Loading and Generating Data

Before you can analyze data, it has to exist somewhere DuckDB can read it. This page surveys the handful of ways analysts get data in, and introduces two DuckDB conveniences you will use constantly: the range() generator and CREATE TABLE AS. These let you build realistic practice datasets in seconds — no files required.

Four ways data arrives

For real work, files dominate — analysts point DuckDB at a CSV or Parquet export and query it directly. In this browser course we lean on generated rows, because they let every example be self-contained and reproducible. Both teach the same SQL.

Literal rows with VALUES

The simplest source is data you type yourself. You have seen this in INSERT, but VALUES can also stand alone as a tiny table — handy for quick experiments.

No CREATE TABLE needed — VALUES becomes a table inline. Analysts use this to sketch a quick example or test an idea.

Generating rows with range()

DuckDB's range(start, stop) produces the integers from start up to but not including stop. Combined with arithmetic, it can fabricate realistic-looking datasets of any size — perfect for practice and for testing a query's behavior at scale.

You can turn the integer into dates, categories, and amounts using arithmetic and DuckDB's list-indexing trick ['a','b','c'][n]:

That same pattern, with range(1, 1000001), would give you a million rows for a stress test — generated instantly, no file in sight.

Capturing a result with CREATE TABLE AS

The most important habit in this lesson: CREATE TABLE AS SELECT (often shortened to CTAS). It runs a query and stores its result as a new table. Analysts use it to "freeze" a cleaned-up or generated dataset so later queries can build on it.

Here the initSql step used CTAS to build daily_sales once; the main query just reads it. This separation — build the dataset, then ask questions of it — mirrors how real analysis is organized.

A note on files (for real-world work)

Outside this browser sandbox, the line you would reach for most is reading a file directly. DuckDB lets you treat a CSV or Parquet file as a table:

-- Real-world DuckDB (needs file access, shown for reference):
SELECT region, SUM(amount) AS revenue
FROM 'sales_2024.parquet'
GROUP BY region;

No import step, no schema declaration — DuckDB infers the columns and scans the file. This "query the file in place" ability is a big reason analysts love it. The browser sandbox here has no file system, so we generate data instead, but the SQL you practice is identical.

Check your understanding

You can now bring data into DuckDB and conjure practice datasets at will. With data in hand, the real fun begins — let us learn how an analyst profiles an unfamiliar dataset.