Better Physics Layers

Forum Thread | Discord | Email | Demo | Documentation | API Reference

Better Physics Layers gives you fine-grained and easy-to-use control over which colliders can collide.

✅ Features:
----------------
↣ Up to 64 layers, totally separate from Unity's built-in layer system
↣ Separate physics layers from rendering layers
↣ Easy to use - Just specify which layer (or layers!) the object belongs to, without worrying about bitmasks
↣ Fine-grained control - Objects can belong to any combination of layers, and can collide with any combination of layers
↣ Mix and match - Any object without a PhysicsMask component will behave exactly like normal and won't be affected
↣ Fast - No measurable performance hit even with thousands of rigidbodies*
↣ Full source included, so you can inspect and modify to your heart's content

❔ Why?
----------------
Normally, whether or not two colliders can collide depends on the GameObject's layer and the physics layer matrix. This is a problem for two main reasons:

- There are only 32 layers, of which 8 are reserved, and these layers also affect rendering (among many other things!)
- Objects can only belong to one layer at a time

Better Physics Layers solves this by making good use of Unity's IgnoreCollision functions. This is extremely fast and doesn't hurt performance.

I created this system because I needed it. While working on my portal system I discovered that it was extremely limiting not being able to have an object belong to more than one physics layer at once, and I quickly used up all 24 of Unity's customisable built-in layers.

❗ *Limitations:
----------------
- Raycasts/spherecasts/etc. can't easily be filtered within the physics engine except by Unity's built-in layers, so there's an included function to filter the results of the cast instead.
- There's a small performance cost to updating the collision ignores, simply due to sheer number of loops required. On my computer this is ~0.2ms in the editor with 100 objects, or ~3ms in the editor with 1000 objects. The cost is O(n^2). The collision ignores only need to be updated when changing layers or adding new objects, so in most cases this performance cost is negligible.
- Currently only supports 3D physics, but I want to add 2D support in future.