Equivalent of 3D Collider.Raycast

The is the very first time I've had this request and it does seem like a very niech call too TBH. 2D physics uses the call on that collider to remove the need to calculate its position as the starting point with the option of ignoring that collider. This, early on, was a widely requested feature often used for projecting rays from a character to its environment ignoring the collider itself.

That's not to say it cannot be added but I do think calling i If I were to add it, I'd add it to the ContactFilter2D so that all 2D physics queries would get the ability to include or exclude a specific Collider2D and not just a specific overload of a Raycast.

In the end, 3D physics isn't a golden standard API so not everything is copied but instead, if there's enough requests or we can see a wider utility in implementing something that also satisfies other needs then it'll likely get added.

I cannot think of a simple way to emulate that feature though TBH. I mean there's ClosestPoint but if you're more interested in that specific direction then it's obviously not going to help. Only option is to filter by layers to reduce the potentials and then look for that collider.

 

It's easy to forgot that your raycast isn't just going to be against a single primitive shape collider such as circle, capsule etc. Box2D has TestPoint and Raycast against primitive shapes only. Take the CompositeCollider2D, TilemapCollider2D or the new CustomCollider2D; those might be composed of tens of thousands of primitives so performing raycast against each primitive shape would take a lot of time and Unity colliders don't have their own "scene graph". Each primitive shape is in the Box2D broadphase and we hook into the low-level algorithm that navigates the broadphase so it's super low-level and very fast. It doesn't get all the hits then perform some post-processing step to filter them out. This is far faster than anything you could do in script returning all results, getting them in a list, searching for a specific collider etc. These results wouldn't go beyond seeing a pointer to a primitive shape, seeing if has a pointer to the Unity collider then accepting it or discounting it.

The ContactFilter2D when set to filter a specific Collider2D would encounter primitives shapes in the broadphase and convert the pointer we have to the Unity collider and check it then ignore it if required. The additional benefit as I've said is that it'll natural work on every single query and not just raycast.

 

I would add that this view is quite skewed towards saying bad performance when in reality it's not a case of doing this against a single collider or every collider but instead it's far more common to filter by "ground" i.e. a layer approach. I know you know this but I wanted that to be clear for anyone else reading this too.

This isn't to say at all that I'm pushing against your idea; quite the opposite in-fact!

 

I completely agree (lots of nodding going on here) but it's something that would be independently 2D physics and/or 3D physics too. I can tell you now that we've got a road-map for 2D physics that includes designs for looking how to provide a different approach to defining interactions moving away from layers which are, as you suggest, an overloaded concept. The "ball and chain" in this design is the overwhelming amount of existing practices and materials surrounding the layer-based approach which means we must be able to switch between models or, as in our design, allow them to compliment each other.