v0.2.0

Version 0.2.0 (2025-03-13)

Fixes

• fix: In CDT, update internal function remove_crossed_edges to be faillible, to avoid a possible infinite loop.

Changes

• changed: defaults to using f64 instead of f32 inside the algorithm for more stable results (benchmarks were already using f64). f32 can still be enabled via the f32 cargo feature.

• changed: tracing dependency is now optional. Error messages (and DebugContext when the debug_context feature is enabled) are returned within a TriangulationError.

• changed: glam dependency now accepts version of glam up to 0.29

• changed: now accept slices as inputs instead of Vec in most functions

• vertices format:

  • added: Vertex2d and Vertex3d traits defining the algorithm vertex input format.
  • changed: constrained_triangulation_from_3d_planar_vertices, constrained_triangulation_from_2d_vertices, triangulation_from_3d_planar_vertices, triangulation_from_2d_vertices and transform_to_2d_planar_coordinate_system to be generic over the input vertex format. This does not affect the vertex format used during the algorithm. This allows to decouple the vertex format used in the algorithm (f32 or f64) from the vertex input format.
    • As an example: it is now possible to give vertices with f32 coordinates as an input, but compute the triangulation using f64.
  • removed: Vector3A type. It was not used except for the plane_normal input format, and it was already being converted immediately to a Vector3.
  • removed: the Vector3 type now superseded by the Vertex3d trait.

Optimizations

• improved: find_vertex_placement hot path for edge side tests, and move the vertex merging to the end
• improved: some triangles neigbor checks

Examples, benchmarks & profiling

• changed: the visual debugger in the examples now always uses glam::f32::Vec3 for vertices storage.
• updated: profiling & benchmarks