mirror of
https://github.com/bevyengine/bevy
synced 2024-11-26 22:50:19 +00:00
c62ad4b2c4
# Objective Closes #10570. #10946 added bounding volume types and traits, but didn't use them for anything yet. This PR implements `Bounded2d` and `Bounded3d` for Bevy's primitive shapes. ## Solution Implement `Bounded2d` and `Bounded3d` for primitive shapes. This allows computing AABBs and bounding circles/spheres for them. For most shapes, there are several ways of implementing bounding volumes. I took inspiration from [Parry's bounding volumes](https://github.com/dimforge/parry/tree/master/src/bounding_volume), [Inigo Quilez](http://iquilezles.org/articles/diskbbox/), and figured out the rest myself using geometry. I tried to comment all slightly non-trivial or unclear math to make it understandable. Parry uses support mapping (finding the farthest point in some direction for convex shapes) for some AABBs like cones, cylinders, and line segments. This involves several quat operations and normalizations, so I opted for the simpler and more efficient geometric approaches shown in [Quilez's article](http://iquilezles.org/articles/diskbbox/). Below you can see some of the bounding volumes working in 2D and 3D. Note that I can't conveniently add these examples yet because they use primitive shape meshing, which is still WIP. https://github.com/bevyengine/bevy/assets/57632562/4465cbc6-285b-4c71-b62d-a2b3ee16f8b4 https://github.com/bevyengine/bevy/assets/57632562/94b4ac84-a092-46d7-b438-ce2e971496a4 --- ## Changelog - Implemented `Bounded2d`/`Bounded3d` for primitive shapes - Added `from_point_cloud` method for bounding volumes (used by many bounding implementations) - Added `point_cloud_2d/3d_center` and `rotate_vec2` utility functions - Added `RegularPolygon::vertices` method (used in regular polygon AABB construction) - Added `Triangle::circumcenter` method (used in triangle bounding circle construction) - Added bounding circle/sphere creation from AABBs and vice versa ## Extra Do we want to implement `Bounded2d` for some "3D-ish" shapes too? For example, capsules are sort of dimension-agnostic and useful for 2D, so I think that would be good to implement. But a cylinder in 2D is just a rectangle, and a cone is a triangle, so they wouldn't make as much sense to me. A conical frustum would be an isosceles trapezoid, which could be useful, but I'm not sure if computing the 2D AABB of a 3D frustum makes semantic sense. |
||
---|---|---|
.. | ||
src | ||
Cargo.toml |