FlatGeobuf

FlatGeobuf is a highly performant binary encoding designed for geographic data, built upon the efficient flatbuffers technology. It is capable of holding collections of Simple Features, including circular interpolations as specified by SQL-MM Part 3.

Inspired by projects like geobuf and flatbush, FlatGeobuf deliberately avoids supporting random writes to maintain simplicity and enable data clustering on a packed Hilbert R-Tree. This design choice facilitates fast bounding box spatial filtering. The spatial index is optional, allowing the format to be efficiently written as a stream, support appending new data, and cater to use cases where spatial filtering is not a primary requirement.

The primary goals of FlatGeobuf are to be suitable for large volumes of static data, provide significantly faster performance than older formats without imposing size limitations on contents or metainformation, and ensure suitability for both streaming and random access scenarios. More in-depth information regarding the challenges posed by legacy formats can be found on switchfromshapefile.org.