient to render only when certain events happen (e.g. user interaction).
Is explicit render, a good pattern to use in three js for static scenes?
Yes, it is. It is a good pattern for 3D applications in general.
Are there pros to render only when needed, instead of calling animate which renders constantly, when this is not needed? In terms of resource load in general
The app requires less resources which is always a good thing. E.g. mobile device will save a lot of battery and laptops or desktops will be more quiet.
May I have a 2D layer for UI, Text, Buttons, etc over the 3D scene in ThreeJS?
Ideally something like engine from PixiJS inside ThreeJS? I've seen PixiJS offers some 3D features so why not combine both libraries in something super-powerful? I just do not want to place any HTML Dom elements over WebGL canvas as this will probably slow down performance on Mobile devices.
One way to solve this issue is to implement the UI as screen space sprites like demonstrated in the following official example (check out how the red sprites are rendered):
https://threejs.org/examples/webgl_sprites
The idea is to render them with a separate orthographic camera and an additional call of WebGLRenderer.render(). Besides, instances of THREE.Sprite do support raycasting which is of course useful when implementing interaction.
Building up on Mugen87's answer, you can also use THREE.Shape to make visual containers adapted to the user screen size :
https://threejs.org/docs/#api/en/extras/core/Shape
You can use THREE.Shape to make mesh-based text, is illustrated in this example :
https://threejs.org/examples/?q=text#webgl_geometry_text_shapes
You should also have a look at three-mesh-ui, an add-on for building mesh-based user interface with three.js :
https://github.com/felixmariotto/three-mesh-ui
I'm playing about with three.js and webGL with an attempt to get html objects overlaid on a webGL model.
I intend to use three.js object loaders to render a 3D model, where I can then allow users to overlay HTML DOM objects in 3D space via a custom UI.
I've mainly been using the various different examples in the three.js docs (https://threejs.org/examples/ and https://codepen.io/AdamEDGE/pen/RRjEwz).
In order to allow the user to move the overlaid HTML objects via the UI, I've added the transform controls (https://threejs.org/examples/misc_controls_transform.html).
Things seemed to be going relatively well up to this point, but now I see that the controls are not hoverable or clickable to move the element.
I suspect I'm either missing code or an issue with the camera not aligning correctly, but I'm very new to this so I'm guessing.
I have also experimented with Jerome Etienne's blending tutorials, but on testing that caused other issues and I think may not be worth the trouble for what I'm going for (http://learningthreejs.com/blog/2013/04/30/closing-the-gap-between-html-and-webgl/index.html).
I have created an example that shows the problem (click on an object and the controls should show).
Its a simple webGL torus with two HTML images overlaid and mapped to a Mesh so I can link the controls to the HTML elements which are on a different scene.
https://jsfiddle.net/mattscotty/h5wuq86y/
control = new THREE.TransformControls(camera, glRenderer.domElement);
control.addEventListener('change', render);
The controls show on all the correct places when clicked as expected and I can use the keyboard to change the transform type (again see https://threejs.org/examples/misc_controls_transform.html for how it should work).
As I can't click or hover, I can't hook up to the relevant events.
I'm brand new to this and I'm quite sure I'm missing something obvious, so any help is appreciated.