sandbox VS rapier

I've made a similar game: https://github.com/JMS55/sandbox, and it looks like we've had a lot of the same ideas :)

If you're interested in game development, you may enjoy playing around with Sandbox. It's a game I created where you place sand, water, acid, fire, etc and watch them interact.

> WebGPU is years away to become usable

As a counterpoint, I've been using WebGPU (through wgpu-rs) for the past 1.5 years. It's been a pleasure to use. For instance, here's the CPU-side code for a glow post-process shader using 4 render passes https://github.com/JMS55/sandbox/blob/master/src/glow_post_p....

I've considered implementing Sandbox using wgpu compute sharers (all the rendering is already done with wgpu). The reason I didn't is because I couldn't figure out how to make particles update in parallel - how to handle conflicts between two particles wanting to move into the same position, updating a particle that's supposed to be destroyed, etc. I'd love to get this working however. My last attempt was a the "multithread" branch where I tried to use rayon as a means of prototyping the game using parallel update logic.

Unrelated to a club, but you might enjoy looking at a game I built with Rust + WebGPU https://github.com/JMS55/sandbox.

Counterpoint: https://flathub.org/apps/details/com.github.jms55.Sandbox

Not a GSOC project, but my game Sandbox is open to contributions! https://github.com/JMS55/sandbox

Overflowing the stack: Use a Vec, or Box<[T; N]> (boxed array). A boxed array is a heap allocated array, where the box part just points to some data on the heap, so it's cheap to move around. Both are pretty much equivalent, boxed array is maybe slightly more faster/correct. Note that if you use a boxed array, due to some missing compiler optimizations, you'll probably just get a stack overflow anyways, as the compiler will first store the array on the stack, and then copy it to the heap. This can be solved with some unsafe code, such as in https://github.com/JMS55/sandbox/blob/master/src/heap_array.rs.

Ok, so this uses https://rapier.rs/ which is very cool

Rapier, alongside https://nalgebra.org/ (which it uses underneath) has seriously good documentation and some advanced features like cross-platform determinism (something made hard by the way floating point differs between platforms)

Code examples: see examples2d, examples3d-f64 and examples3d directories

What are the pros and cons compared to something like Rapier? When should one use Bevy XPBD instead of Rapier, or vice versa?

Still using Rust in a browser-based multiplayer party game I'm working on! I'm using Actix Web for the backend and rapier2d to handle my game's physics. I'm looking to make some more connections amongst the developer / gaming community through my game down the line.

I have noticed that Rapier 0.17 contains a RopeJoint struct, which constrains maximum distance between two dynamic bodies. It doesn't seem to have found its way into bevy_rapier3d yet so I haven't understood whether it supports (half of) my use case, and looking at the code (https://github.com/dimforge/rapier/blob/master/src/dynamics/joint/rope_joint.rs) I really cannot see how it is actually implemented and how I could extend it to also set a minimum distance, but this may be a place to start.

I was also frustrated with the kcc, so I opened a PR that fixes most of my issues: https://github.com/dimforge/rapier/pull/446

Likewise, see the js bindings of Rapier.

Maybe check out how Rapier does it. Rapier is likely the most advanced physics engine in Rust. (Though Embark Studios is doing some crazy stuff with ML physics.)

rapier