CGAL.jl - CGAL meets Julia

This is the Concept DOI. For a specific version, please follow the link.

A package exposing a series of types, constructs, functions, predicates, and algorithms from CGAL (Computational Geometry Algorithms Library), a powerful, reliable, and efficient C++ library

This package is supported by a C++ wrapper around CGAL in the form of libcgal-julia, itself powered by JlCxx.

Since the kernel is being fixed on the C++ side, the usual typedefs you see in CGAL's examples aren't as common. Here's one of CGAL's examples in a side-by-side comparison with a Julia translation using this package:

// points_and_segment.cpp
#include <iostream>
#include <CGAL/Simple_cartesian.h>

typedef CGAL::Simple_cartesian<double> Kernel;
typedef Kernel::Point_2 Point_2;
typedef Kernel::Segment_2 Segment_2;

int main()
{
  Point_2 p(1,1), q(10,10);

  std::cout << "p = " << p << std::endl;
  std::cout << "q = " << q.x() << " " << q.y() << std::endl;

  std::cout << "sqdist(p,q) = "
            << CGAL::squared_distance(p,q) << std::endl;

  Segment_2 s(p,q);
  Point_2 m(5, 9);

  std::cout << "m = " << m << std::endl;
  std::cout << "sqdist(Segment_2(p,q), m) = "
            << CGAL::squared_distance(s,m) << std::endl;

  std::cout << "p, q, and m ";
  switch (CGAL::orientation(p,q,m)){
  case CGAL::COLLINEAR:
    std::cout << "are collinear\n";
    break;
  case CGAL::LEFT_TURN:
    std::cout << "make a left turn\n";
    break;
  case CGAL::RIGHT_TURN:
    std::cout << "make a right turn\n";
    break;
  }

  std::cout << " midpoint(p,q) = " << CGAL::midpoint(p,q) << std::endl;
  return 0;
}

# points_and_segment.jl
using CGAL

p, q = Point2(1, 1), Point2(10, 10)

println("p = $p")
println("q = $(x(q)) $(y(q))")

println("sqdist(p,q) = $(squared_distance(p, q))")

s = Segment2(p, q)
m = Point2(5, 9)

println("m = $m")
println("sqdist(Segment2(p,q), m) = $(squared_distance(s, m))")

print("p, q, and m ")
let o = orientation(p, q, m)
    if     o == COLLINEAR  println("are collinear")
    elseif o == LEFT_TURN  println("make a left turn")
    elseif o == RIGHT_TURN println("make a right turn")
    end
end

println(" midpoint(p,q) = $(midpoint(p, q))")

Drop into a REPL and type the following:

julia> import Pkg; Pkg.add("CGAL")

Alternatively, in a blank REPL, after hitting ],

pkg> add CGAL

Currently, two different binary libraries are made available in libcgal-julia: one compiled with the Exact_predicates_inexact_constructions_kernel, and another one with Exact_predicates_exact_constructions_kernel_with_sqrt (i.e. using doubles => Float64, and CORE::Expr => FieldType). By default, this package uses the inexact variant, trading better performance for minor, even sometimes negligible, dare I say, inexact results.

There are two different ways of loading a custom library, both of which come in the form of defining environment variables. One of the glaring downsides of this approach is the package must be rebuilt in order to pick up the change before loading the module due to precompilation.

On the julia side of things, this distinction is made based on the existence of a FieldType type mapped from the C++ side. Take a look inside kernel.jl to see how this is handled.

In order to use a kernel with exact constructions, one must define the JLCGAL_EXACT_CONSTRUCTIONS environment variable. The variable's value is ignored. It only needs to exist. For example,

$ export JLCGAL_EXACT_CONSTRUCTIONS="I'm willing to pay for some performance penalties"
$ julia
julia> build CGAL
...
julia> using CGAL
...
julia> FT
FieldType # numeric type from CGAL. when using inexact constructions => Float64

The full path to an alternative version of the wrapper library can be specified by defining the JLCGAL_LIBPATH environment variable. This will override the JLCGAL_EXACT_CONSTRUCTIONS definition since the former is more specific than the latter.