Versions:
DataPress, published by DataPress, is a high-performance HTTP server designed to expose Parquet and Delta datasets stored in object storage as fast, typed HTTP APIs. Falling squarely into the data infrastructure and analytics tooling category, it addresses the common challenge of making large columnar datasets accessible to applications without requiring a full database deployment or complex ETL pipelines. The server supports two output formats — JSON and Arrow IPC — allowing both conventional web clients and high-throughput analytical consumers to query the same underlying data efficiently. A distinguishing architectural feature of DataPress is its multi-backend design: it can execute queries using either DuckDB or Apache Arrow with DataFusion, and the single `datapress` binary bundles both engines, selecting one at runtime based on the `server.backend` setting in the user's `datasets.toml` configuration file. This flexibility lets operators choose the query engine best suited to their workload characteristics, hardware, or compatibility requirements without changing binaries or deployment workflows. Typical use cases include serving analytical datasets directly from object storage to dashboards, microservices, and data applications; building lightweight data APIs over lakehouse-style storage; and providing low-latency, typed access to columnar data for downstream consumers that benefit from Arrow IPC's zero-copy transfer model. Because datasets remain in their native Parquet or Delta formats, DataPress fits naturally into modern lakehouse architectures where storage and compute are decoupled. The current version is 0.8.6, and the project has released 22 versions in total, indicating an active development history with ongoing iteration on features, engine support, and stability. As a pre-1.0 release, the software is still evolving, but its focused scope — turning Parquet and Delta datasets into production-ready HTTP APIs with a choice of two proven query engines — makes it a practical option for teams seeking simple, performant data serving over object storage.
Tags: