Skip to content
BENIBIX Learn

DuckDB

Data📊 Analytics & BIQuery Engines

Also known as Duck DB

A small database for sums and counts. It runs inside your own program.

Think of it like

It is like a small tool you keep in your pocket for data. There is no server to set up. It just runs where you are.

Example

An analyst downloads 2 GB of Parquet files. On her laptop, DuckDB queries them in seconds. She never needs a login to the warehouse.

Why it matters

Much analysis does not need a cluster. DuckDB makes a laptop enough for data that once needed a warehouse.

Definition

DuckDB is a free, open source database for crunching numbers, and it runs inside your own program. There is no server to install or start. It stores data in columns. So sums and group-bys over millions of rows are fast, even on a laptop. It can query CSV and Parquet files directly, and work with pandas or Polars data in Python. People often call it the SQLite of data work.

Example code

sql
SELECT country, COUNT(*) AS orders
FROM 'orders.parquet'
GROUP BY country
ORDER BY orders DESC;
DuckDB reads the Parquet file straight from disk. There is no load step and no server.
No server on the right. The database is a library inside the program.

Common mistake

Using it as the main database for a busy app. It is built for analysis, not many small writes from many users.

Key takeaways

  1. 1
    In-process. Runs inside your program.
  2. 2
    Column storage. Fast sums and group-bys.
  3. 3
    Reads files. CSV and Parquet, no loading.

Compare