DuckDB

DuckDB is an in-process analytical database, optimized for fast aggregations, joins and reads over columnar data. It has no server: you point QoreDB at a file (or use an in-memory database) and you're connected.

Quick connect

Open New Connection, choose DuckDB, and either browse to a file or paste a path.

The driver also accepts a path of :memory: for an ephemeral in-memory database.

Connection fields

Like SQLite, there is no host, port, user or password. QoreDB creates the file automatically if it doesn't exist; first write persists it to disk.

Supported file extensions

The driver validates the file extension. Two are accepted:

• .duckdb
• .db

If you have a different extension, rename the file or use :memory:.

In-memory databases

Use :memory: (or duckdb::memory:) for a database that lives only in RAM and disappears when the connection closes. Useful for:

• One-off exploratory queries
• Testing analytical queries before persisting them
• Loading a CSV or Parquet, doing some transformations, exporting the result

What DuckDB is good at

DuckDB is a columnar engine designed for analytical workloads (OLAP). It shines at:

• Aggregations and joins over millions of rows on a single machine, without a server.
• Reading Parquet and CSV directly with SQL (SELECT * FROM 'data.parquet' works out of the box).
• Embedded analytics: you can ship a .duckdb file with a desktop app and query it locally.
• Time-series and analytical queries that would be slow on row-stored OLTP databases.

For high-throughput transactional workloads (lots of single-row inserts and updates), pick PostgreSQL or another row-store.

Cross-database federation

DuckDB is also the engine behind QoreDB's cross-database federation: it can attach external databases (Postgres, MySQL, SQLite, and others) and query them as if they were local DuckDB tables. This lets you join data across heterogeneous sources from a single SQL query.

The federation feature is set up at the workspace level rather than per-connection. The DuckDB driver itself is for connecting to a regular DuckDB file or in-memory instance; the federation engine reuses DuckDB internally but isn't exposed through the same connection dialog.

Common issues

Where to go next

• SQLite for the row-store equivalent on a single file
• Querying: editor and autocomplete for the SQL editor
• Reference: connection URLs for the full URI grammar