libcgal-julia

This library exposes a series of types and functions from CGAL (Computational Geometry Algorithms Library) to the julia language. It serves as the supporting library for the CGAL.jl package (not yet published).

CGAL is a powerful easily-accessible C++ library that aims to provide algorithms in computational geometry (hence the name) with emphasis on reliability and efficiency while still providing robust results.

Building

The project can be built manually, albeit discouraged since setting up the required dependencies may be an arduous task. Building the project for various julia versions and platforms now is highly reliant on Yggdrasil.

Currently, two shared libraries are being built: One with inexact constructions, i.e., using the Exact_predicates_inexact_constructions_kernel linear kernel, and another with exact constructions, i.e., using the Exact_predicates_exact_constructions_kernel_with_sqrt, each library being respectively suffixed with _inexact and _exact. However, due to pre-compilation, the current mechanism to swap between the two is made virtually impossible. Although still present in the source code, alternatives should be looked into to support different kernels. Ideally, support them with proper templating instead of opaquely mapping kernel objects.

Using build_tarballs.jl (outdated)

NOTE: This method is relatively outdated. It might work, but be warned. YMMV. I haven't relied on it since adopting the jll artifacts system.

You're only required to have julia ≥ 1.0 installed. Once that requirement is met, just run:

$ julia .github/scripts/build_tarballs.jl [optional,list,of,comma,separated,triplets]

You can provide a --help argument to get more information on BinaryBuilder's build script options. Check out BinaryBuilder.jl for more information.

By default, it will build for Windows (mingw) 32- and 64-bit architectures, and GNU/Linux and MacOS 64-bit architectures. If you'd like to build for a specific (set of) target(s), explicitly specify the targets you wish to build for. For example, if you wish to build for Windows x86_64 using GCC 7 targetting C++11 ABI upward, run:

$ julia .github/scripts/build_tarballs.jl x86_64-w64-mingw32-gcc7-cxx11

Manually

Requirements:

• CMake ≥ 3.14
• CGAL 5.3
  • Boost ≥ 1.66
  • GMP ≥ 4.2
  • MPFR ≥ 2.2.1
• Julia ≥ 1.3
• JlCxx ~0.8

Depending on your system, it may be easier to aggregate and install the listed requirements, with the exception of JlCxx, which, to my knowledge, isn't in as readily available on any platform I've tested. However, the former ones are. On Windows, (most) via neatly packed installers; on Linux & co., via the distribution's package manager; on macOS, maybe homebrew, although I'm not sure.

Afterwards, it is a mostly typical run-of-the-mill CMake build:

cmake -B build -DJlCxx_DIR=<path/to/libcxxwrap-julia>
cmake --build build

We can see above that JlCxx_DIR is specified. However, if it's reachable by CMake, you won't need it (i.e., libcxxwrap-julia is installed in your system in a standard path).

TODO

• Kernel
  • Objects
  • Global Functions
    • 2D
    • 3D
    • Circular
    • Spherical
  • ...
• Type Parametrization?
• Iterators
• Circulators
• Other packages, such as
  • 2D Convex Hulls
  • 3D Convex Hulls
  • 2D Voronoi Diagram
  • ...

Related Projects

• cgal-swig-bindings - CGAL bindings using SWIG
• scikit-geometry - Scientific Python Geometric Algorithms Library (f.k.a. PYGAL, CGAL python bindings using pybind11 by wolfv