2D triangulation library. Allows translating lines and polygons (both based on points) to the language of GPUs.

Features normal and miter joint line triangulation. Handles polygons using ear clipping algorithm with hole elimination included.

For reference: triangulates 99.76% of 75 thousand buildings in Cracow under 3.43s on average programmer notebook (single threaded).

• Installation
• Usage
• Examples
• Features
  • API
  • Types
• Tests
• Benchmarks
  • Flame Graphs
• Future plans
• Contributing
• License

Nothing surprising

go get github.com/tchayen/triangolatte

vertices := []Point{{10, 20}, {30, 40}, {50, 60}}
t, err = triangolatte.Polygon(vertices)

In /examples you can find:

• buildings – full-blown WebGL previewer of buildings triangulated in city example
• city – triangulation of whole city downloaded from Open Street Map
• gpx – GPX format parsing and triangulation of its data
• wireframe – desktop OpenGL wireframe previewer for triangulated shapes

You will find instructions for running the code there.

Takes array of points and produces array of triangle coordinates.

Based on the following paper and inspired by EarCut.

Removes holes, joining them with the rest of the polygon. Provides preprocessing for Polygon. First element of the points array is the outer polygon, the rest of them are considered as holes to be removed.

Takes array of points and triangulates them to resemble a line of given width. Returns array of two-coordinate CCW triangles one after another.

To select method of joining line segments.

type Joint int

const (
	// No joint correction.
	Normal Joint = 0
	// Producing miter joints, i.e. extending the lines until they meet at some point.
	Miter Joint = 1
)

For calculations using points.

type Point struct {
  X, Y float64
}

A wrapper for Point used in cyclic list.

type Element struct {
	Prev, Next *Element
	Point      Point
}

You can have a look at helpers.go file. It stores triangolatte's helper functions used mostly by tests and examples. They are not exported because I don't want to commit to supporting them in the future, but they might turn out useful for you.

Code is (more or less) covered in tests. You can run them like this:

go test -v

You can also run benchmarks for selected functions (refer to the *_test.go files for availability). For example:

go test -run NONE -bench IsInsideTriangle

NOTE: This section contains work in progress. Numbers below are better reference point than nothing, but still far from perfect.

Polygon() on shape with 10 vertices takes 754ns on average.

Triangulation of 75 thousand buildings runs for around 3.43s.

Using average programmer's notebook. Expect speed up on faster CPUs or while splitting execution into separate threads.

CPU time % usage snaphost using Flame Graphs:

Want to learn what is it or maybe you are willing to generate one yourself? Check FlameGraphs document in this repository.

NOTE: this library is developed mostly with map data triangulation in mind and it will be its main performance target.

• explore possibilities for optimizations in JoinHoles(...)
• maybe allow reusing point array for massive allocation reduction

• provide more examples (e.g. desktop OpenGL usage, mobile app, live rendering pipeline, other unusual use cases...)
• add benchmarks with comparison to libraries in other languages

One of the core plans for this library's development is creating, as soon as it becomes possible, some kind of WebAssembly module for use in JS.

You are welcome to create an issue or pull request if you've got an idea what to do. It is usually a good idea to visit Gitter and discuss your thoughts.

Don't have one, but still want to contribute? Get in touch with us and we can brainstorm some ideas.

MIT License – refer to the LICENSE file.