An Evaluation of Modern Spatial Libraries

Varun Pandey, Alexander van Renen, Andreas Kipf, Alfons Kemper

Abstract: Applications such as Uber, Yelp, and Tinder rely on spatial data or locations from their users. These applications and services either build their own spatial data management systems or rely on existing solutions. The JTS Topology Suite (JTS), its C++ port GEOS, Google S2, ESRI Geometry API, and Java Spatial Index (JSI) are among the spatial processing libraries that these systems build upon. Applications and services depend on the indexing capabilities available in such libraries for high-performance spatial query processing. However, limited prior work has empirically compared these libraries. Herein, we compare these libraries qualitatively and quantitatively based on four popular spatial queries and using two real-world datasets. We also compare a lesser known library (jvptree) which utilizes Vantage Point Trees. In addition to performance evaluation, we also analyzed the construction time, and space overhead, and identified the strengths and weaknesses of each libraries and their underlying index structures. Our results demonstrate that there are vast differences in space consumption (up to 9.8 x), construction time (up to 5 x), and query runtime (up to 54 x) between the libraries evaluated.