DuckDB

DuckDB 2.0 lets you export and reimport aggregate state for quantile, median, mad, approx_top_k, approx_quantile, string_agg and list-based aggregates, adds asynchronous Parquet reads and writes, allows MERGE INTO inside triggers, adds RETURNING support for tables with triggers, a dialect_compatibility_mode='spark' setting, and static linking of C-API extensions. Because this is a major version release, teams should re-test existing C-API extensions and any code relying on internal identifier handling or plan-time default resolution before upgrading, since these internals were reworked.

Matters for: Relevant to teams building custom DuckDB extensions through the C-API and to analytical pipelines that read or write large Parquet datasets and need faster concurrent I/O.