Space-divison-based Geometry Compression
This is my current project that I am working on while attending Georgia Tech.
The main idea behind this concept is that, given a set of coordinates, we can create an
n x n grid within
the extent of those coordinates. As
n increases, there will be centers of grid cells that
tend to the initial set of coordinates. Once we have a “fine-enough” grid, we can create an
index over the grid (using space-filling curves),
and take only the cells that align to a coordinate (or set of close coordinates) to form a 1D array of integers.
Since we have an array of integers (resembling an inverted index),
we can use integer-specific compression techniques to reduce the size of this array.
Some of the work that I’ve done on this can be seen here: spress, flatpoints.
GEOM is a geometry and topology engine written in Fortran. It is an exploratory project at manipulating geometry in SoA format, as opposed to the traditional AoS format represented by the simple features standard. The goal of this is to explore the ways that geometry operations can take advantage of SIMD instructions.
rapture (Raster Capture)
rapture is a tool written in Go for visualizing large point datasets as rasters. It draws inspiration from similar tooling, such as datashader and rasterly.
I maintain various R packages. A (non-exhaustive) list of them is as follows: