The goal of decido is to provide an R binding to the Mapbox library earcut.hpp for constrained polygon triangulation. Decido is aimed at package developers at the moment, there are not high-level classes or objects but the earcut functionality can be easily used in higher-level tools or just used directly.

Ear cutting (or ear clipping) applies constrained triangulation by successively ‘cutting’ triangles from a polygon defined by path/s. Holes are supported, the earcut library works with single-island-with-holes polygons, analogous to the POLYGON type in simple features.

This augments the Javascript version available in rearcut (also an R wrapper of the JS version by Mapbox). Only very minimal comparison testing has yet been done to compare these implementations.

## Installation

Install the released version from CRAN.

install.packages("decido")

You can install the development version from GitHub with the following code. You will need the set of development tools for building source packages with binary code.

## install.packages("devtools")
devtools::install_github("hypertidy/decido", build_vignettes = TRUE)

## Example

This is a basic example of triangulating a single-ring polygon. The output is a vector of triplet indices defining each triangle.

library(decido)
x <- c(0, 0, 0.75, 1, 0.5, 0.8, 0.69)
y <- c(0, 1, 1, 0.8, 0.7, 0.6, 0)
earcut(cbind(x, y))
#>  [1] 2 1 7 7 6 5 5 4 3 2 7 5 5 3 2

See the documentation and vignette for more.

Open the getting started vignette.

vignette("decido", package = "decido")

## Development

This is motivated by the topology aspirations of hypertidy/silicate. We need tools for decomposing shape data into primitives for analysis and visualization. Decomposition into graph types is already well supported and exercised, but triangulations of paths versus triangulations from edges are two key facilities needed for greater control.

This broader project is fairly well advanced in silicate and show-cased in hypertidy/anglr.

## Other implementations

Ear clipping (or ear cutting) is also available in the rgl function triangulate (implemented in R), and in the lawn function lawn_tesselate (implemented via the Mapbox Javascript library earcut). In rgl the function also classifies input coordinates according to their nesting, a necessary first step if the relaionship between holes and islands is not known. The INLA package has some kind of constraint-based triangulation, but I don’t yet know the details.

In comparison to path-based ear-clipping, other libraries ‘Triangle’ and ‘CGAL’ provide edge-based mostly Delaunay triangulation. The Triangle library is available in the R package RTriangle. Experimental implementations binding CGAL are in rcgal and laridae.

Do you know of others? Let me know! Triangulation is common across many R packages, but constrained algorithms are pretty rare (it’s hard). There are many Delaunay and other non-constrained implementations in many packages, and I’m compiling a list of those as well. OTOH there’s rgeos, sf, deldir, geometry, tripack, spatstat, akima, several mesh-related packages Rvcg, meshsimp, icosa, webglobe …

Please note that the decido project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.