SQL Formatter & Query Beautifier

Selecting all columns prevents the database optimizer from using Index-Only Scans, forces the engine to read large text/blob pages from disk into the buffer cache, and increases network serialization overhead. Always explicitly list only the required columns.

-- Anti-pattern: SELECT * FROM orders WHERE user_id = 42;
-- Optimized: SELECT id, order_number, total_amount, status FROM orders WHERE user_id = 42;

Wrapping indexed columns in scalar functions (e.g. YEAR(created_at) = 2026 or LOWER(email) = ...) invalidates standard B-Tree indexes, forcing a sequential table scan across millions of rows. Rewrite conditions using range boundaries.

-- Anti-pattern: WHERE YEAR(created_at) = 2026
-- Optimized: WHERE created_at >= '2026-01-01' AND created_at < '2027-01-01'

A correlated subquery in the SELECT clause executes once for every single row returned by the outer query (quadratic time). Refactoring into a LEFT JOIN with aggregation or a Common Table Expression (CTE) allows the optimizer to use hash joins or merge joins.

-- Anti-pattern: SELECT u.name, (SELECT COUNT(*) FROM orders o WHERE o.user_id = u.id) FROM users u;
-- Optimized: SELECT u.name, COUNT(o.id) FROM users u LEFT JOIN orders o ON u.id = o.user_id GROUP BY u.id, u.name;

Instead of self-joining a table to find the latest record or top customer by country, use ROW_NUMBER(), RANK(), or DENSE_RANK() OVER (PARTITION BY ... ORDER BY ...). The database calculates row ranks in a single streaming scan.

SELECT customer_id, total_amount, ROW_NUMBER() OVER (PARTITION BY country_code ORDER BY total_amount DESC) as rank FROM orders;

Never rely on query execution elapsed time alone, as warm buffer caches mask disk I/O bottlenecks. Inspect EXPLAIN ANALYZE to verify whether the optimizer performed an Index Scan, Bitmap Heap Scan, or costly Sequential Scan.

EXPLAIN (ANALYZE, BUFFERS, COSTS) SELECT u.id, o.total FROM users u JOIN orders o ON u.id = o.user_id WHERE u.status = 'active';