Don't Panic — DuckDB Fundamentals

DuckDB is the analytical SQL engine that decided the warehouse could live inside the process you already have open. No separate server for the common case. You open memory or a file, type SQL, and it answers aggregations and joins where the data already sits.

Before this shape of tool, the usual forks were painful. Either you stood up a warehouse for a question that fit on a laptop, or you dragged millions of rows through a row-at-a-time engine and watched the CPU negotiate with each value individually. DuckDB takes SQLite’s embedding habit and aims it at OLAP: many rows, few columns, heavy scans.

Two ideas carry almost everything else. Columnar storage keeps each column together so a query that needs three fields does not haul the other ninety. Vectorized execution processes batches of values instead of walking row by row. Remember those, and the rest of the vocabulary starts attaching to something real.

The pleasant surprise is files as tables. A Parquet path can appear in FROM. Globs pull a folder into one scan. You materialize a native table only when repetition justifies it. Extensions then widen the world: HTTPS, S3, JSON, and friends load when needed instead of shipping in every binary.

The less pleasant surprise is concurrency. Inside one process, threads can write when they do not fight over the same row. Across processes, either many readers open the file read-only, or one process holds the pen. That is not a missing checkbox. It is how the engine keeps catalogs and pages cached for speed. Shared writable analytics belongs on MotherDuck, DuckLake, or the maturing Quack path, not on hope and NFS locks.

If a week from now you remember only this tab, keep three moves: open the right mode (memory versus file), query the file before you invent a schema, and treat multi-writer sharing as a deployment choice rather than a PRAGMA.

Next reads: the Intro for the full map, the Cheatsheet when you need the decision tables, Practice when you want the commands under your fingers, and Field Notes when you want the production bruises the cheerful docs understate.

Don't Panic — DuckDB Fundamentals

Relevant careers