Practical window function patterns in PostgreSQL

Introduction

Window functions help you analyze data by looking at rows related to the current row without grouping them. They let you calculate running totals, ranks, and moving averages easily.

When you want to rank items like top salespeople without losing detail about each sale.

When you need a running total of sales over time to see trends.

When you want to compare each row to others in the same group, like finding the difference from the previous month.

When you want to assign row numbers to results for pagination or ordering.

When you want to calculate averages or sums over a sliding window of rows.

Syntax

PostgreSQL

SELECT column1, column2, 
       window_function() OVER (PARTITION BY columnX ORDER BY columnY ROWS BETWEEN ... ) AS alias
FROM table_name;

OVER() defines the window for the function.

PARTITION BY divides rows into groups to apply the function separately.

Examples

This ranks employees by salary within each department.

PostgreSQL

SELECT employee_id, department_id, salary,
       RANK() OVER (PARTITION BY department_id ORDER BY salary DESC) AS dept_rank
FROM employees;

This calculates a running total of order amounts ordered by date.

PostgreSQL

SELECT order_id, order_date, amount,
       SUM(amount) OVER (ORDER BY order_date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS running_total
FROM orders;

This gets the previous sale amount for each product by date.

PostgreSQL

SELECT product_id, sale_date, amount,
       LAG(amount, 1) OVER (PARTITION BY product_id ORDER BY sale_date) AS previous_sale
FROM sales;

Sample Program

This example creates a sales table, inserts sample data, and calculates a running total of sales amount for each product ordered by sale date.

PostgreSQL

CREATE TABLE sales (
  sale_id SERIAL PRIMARY KEY,
  product_id INT,
  sale_date DATE,
  amount NUMERIC
);

INSERT INTO sales (product_id, sale_date, amount) VALUES
(1, '2024-01-01', 100),
(1, '2024-01-05', 150),
(1, '2024-01-10', 200),
(2, '2024-01-03', 300),
(2, '2024-01-07', 100);

SELECT product_id, sale_date, amount,
       SUM(amount) OVER (PARTITION BY product_id ORDER BY sale_date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS running_total
FROM sales
ORDER BY product_id, sale_date;

Important Notes

Window functions do not reduce the number of rows returned, unlike GROUP BY.

Ordering inside OVER() is important for functions like ROW_NUMBER(), RANK(), and running totals.

Use PARTITION BY to reset calculations for each group, like per department or product.

Summary

Window functions let you perform calculations across related rows without grouping.

Common patterns include ranking, running totals, and comparing current row to previous rows.

Use PARTITION BY and ORDER BY inside OVER() to control how the window function works.

Practice

1. What does the ROW_NUMBER() OVER (PARTITION BY department ORDER BY salary DESC) window function do in a query?

easy

A. Counts the number of employees in each department.

B. Assigns a unique rank to each employee within their department based on salary, starting at 1 for the highest salary.

C. Calculates the total salary for each department.

D. Returns the average salary of all employees.

Solution

1. Step 1: Understand ROW_NUMBER() with PARTITION BY

   The function assigns a unique sequential number to rows within each department group.

2. Step 2: ORDER BY salary DESC ranks highest salary first

   Ordering by salary descending means the highest salary gets row number 1 in each department.

3. Final Answer:

   Assigns a unique rank to each employee within their department based on salary, starting at 1 for the highest salary. -> Option B

4. Quick Check:

   ROW_NUMBER() with PARTITION BY and ORDER BY = ranked rows [OK]

Hint: ROW_NUMBER() numbers rows per group ordered by criteria [OK]

Common Mistakes:

• Confusing ROW_NUMBER() with COUNT() or SUM()
• Ignoring PARTITION BY and thinking it ranks globally
• Assuming it calculates totals or averages

2. Which of the following is the correct syntax to calculate a running total of sales ordered by date using window functions in PostgreSQL?

easy

A. SELECT date, sales, SUM(sales) PARTITION BY date FROM sales_data;

B. SELECT date, sales, SUM(sales) GROUP BY date ORDER BY running_total FROM sales_data;

C. SELECT date, sales, SUM(sales) WHERE ORDER BY date FROM sales_data;

D. SELECT date, sales, SUM(sales) OVER (ORDER BY date) AS running_total FROM sales_data;

Solution

1. Step 1: Identify correct window function syntax

   SUM(sales) OVER (ORDER BY date) calculates running total by ordering rows by date.

2. Step 2: Check other options for syntax errors

   The incorrect options misuse PARTITION BY without OVER(), use WHERE incorrectly, or confuse GROUP BY with window functions.

