This repository contains a collection of vector processing benchmarks for Python and R packages. The tests cover the most common operations such as loading and saving a geopackage file, sampling points in a polygon, creating buffers, transformating the coordinate system (CRS), calculating the distance between points, and intersecting geometries.

Note that all operations were performed in the Cartesian coordinate system excluding s2 package, where calculations were performed on the sphere (this may affects the longer calculation times). For more information, see the "Spherical geometry in sf using s2geometry" article and presentation at the FOSS4G 2021 conference.

The detailed results are available at https://kadyb.github.io/vector-benchmark/report.html.

For high-performance data frames processing in R, check data.table and collapse.

You may also be interested in the raster processing benchmarks.

Python:

• geopandas

R:

• sf
• terra
• s2
• geos

Julia:

• GeometryOps.jl

1. Generate the data from data/ folder in R.
2. Run all benchmarks using batch script (run_benchmarks.sh) or single benchmarks files.

Batch script

cd vector-benchmark
./run_benchmarks.sh

Single benchmark

Rscript sf/buffer.R

python3 geopandas/buffer.py

The dataset is synthetically generated and consists of 500,000 points in a planar coordinate system.

• CPU: AMD Ryzen 9 5900X @ 3.7 GHz
• RAM: 64 GB
• OS: Ubuntu 22.04.5 LTS