What is DuckDB?

DuckDB is a columnar relational database, OLAP-oriented (analytical) and embedded. It is designed to be integrated into applications, data analysis tools, or data science environments. DuckDB is optimized for complex analytical queries on large datasets, offering high performance through its columnar architecture and efficient processing engine.

What’s new in version 1.4?

I’ll let you discover all the new features in the blog post, but two features seem particularly interesting:

  • Writing to Iceberg tables! DuckDB joins the (increasingly less) small group of software capable of writing to Iceberg tables, notably including Dremio, Olake, Trino…​ With its ability to read data from multiple different sources, its low resource consumption, …​, and now the ability to write to Iceberg, DuckDB becomes an even more essential tool in the modern data ecosystem, potentially replacing solutions like Spark or Trino when technical constraints don’t require distributed computing.

  • MERGE statement support has been added. This is a highly anticipated feature that allows for efficient and atomic "upsert" operations (update + insert). This is particularly useful for data management scenarios where you need to synchronize tables or apply conditional modifications.