What it is

DuckDB is an analytical, in-process SQL database management system written in C++ and released under the MIT license. It runs inside the application process rather than as a separate server, and it is distributed as a standalone command-line application as well as through clients for Python, R, Java, and WebAssembly. The project lives in the analytical SQL ecosystem, where the topics listed for it are analytics, database, embedded-database, OLAP, and SQL. It has been developed for about eight years, with the last push recorded on 15 September 2026.

The concrete problem DuckDB solves is analytical query processing on data that already sits in files or in the host process, without a server to install, configure, and administer. Files can be queried directly: referencing myfile.csv or myfile.parquet in the FROM clause of a statement is enough to read them. That makes it usable for local analysis, embedded analytics, and pipelines where a separate database deployment would add operational weight. The SQL dialect goes past basic SQL, covering arbitrary and nested correlated subqueries, window functions, collations, and complex types such as arrays, structs, and maps, plus extensions intended to make SQL friendlier to write.

Key capabilities

• Analytical (OLAP) query engine designed to be fast, reliable, portable, and easy to use.
• Rich SQL dialect supporting arbitrary and nested correlated subqueries, window functions, and collations.
• Complex data types, including arrays, structs, and maps.
• Direct import of CSV and Parquet files by naming the file in the FROM clause, with no separate load step required for the simple case.
• Standalone CLI client for interactive use, plus language clients for Python, R, Java, and WebAssembly.
• Deep integration with pandas for Python users and a dplyr API (duckplyr) for R users.
• Extension mechanism, including extensions that simplify SQL writing.

Who uses it and how

• Analysts and engineers who query CSV or Parquet files in place, using the CLI or a language client rather than a server-backed warehouse.
• Python users who work on pandas data frames and want SQL semantics over them through the documented SQL-on-pandas integration.
• R users who work with dplyr and want to keep that interface while querying through the duckplyr API.
• Application developers who embed the

SELECT * FROM 'myfile.csv';
SELECT * FROM 'myfile.parquet';