What it does
Loads a CSV file from your device, or a built-in sample table of 30 orders (click Try sample data). The table is called
data.Runs SELECT queries with DISTINCT, WHERE (including LIKE, IN, BETWEEN and IS NULL), GROUP BY, HAVING, ORDER BY, LIMIT and OFFSET, and the aggregates COUNT, SUM, AVG, MIN, MAX and GROUP_CONCAT.
Works out each column's type from its values, so number comparisons and sums behave as expected.
Runs entirely in the page. The file is never uploaded, and nobody needs an account, which keeps it usable in schools with strict data protection rules.
Getting data in
Drop a CSV file onto the page, choose one, paste the text, or click Try sample data. Under Options you can set the delimiter (detected automatically, or comma, semicolon, tab or pipe), say whether the first row holds the column names, and edit the raw data. The Columns list shows every column with the type the playground found (number, date, text and so on); click a name to add it to the query.
Writing and running queries
The query runs again a moment after you stop typing, so the result is always up to date. Run or Ctrl+Enter (Cmd+Enter on a Mac) runs it at once.
The Try links load five example queries: Everything, Filter, Group and total, Having and Text search.
The result can be copied, or downloaded as
query-result.csv.Besides the aggregates, there are functions for text and numbers: LOWER, UPPER, TRIM, LENGTH, SUBSTR, REPLACE, ROUND, ABS, COALESCE and IFNULL.
||joins text:customer || ' (' || company || ')'.Column names with spaces go in double quotes, square brackets or backticks:
"order date".
What it leaves out
It queries one table. JOIN, UNION and window functions are rejected with a clear error that names the missing feature, and so is anything that writes data (INSERT, UPDATE, DELETE, CREATE). A subquery, such as WHERE revenue > (SELECT AVG(revenue) FROM data), stops with a syntax error instead. CASE expressions and date functions such as strftime are not available either; for months, SUBSTR(date, 1, 7) turns 2026-01-04 into 2026-01. That makes it a good first step before a full database system, not a replacement for one.
Where it differs from SQLite
Learners who move on to the SQL sandbox, which runs real SQLite, will notice a few differences:
In the playground | In SQLite |
|---|---|
|
|
An empty cell counts as NULL, so | An empty text and NULL are different |
| The table name must be right |
GROUP BY needs the expression itself, not its alias | GROUP BY can use an alias |
No subqueries, joins or CASE | All available |
ORDER BY and HAVING can use an alias in both: HAVING orders > 2 works.
A first query
SELECT country, COUNT(*) AS orders
FROM data
GROUP BY country
ORDER BY orders DESCOn the sample data this lists nine countries: Germany first with 6 orders, then the Netherlands and the United States with 5 each.
A second step: the share of each channel in paid revenue. Paid revenue is 9,469 in total.
SELECT channel, ROUND(SUM(revenue) * 100.0 / 9469, 1) AS share
FROM data
WHERE status = 'paid'
GROUP BY channelDirect 48.2 percent, Marketplace 26.1 percent, Partner 25.7 percent.
Reading error messages
The playground names the problem in plain words. The ones learners meet first:
No such column: contry. The name is misspelt, or an alias was used in GROUP BY. Check the Columns list.
Expected BY but found country. GROUP or ORDER without BY. "Expected ... but found ..." always names what is missing.
Query ended unexpectedly. The query stops too early, for example right after WHERE or ORDER BY.
Unclosed call to SUM(). A closing bracket is missing.
JOIN is not supported, this engine queries a single table. The feature exists in SQL, but not here. Use the SQL sandbox for joins.
One mistake gives no error at all: a missing comma. SELECT country revenue FROM data shows the country column under the heading "revenue", because a name right after a column counts as its alias. SQLite does the same. If a column seems to be missing from the result, look for a missing comma.
Ideas for class
A first SQL lesson. The 45-minute exercise on sales data has six tasks with solutions.
Open data. Download a CSV from an open data portal and let learners ask their own questions of it.
Spreadsheet versus SQL. Answer the same question once with a pivot table and once with a query, then compare which one you could still read a week later.
Ten questions, ten queries. The sample orders dataset has ten tasks with checked answers that all run in the playground.
Next step. Move to the SQL sandbox for joins across several tables, subqueries and creating tables.
Using this page
The tool is free to use. This description and the ideas on it are licensed under CC BY 4.0.
