detria - a Delaunay triangulation library

Features

• Delaunay triangulation of a point set
• Constrained delaunay triangulation
• Support for multiple outlines and holes
• Support for Steiner points
• Geometric robustness - results will always be exact, no errors because of floating-point inaccuracy
• Properly handle edge cases (e.g. collinear or cocircular points)
• Allows custom user types to be used (e.g. custom point type, custom allocator, etc.)
• Single file library

Examples

Delaunay triangulation of a point set:
Constrained delaunay triangulation:
Interior triangles only:
With Steiner points added:
Note that the Steiner points are not auto-generated, they must be added manually.

Holes inside holes:

Requirements

Requires C++17 or later.
No external dependencies are used, only the C++ standard library.

Usage

Just include detria.hpp in your project.
Basic code example:

// Create a square, and triangulate it

// List of points (positions)
std::vector<detria::PointD> points =
{
    { 0.0, 0.0 },
    { 1.0, 0.0 },
    { 1.0, 1.0 },
    { 0.0, 1.0 }
};

// List of point indices
std::vector<uint32_t> outline = { 0, 1, 2, 3 };

bool delaunay = true;

detria::Triangulation tri;
tri.setPoints(points);
tri.addOutline(outline);

bool success = tri.triangulate(delaunay);

if (success)
{
    bool cwTriangles = true;

    tri.forEachTriangle([&](detria::Triangle<uint32_t> triangle)
    {
        // `triangle` contains the point indices

        detria::PointD firstPointOfTriangle = points[triangle.x];
        detria::PointD secondPointOfTriangle = points[triangle.y];
        detria::PointD thirdPointOfTriangle = points[triangle.z];
    }, cwTriangles);
}

Documentation

Geometric robustness

This library uses the Fast Robust Predicates for Computational Geometry library.
Placed in the public domain by Jonathan Richard Shewchuk.

All floating point calculations are handled by either:

• Direct floating point comparisons (e.g. a < b, a == b), which are always exact
• Calls to orient2d from the Robust Predicates library, which exactly decides the orientation of three points (clockwise, counter-clockwise, or collinear)
• In delaunay triangulations, the incircle function is used to decide if a circle (described by three points) contains a fourth point, or doesn't (or if all four points are cocircular)

Future improvements

These might be added in the future:

• User-defined edge flip conditions (e.g. flip based on edge lengths, triangle aspect ratios, etc.)
• Performance improvements

Non-goals

There are currently no plans to implement the following:

• Adding new points to the triangulation (e.g. conforming delaunay triangulation, triangulation refinement, resolve edge intersections)

Benchmarks

See performance comparisons with other triangulation libraries here.

License

Licensed under either the WTFPL or the MIT License, at your choice.