[QGIS-Developer] QGIS 3D: Move away from Qt 3D?

[David Koňařík]

On 2/12/26 23:34, Even Rouault via QGIS-Developer wrote:
> Hi,
>> High level 3D engines are certainly something to consider as well, and 
>> I have evaluated a couple of them. Overall my impression was that 
>> those are generally massive pieces of software with lots of new 
>> dependencies, that are generally not built for being embedded in 
>> existing applications. (e.g. Godot, O3DE) I would be happy to hear any 
>> recommendations for higher level 3D engines that would be easy to embed!
> 
> Not necessarily a recommendation, but there's also Ogre3D (https:// 
> www.ogre3d.org/). In a project in a previous life (~15 years ago) we 
> used it for flight preview.  If I remember well, there was no terrain 
> tile loader shipped with the engine, but we plugged one. We didn't have 
> lots of objects, a few OBJs, so not sure how that would behave with big 
> models. At the time we used it through a GTK integration, but there's 
> apparently a Qt one: https://ogrecave.github.io/ogre/api/latest/ 
> class_ogre_bites_1_1_application_context_qt.html
Hi, you're right that 3D engines in a library form do exist, but sadly 
as far as I know they are mostly rather old. Ogre3D, Irrlicht, 
OpenSceneGraph are all products of the OpenGL 2 era, but nowadays our 
targets should be Vulkan and Metal, with very different paradigms.

There is OGRE-Next and VulkanSceneGraph (which even has some GIS-esque 
examples). I can't say right now how well they'd integrate with our 
existing Qt-heavy codebase or how much of their code we'd use, though. 
That's always a question when including a new C++ library.

Personally I think writing our own 3D view on top of a low-level API 
like QRhi or WebGPU is preferable to using a library that's not a good 
fit, but maybe there already exists one that would be a right for us, I 
just haven't seen it yet.

David Koňařík