SQLite Database Lab

Query an export without setting up a server

Import your CSV or JSON under the default table name records. The sample groups sales by category and calculates their totals. Change the query to select only the columns or rows you need.

SELECT category, ROUND(SUM(amount), 2) AS total
FROM records
GROUP BY category
ORDER BY total DESC;

Join two local datasets

Import the first file as orders and the second as customers, then join them by their common identifier. Choose a new name for each import: importing never overwrites an existing table.

SELECT c.name, SUM(o.total) AS revenue
FROM orders AS o
JOIN customers AS c ON c.id = o.customer_id
GROUP BY c.id, c.name
ORDER BY revenue DESC;

Repeat a query as a workflow

The CSV → SQLite query → TypeScript workflow carries data through connected steps. Change the query, inspect its JSON result, then generate types, Zod schemas or CSV. Save the workflow locally to run the same task on a new export.

Storage, precision and supported SQL

The database lives in memory in this tab. Download it before closing or refreshing the page to retain imported tables and SQL changes. Opening a database file replaces the working copy after the file has been validated. A failed or cancelled operation keeps your last successful copy.

Imported columns use SQLite’s dynamic storage without column affinity, so a JSON string such as "00123" remains text. CSV inference is optional. Signed 64-bit integers stay exact; wider integers or numbers that cannot be represented safely are stored as text. Arithmetic, casts and user-written SQL follow SQLite rules and can change precision. Blob results are exported as hexadecimal strings, and non-finite floating-point values are exported as strings because JSON cannot represent them.

SELECT, CTEs, joins, aggregations, VALUES, EXPLAIN and CREATE/ALTER/DROP/INSERT/UPDATE/DELETE statements are supported. Transaction-control statements, ATTACH, PRAGMA and server-specific SQL are excluded. A script’s results come from its last row-producing statement. CSV downloads protect text beginning with spreadsheet formula characters. Duplicate result column names receive suffixes in exported files so no value is dropped.

For larger datasets, reduce the file first or use a native database. This lab limits imports to 5 million text characters and 250,000 cells, database files to 20 MB, and displayed/exported results to 5,000 rows. Save basic tools offline while connected to include the SQLite processor and WebAssembly binary.

For analytics over Parquet or multiple imported file formats, try DuckDB Playground. Its queries start with a fresh database each time; this SQLite lab is the choice when you need to inspect, change or download a SQLite database.

Frequently asked questions