// the find
apache/sedona-db
A single-node analytical database engine with geospatial as a first-class citizen
SedonaDB is a single-node analytical database for geospatial workloads, built on Apache Arrow and DataFusion with a Rust core. It's the single-node companion to the distributed Apache Sedona project, targeting GIS analysts and developers who need fast spatial SQL on datasets that fit on one machine. If you're doing point-in-polygon joins, KNN queries, or map algebra without wanting to spin up Spark, this is the niche it fills.
Built on DataFusion and Arrow means you get zero-copy columnar processing and interop with GeoPandas, DuckDB, and Polars for free — no serialization overhead when passing data between tools. CRS tracking is propagated through the query engine rather than left to the caller, which is the right call and something PostGIS users will appreciate. The benchmark numbers against DuckDB and GeoPandas at SF10 are credible enough to take seriously given the DataFusion foundation. GPU acceleration path via CUDA/RAPIDS is already present in the codebase, not just a roadmap item.
Raster support is 'coming soon' — prominently advertised but not shipped, so if that's your use case you're waiting. At 465 stars the user base is thin, which means production battle-testing is limited and you'd be an early adopter eating edge cases. The dependency chain is heavy: GEOS, PROJ, GDAL, S2, and a CUDA stack for GPU — getting this to build cleanly on an arbitrary CI environment will cost you time. Single-node by design means there's no answer for datasets that outgrow one machine other than 'switch to SedonaSpark', which is a full rewrite of your pipeline.