Hyperreal

[timschmidt]

High-level PR Summary

This PR documents the porting strategy for migrating the csgrs geometry stack to use the hyperreal numeric backend and improves error handling across the codebase. The porting plan establishes branching discipline, ownership boundaries between crates (hyperreal, hyperlattice, hyperlimit, hypersolve), and integration of symbolic variables with infinitesimal perturbation machinery for robust geometric predicates. On the implementation side, error types are enhanced with thiserror derives and detailed variants, several methods that previously returned Option or unwrapped results now properly return Result types, and tests are updated to handle the new error-aware APIs.

⏱️ Estimated Review Time: 15-30 minutes

💡 Review Order Suggestion

Order	File Path
1	PORTING_PLAN.md
2	src/errors.rs
3	src/mesh/mod.rs
4	src/mesh/shapes.rs
5	src/sketch/extrudes.rs
6	src/polygon.rs
7	src/tests.rs
8	src/wasm/mesh_js.rs
9	src/io/stl.rs
10	src/csg.rs
11	src/main.rs

[gemini-code-assist]

Code Review

This pull request updates the porting plan documentation and significantly improves error handling by transitioning several Mesh methods to return Result<T, ValidationError> instead of panicking or returning Option. It also includes performance optimizations in triangle subdivision and extrusion logic. A review comment highlights that the changes to test_difference_metadata mask an underlying inconsistency in how CSG operations handle metadata propagation between optimized and fallback paths, suggesting that the test remain strict to ensure consistent behavior.

[gemini-code-assist]

The assertion in test_difference_metadata has been weakened to allow metadata from either operand (Cube1 or Cube2). In a CSG difference operation ($A - B$), the resulting polygons should ideally inherit the metadata of the primary object ($A$), including the surfaces created by the cut. The current implementation of difference in src/mesh/mod.rs appears inconsistent: the optimized path (mesh-bbopt) correctly retags polygons from the second operand, while the fallback path does not. This test change masks that inconsistency. It is recommended to keep the test strict and ensure metadata propagation is consistent across all implementation paths.