Your first queries

Start the installed shell with a database filename:

./bin/roastedDuck example.duckdb

Create a tiny dataset and aggregate it:

CREATE TABLE sales (category VARCHAR, amount INTEGER);
INSERT INTO sales VALUES ('books', 20), ('coffee', 8), ('books', 15);

SELECT category, sum(amount) AS total
FROM sales
GROUP BY category
ORDER BY total DESC;

The totals are books: 35 and coffee: 8. With a database filename, committed changes persist when you close and reopen the shell. Without a filename, the database is in memory. End statements with a semicolon; use Ctrl+D to exit.

Read a local file

COPY sales TO 'sales.parquet' (FORMAT PARQUET);
SELECT * FROM 'sales.parquet' ORDER BY category, amount;

CSV and Parquet readers, joins, aggregates, and window functions are part of the compatibility target. SQL can access files and configured network resources; run untrusted SQL inside an appropriate application or operating-system sandbox.

Run a query from the command line

./bin/roastedDuck -csv -c 'SELECT i, i * i AS square FROM range(5) t(i);'

Check before migrating

Work on a copy of an existing database during evaluation. The candidate has recorded bidirectional persistence/WAL tests against the pinned reference, but that is not a promise of compatibility with every database or future DuckDB version. Read the compatibility contract.