3. Final Answer:

   SELECT date, sales, SUM(sales) OVER (ORDER BY date) AS running_total FROM sales_data; -> Option D

4. Quick Check:

   SUM() OVER (ORDER BY ...) = running total syntax [OK]

Hint: Use OVER() with ORDER BY for running totals [OK]

Common Mistakes:

• Using PARTITION BY without OVER()
• Trying to use WHERE inside window function
• Confusing GROUP BY with window functions

3. Given the table sales(date, amount) with rows:
('2024-01-01', 100), ('2024-01-02', 150), ('2024-01-03', 200)
What is the output of this query?

SELECT date, amount, SUM(amount) OVER (ORDER BY date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS running_total FROM sales ORDER BY date;

medium

A. [('2024-01-01', 100, 450), ('2024-01-02', 150, 350), ('2024-01-03', 200, 200)]

B. [('2024-01-01', 100, 100), ('2024-01-02', 150, 150), ('2024-01-03', 200, 200)]

C. [('2024-01-01', 100, 100), ('2024-01-02', 150, 250), ('2024-01-03', 200, 450)]

D. [('2024-01-01', 100, NULL), ('2024-01-02', 150, NULL), ('2024-01-03', 200, NULL)]

Solution

1. Step 1: Understand SUM() OVER with frame

   The frame includes all rows from the start up to current row ordered by date, so it sums cumulatively.

2. Step 2: Calculate running totals row by row

   Row 1: 100; Row 2: 100+150=250; Row 3: 250+200=450.

3. Final Answer:

   [('2024-01-01', 100, 100), ('2024-01-02', 150, 250), ('2024-01-03', 200, 450)] -> Option C

4. Quick Check:

   Running total sums up to current row [OK]

Hint: ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW sums cumulatively [OK]

Common Mistakes:

• Ignoring frame clause and expecting total sum on all rows
• Confusing running total with simple SUM() aggregation
• Assuming NULLs appear without reason

4. You wrote this query to rank employees by salary within each department:

SELECT employee_id, department, salary, RANK() OVER (ORDER BY salary DESC) AS dept_rank FROM employees;

But the query returns the same rank for all employees regardless of department. What is the likely mistake?

medium

A. Missing PARTITION BY clause inside OVER()

B. ORDER BY salary ASC instead of DESC

C. Using RANK() instead of ROW_NUMBER()

D. Not ordering the final result by department

Solution

1. Step 1: Check PARTITION BY usage

   If PARTITION BY is missing or incorrect, ranking happens over entire table, ignoring departments.

2. Step 2: Confirm ORDER BY and function choice

   ORDER BY DESC is correct for highest salary first; RANK() is appropriate for ties; ordering final result doesn't affect ranking.

3. Final Answer:

   Missing PARTITION BY clause inside OVER() -> Option A

4. Quick Check:

   PARTITION BY groups rows for ranking [OK]

Hint: Always include PARTITION BY to rank within groups [OK]

Common Mistakes:

• Forgetting PARTITION BY causes global ranking
• Confusing RANK() and ROW_NUMBER() effects
• Thinking ORDER BY outside OVER() affects ranking

5. You want to find each employee's salary difference compared to the previous employee in the same department ordered by salary descending. Which query correctly uses window functions to achieve this?

hard

A. SELECT employee_id, department, salary, salary - LAG(salary) OVER (PARTITION BY department ORDER BY salary DESC) AS salary_diff FROM employees;

B. SELECT employee_id, department, salary, salary - LEAD(salary) OVER (ORDER BY salary) AS salary_diff FROM employees;

C. SELECT employee_id, department, salary, salary - LAG(salary) FROM employees GROUP BY department ORDER BY salary DESC;

D. SELECT employee_id, department, salary, salary - FIRST_VALUE(salary) OVER (PARTITION BY department) AS salary_diff FROM employees;

Solution

1. Step 1: Use LAG() to get previous row's salary per department

   LAG(salary) OVER (PARTITION BY department ORDER BY salary DESC) gets previous salary in descending order within department.

2. Step 2: Subtract previous salary from current salary

   salary - LAG(salary) calculates difference compared to previous employee.

3. Final Answer:

   SELECT employee_id, department, salary, salary - LAG(salary) OVER (PARTITION BY department ORDER BY salary DESC) AS salary_diff FROM employees; -> Option A

4. Quick Check:

   LAG() gets previous row value for difference [OK]

Hint: Use LAG() with PARTITION BY and ORDER BY for previous row values [OK]

Common Mistakes:

• Using LEAD() instead of LAG() for previous row
• Omitting PARTITION BY causing wrong grouping
• Trying to use aggregate functions without OVER()