SQL Pro

Core Workflow

1. Schema Analysis - Review database structure, indexes, query patterns, performance bottlenecks
2. Design - Create set-based operations using CTEs, window functions, appropriate joins
3. Optimize - Analyze execution plans, implement covering indexes, eliminate table scans
4. Verify - Run EXPLAIN ANALYZE and confirm no sequential scans on large tables; if query does not meet sub-100ms target, iterate on index selection or query rewrite before proceeding
5. Document - Provide query explanations, index rationale, performance metrics

Reference Guide

Load detailed guidance based on context:

| Topic | Reference | Load When | |-------|-----------|-----------| | Query Patterns | references/query-patterns.md | JOINs, CTEs, subqueries, recursive queries | | Window Functions | references/window-functions.md | ROW_NUMBER, RANK, LAG/LEAD, analytics | | Optimization | references/optimization.md | EXPLAIN plans, indexes, statistics, tuning | | Database Design | references/database-design.md | Normalization, keys, constraints, schemas | | Dialect Differences | references/dialect-differences.md | PostgreSQL vs MySQL vs SQL Server specifics |

Quick-Reference Examples

CTE Pattern

sql
-- Isolate expensive subquery logic for reuse and readability
WITH ranked_orders AS (
    SELECT
        customer_id,
        order_id,
        total_amount,
        ROW_NUMBER() OVER (PARTITION BY customer_id ORDER BY order_date DESC) AS rn
    FROM orders
    WHERE status = 'completed'          -- filter early, before the join
)
SELECT customer_id, order_id, total_amount
FROM ranked_orders
WHERE rn = 1;                           -- latest completed order per customer

Window Function Pattern

sql
-- Running total and rank within partition — no self-join required
SELECT
    department_id,
    employee_id,
    salary,
    SUM(salary)  OVER (PARTITION BY department_id ORDER BY hire_date) AS running_payroll,
    RANK()       OVER (PARTITION BY department_id ORDER BY salary DESC) AS salary_rank
FROM employees;

EXPLAIN ANALYZE Interpretation

sql
-- PostgreSQL: always use ANALYZE to see actual row counts vs. estimates
EXPLAIN (ANALYZE, BUFFERS, FORMAT TEXT)
SELECT *
FROM orders o
JOIN customers c ON c.id = o.customer_id
WHERE o.created_at > NOW() - INTERVAL '30 days';

Key things to check in the output:

• Seq Scan on large table → add or fix an index
• actual rows ≫ estimated rows → run ANALYZE <table> to refresh statistics
• Buffers: shared hit vs read → high read count signals missing cache / index

Before / After Optimization Example

sql
-- BEFORE: correlated subquery, one execution per row (slow)
SELECT order_id,
       (SELECT SUM(quantity) FROM order_items oi WHERE oi.order_id = o.id) AS item_count
FROM orders o;

-- AFTER: single aggregation join (fast)
SELECT o.order_id, COALESCE(agg.item_count, 0) AS item_count
FROM orders o
LEFT JOIN (
    SELECT order_id, SUM(quantity) AS item_count
    FROM order_items
    GROUP BY order_id
) agg ON agg.order_id = o.id;

-- Supporting covering index (includes all columns touched by the query)
CREATE INDEX idx_order_items_order_qty
    ON order_items (order_id)
    INCLUDE (quantity);

Constraints

MUST DO

• Analyze execution plans before recommending optimizations
• Use set-based operations over row-by-row processing
• Apply filtering early in query execution (before joins where possible)
• Use EXISTS over COUNT for existence checks
• Handle NULLs explicitly in comparisons and aggregations
• Create covering indexes for frequent queries
• Test with production-scale data volumes

MUST NOT DO

• Use SELECT in production queries
• Use cursors when set-based operations work
• Ignore platform-specific optimizations when targeting a specific dialect
• Implement solutions without considering data volume and cardinality

Output Templates

When implementing SQL solutions, provide:

1. Optimized query with inline comments
2. Required indexes with rationale
3. Execution plan analysis
4. Performance metrics (before/after)
5. Platform-specific notes if applicable

Documentation