SQL Concepts Lab — interactive SQL trainer running entirely on DuckDB-Wasm

[oluies]

I built SQL Concepts Lab — an interactive SQL trainer that runs entirely on DuckDB-Wasm, with no server or backend of any kind. Everything (the database, the example queries, and the exercise grading) happens inside the browser tab.

• 🦆 Live: https://oluies.github.io/sql-concepts-lab/
• 📦 Source: https://github.com/oluies/sql-concepts-lab

What it is

Ten lessons that build from SELECT/WHERE up through joins, GROUP BY/HAVING, CTEs, window functions, set operations, EXPLAIN/query plans, and a SQL-dialect comparison (T-SQL vs DuckDB vs standard) — plus a free-form sandbox. Each lesson has runnable examples and an exercise that you answer by writing a query.

It runs against a small synthetic trading dataset (traders, instruments, trades) and leans into DuckDB's friendly-SQL features as teaching moments: FROM-first queries, GROUP BY ALL, QUALIFY, UNION ALL BY NAME, and literal ANTI/SEMI JOIN.

How it uses DuckDB-Wasm

• Built with Vite + TypeScript; bundles the MVP and EH builds via selectBundle so it picks the best target per browser.
• The exercise checker runs the user's query and a reference solution against the same in-browser connection, then compares canonicalized result sets — so grading is just DuckDB checking DuckDB, no expected-output fixtures to maintain.
• Pinned to @duckdb/duckdb-wasm@1.33.1-dev45.0 (DuckDB v1.5.1 in the tab).
• Hosted as a static GitHub Pages site — the whole thing is just files on a CDN.

Why DuckDB-Wasm made this possible

The pitch of the site is "every number on this page is computed by a real analytical SQL engine running in your browser tab, offline-capable after first load." That's only honest because DuckDB-Wasm is a genuine engine, not a teaching toy — EXPLAIN ANALYZE, window functions, and ROLLUP all behave exactly as they would in a server DuckDB, which is the entire point of a trainer.

Huge thanks to the DuckDB-Wasm team — feedback very welcome. 🙏