What is inside Query Belt
A real SQL engine, tasks you can trust, mock interviews and tools that show what the database is doing. Everything works offline.
SQLite sandbox with four practice databases
Run any query on a shop, an HR database, a streaming service and a logistics warehouse. The engine is SQLite, bundled in the app so it behaves the same on every phone. Your history is kept on the device, results can be shared as CSV through the system share sheet, and one tap resets a database.
180 tasks and 36 lessons in seven belts
From SELECT and WHERE to JOINs, subqueries, CTEs, window functions, interview patterns and data changes with indexes. Lessons take about two minutes and include runnable examples. All data is synthetic.
- White belt
SELECT, WHERE, ORDER BY, NULL basics - Yellow belt
Aggregates, GROUP BY, HAVING, CASE - Orange belt
JOINs and set operations - Green belt
Subqueries, EXISTS, CTEs - Blue belt
Window functions - Brown belt
Interview patterns: dedupe, gaps and islands - Black belt
Data changes, indexes, query plans
Answer checks and the row-by-row view
Your result is compared with the reference result, not your query text. When rows differ, you see the extra and missing rows side by side and a likely reason: duplicates from a JOIN, a NULL trap, a boundary value in BETWEEN, a different column order, or text that differs only in case.
Tested on several datasets
Each task is checked against more than one version of the data, so a query written for the sample rows alone is caught. Every reference answer is run against all datasets when the content is built.
Mock interviews by role
Data Analyst, Backend and BI presets with different topic weights. Quick (2 tasks, 15 minutes), Standard (3 tasks, 30 minutes) and Full (4 tasks, 45 minutes). No hints, one submit per task, then a report with a map of weak topics and a practice score. One Quick interview is free.
Query Plan Lens
SQLite can tell you how it will run a query. The lens translates that into plain language and lets you create an index in the sandbox to see the plan change.
SCAN employeesSEARCH employees USING INDEX idx_employees_department (department_id=?)Illustration: real plan lines from SQLite for WHERE department_id = 2, before and after CREATE INDEX. The lens shows the structure of the plan, not timings.
Why is it empty?
For a plain SELECT that returns nothing, the app removes conditions one by one, or turns a JOIN into a LEFT JOIN, and tells you at which step your rows disappeared. For queries it cannot take apart, it says so.
Pattern cards and spaced review
Top N per group, running totals, second highest, deduplication, gaps and islands. Each card has the idea, two or three ways to solve it and a check in the sandbox. Tasks you missed come back at growing intervals.
A keyboard made for SQL
A snippet bar above the keyboard with keywords and symbols, templates for common clauses, suggestions from the current schema, undo and redo, and a font size setting. Typing never fights autocorrect. An optional mode without the bar is there if you want to type everything by hand.
Daily query and a gentle streak
One short task a day. A missed day is covered by a weekly freeze instead of resetting your streak. Reminders are optional, local and never exact-time alarms.
What we do not do
- No ads and no advertising SDKs.
- No account and no server for your practice.
- No certificates, no job promises and no countdown timers on purchases.
- No AI chat that makes up answers: checks are deterministic.
Not sure about the price? Read the Free and Pro table before you install.