Preparing documentation for 2.1 release

README.md

@@ -11,6 +11,7 @@ Delaunay triangulations for the rust ecosystem.
 - Allows both incremental and bulk loading creation of triangulations
 - Support for vertex removal
 - 2D constrained Delaunay triangulation (CDT)
+- [Delaunay refinement](https://en.wikipedia.org/wiki/Delaunay_refinement)
 - Uses precise geometric predicates to prevent incorrect geometries due to rounding issues
 - Supports extracting the Voronoi diagram
 
@@ -32,7 +33,6 @@ Project goals, in the order of their importance:
 
 For Spade 2.x:
  - Add back the removed interpolation methods (natural neighbor interpolation, #67)
- - Delaunay refinement (#66)
 
 For Spade 3:
  - Possibly base `spade` on `nalgebra` as underlying vector and matrix library. Not much else planned yet!

src/delaunay_core/refinement.rs

@@ -473,10 +473,19 @@ where
  /// Use [RefinementResult::refinement_complete] to identify if a refinement operation has succeeded without running out of
  /// vertices.
  ///
- /// To prevent this from happening, consider either lowering the minimum angle limit
+ /// For mitigation, consider either lowering the minimum angle limit
  /// (see [RefinementParameters::with_angle_limit]) or introduce a
  /// [minimum required area](RefinementParameters::with_min_required_area).
  ///
+ /// Meshes with very small input angles (angles between two constraint edges) may lead to poorly refined results.
+ /// Please consider providing a bug report if you encounter an input mesh which you think isn't refined well.
+ ///
+ /// # Stability guarantees
+ ///
+ /// While changing the interface of this method is considered to be a breaking change, changes to the specific
+ /// refinement process (e.g. which faces are split in which order) are not. Any patch release may change how
+ /// the same input mesh is being refined.
+ ///
  /// # References
  ///
  /// This is an adaption of the classical refinement algorithms introduced by Jim Ruppert and Paul Chew.