Post 3.0 Roadmap

[starseeker]

@elalish now that 3.0 has been released, what are the next priorities?

[elalish]

Well, I suppose I should turn that question around - what would you like to contribute? I'm certainly hoping we can stay away from breaking changes for awhile now. There are lots of things I'd like to do, but my time is pretty constrained. As usual, bugs get top priority. This is my personal list:

• better handling of tolerance for simplification #1017/Better faces #1040 are my current focus. I'm hoping that turns into a solid generic mesh decimator.
• Potential boolean bug #970/Triangulation failure with TBB enabled in SimpleOffset test #1006/Add MinkowskiSum and MinkowskiDifference #666; it seems like offsetting is a good edge case test that's catching some Boolean issue. Need to investigate.
• Refactor to avoid using SparseIndices #972 I think could bring a significant perf improvement and reduce lines of code. I'm probably the only person with the background for this particular refactor, as it will be large.
• Overlap removal #289 is the big new algorithm I want to write that would open up a lot of use cases. I'll need to find a serious chunk of time to tackle that.
• I wrote a novel straight skeleton algorithm years ago I've been meaning to turn into C++: it turns a polygon into a manifold 3D "roof". If anyone thinks that's useful, I might prioritize it higher.

I'm curious what @pca006132 wants to tackle - I bet it involves perf improvements.

What does everyone else want? What might you like to contribute?

• There's much talk of Rust bindings, e.g. Rust bindings #1046.
• It'd be nice to release the worker of ManifoldCAD.org as its own npm package.
• It would be especially cool to then use that package to make a VSCode extension that's effectively ManifoldCAD.org, but with file system support. Especially if that made it easy to import and export design libraries like OpenSCAD does. That would be an awesome project for someone who wants to help, but not get into the nitty-gritty of geometric algorithms.
• I wrote another glTF extension (just a draft) to encode our smoothing tangents. It would be cool to make a lighter-weight JS decoder for that to demonstrate the benefits of local tessellation. Anyone interested?

[pca006132]

My plan:

1. Add debug code to dump the boolean operands that produce an error in csg_tree, to help debugging things like Potential boolean bug #970 and Triangulation failure with TBB enabled in SimpleOffset test #1006.
2. Fix accidental quadratic runtime in csg tree #989, I have some idea about how to fix it, but it will be a big refactor for csg_tree, due to c++ having no support for tail recursion...
3. Look into Further collider optimizations #982, I think we have some low-hanging fruits there, but that may require some refactoring. And there are potential algorithmic improvements as well.
4. Compile manifold to wasm without emscripten. I think it will take a long time until the rust folks add wasm-unknown-emscripten to wasm-bindgen, so it will probably be easier to support compiling manifold to wasm-unknown-unknown directly. Requires some hacking, but I think this will be a fun one.