A Preview of DuckDB v2.0
Points and comments are a snapshot, not live.
DuckDB v2.0 adds server mode, triggers, VARIANT type, asynchronous I/O, and a new SQL parser.
DuckDB v2.0, due this fall, introduces a client/server mode via the `quack` extension and `CONNECT` statement. It adds triggers, the `VARIANT` type with shredded execution, asynchronous I/O for faster network reads, a new PEG-based SQL parser with dialect compatibility, and a new storage format. Other features include `NEAREST` joins, DML in CTEs, nested schemas, variable syntax, JSON mutation functions, and a reimplementation of the ICU library for timezones and collations. Performance improvements include a rewritten recursive CTE engine that is about 40x faster for a reachability query on a million-edge graph.
What commenters are saying
Commenters are enthusiastic about DuckDB's evolution from an in-process engine toward a server foundation for cloud data warehouses. Many share concrete use cases: in-browser dashboards with WASM, ETL on Kubernetes, SvelteKit apps querying .db files on cloud buckets, and replacing Pandas for transformations. One commenter noted the new async I/O will help serve HTTP traffic. A few asked about procedural features like PL/pgSQL and WASM binary size. Some detected AI-generated prose in the post. Others compared DuckDB favorably to SQLite for structured data and to Athena for local processing.