This package provides a set of geometric primitive types (balls, cuboids, cylinders, and so on) and operations on them designed to enable piecewise definition of functions, especially for finite-difference and finite-element simulations, in the Julia language.
For example, suppose that you are discretizing a PDE like the Poisson equation ∇⋅c∇u = f,
and you want to provide a simple user interface for the user to specify the function c(x).
In many applications, c will be piecewise constant, and you want to be able to specify
c = 1 in one box, c = 2 in some cylinders, etcetera. The GeometryPrimitives package
allows the user to provide a list of shapes with associated data (in this case, the value of
c) to define such a c(x).
Furthermore, the application to discretized simulations imposes a couple of additional requirements:
-
One needs to be able to evaluate
c(x)a huge number of times (once for every point on a grid). So, we provide a fast O(log n) K-D tree data structure for rapid searching of shapes. -
Often, one wants to compute the average of
c(x)over a voxel, so we provide routines for rapid approximate voxel averages. -
Often, one needs not only the value
c(x)but the normal vector to the nearest shape, so we provide normal-vector computation.
This package was inspired by the geometry utilities in Steven G. Johnson's [Libctl] (http://ab-initio.mit.edu/wiki/index.php/Libctl) package.