vortrace

Fast one-dimensional integrals through unstructured Voronoi meshes.

One-dimensional integrals through Voronoi meshes can be expensive. Because the mesh is unstructured, it is not obvious a priori where a line intersects the faces of which cells. Brute-force methods that sample large numbers of points struggle with systems that have a large dynamic range in cell size, like cosmological simulations.

vortrace performs these integrals with the fewest nearest-neighbor calls possible. With an optimized C++ backend and a user-friendly Python frontend, it’s easy to get started:

import vortrace as vt

pc = vt.ProjectionCloud(pos, density)
image = pc.grid_projection(extent, npix, bounds, center)

Features

  • Fast ray tracing – recursive split-point algorithm minimizes kDTree queries

  • Multiple reduction modes – integrate, max, min, and volume rendering

  • Python and C++ APIs – thin Python wrapper over a standalone C++ library

  • Multi-field support – project multiple scalar fields simultaneously

  • Custom rotations – Tait-Bryan yaw/pitch/roll for arbitrary viewing angles

  • Periodic boundaries – full support for cosmological simulation boxes

Getting Started

Tutorials

Concepts

API Reference

Project