Window frame specification (ROWS BETWEEN) in SQL

Introduction

Window frame specification lets you control which rows are included when calculating values over a group of rows in a table.

When you want to calculate running totals in a sales report.

When you need to find the average of the last 3 days' temperatures for each day.

When you want to compare each row's value with the previous two rows in a list.

When you want to calculate a moving average or sum over a specific range of rows.

Syntax

SQL

OVER (ORDER BY column_name ROWS BETWEEN frame_start AND frame_end)

The frame_start and frame_end define the range of rows to include relative to the current row.

You can use keywords like UNBOUNDED PRECEDING, CURRENT ROW, or a number of rows before or after.

Examples

Calculates a running total of sales from the first row up to the current row.

SQL

SUM(sales) OVER (ORDER BY date ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW)

Calculates the average score for the current row and the two rows before it.

SQL

AVG(score) OVER (ORDER BY game_date ROWS BETWEEN 2 PRECEDING AND CURRENT ROW)

Counts the number of rows from the current row to the next three rows.

SQL

COUNT(*) OVER (ORDER BY id ROWS BETWEEN CURRENT ROW AND 3 FOLLOWING)

Sample Program

This query calculates the sum of the current sale and the two previous sales ordered by date, showing a running sum over the last 3 rows.

SQL

CREATE TABLE sales_data (
  id INT,
  sale_date DATE,
  amount INT
);

INSERT INTO sales_data VALUES
(1, '2024-01-01', 100),
(2, '2024-01-02', 150),
(3, '2024-01-03', 200),
(4, '2024-01-04', 50),
(5, '2024-01-05', 300);

SELECT
  id,
  sale_date,
  amount,
  SUM(amount) OVER (ORDER BY sale_date ROWS BETWEEN 2 PRECEDING AND CURRENT ROW) AS running_sum_last_3
FROM sales_data
ORDER BY sale_date;

Important Notes

The frame moves with each row, so calculations update as you go down the table.

ROWS BETWEEN counts physical rows, not values, so duplicates are included as separate rows.

If you omit ROWS BETWEEN, the default frame is RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW, which behaves differently.

Summary

Window frame specification controls which rows are included in window function calculations.

ROWS BETWEEN lets you specify a range of rows relative to the current row.

It is useful for running totals, moving averages, and other calculations over a sliding window.

Practice

1. What does the ROWS BETWEEN clause specify in a window function?

easy

A. The range of rows to include relative to the current row for calculation

B. The order in which rows are sorted before aggregation

C. The columns to group by in the query

D. The filter condition for rows in the table

Solution

1. Step 1: Understand window function frame

   The window frame defines which rows are considered for each calculation in a window function.

2. Step 2: Role of ROWS BETWEEN

   The ROWS BETWEEN clause specifies the exact range of rows relative to the current row to include in the calculation.

3. Final Answer:

   The range of rows to include relative to the current row for calculation -> Option A

4. Quick Check:

   ROWS BETWEEN controls row range = D [OK]

Hint: ROWS BETWEEN defines which rows to include around current row [OK]

Common Mistakes:

• Confusing ROWS BETWEEN with ORDER BY
• Thinking it filters rows from the table
• Mixing it up with GROUP BY

2. Which of the following is the correct syntax to specify a window frame from 2 rows before to current row using ROWS BETWEEN?

easy

A. ROWS BETWEEN 2 PRECEDING AND CURRENT ROW

B. ROWS BETWEEN CURRENT ROW AND 2 PRECEDING

C. ROWS BETWEEN 2 FOLLOWING AND CURRENT ROW

D. ROWS BETWEEN CURRENT ROW AND 2 FOLLOWING

Solution

1. Step 1: Understand frame boundaries order

   The syntax is ROWS BETWEEN X AND Y where X is the start and Y is the end of the frame.

2. Step 2: Identify correct order for 2 rows before to current

   Since 2 rows before is the start and current row is the end, it must be ROWS BETWEEN 2 PRECEDING AND CURRENT ROW.

3. Final Answer:

   ROWS BETWEEN 2 PRECEDING AND CURRENT ROW -> Option A

4. Quick Check:

   Start to end order matters = A [OK]

Hint: Start boundary comes first, then end boundary in ROWS BETWEEN [OK]

Common Mistakes:

• Reversing the order of boundaries
• Using FOLLOWING instead of PRECEDING incorrectly
• Omitting CURRENT ROW keyword

3. Given the table sales(date, amount) ordered by date, what is the output of this query?

SELECT date, amount, SUM(amount) OVER (ORDER BY date ROWS BETWEEN 1 PRECEDING AND CURRENT ROW) AS running_sum FROM sales;

Assuming data:
date | amount
2024-01-01 | 100
2024-01-02 | 200
2024-01-03 | 300

medium

A. [{"date": "2024-01-01", "amount": 100, "running_sum": 300}, {"date": "2024-01-02", "amount": 200, "running_sum": 500}, {"date": "2024-01-03", "amount": 300, "running_sum": 600}]

B. [{"date": "2024-01-01", "amount": 100, "running_sum": 100}, {"date": "2024-01-02", "amount": 200, "running_sum": 200}, {"date": "2024-01-03", "amount": 300, "running_sum": 300}]

C. [{"date": "2024-01-01", "amount": 100, "running_sum": 100}, {"date": "2024-01-02", "amount": 200, "running_sum": 300}, {"date": "2024-01-03", "amount": 300, "running_sum": 500}]

D. [{"date": "2024-01-01", "amount": 100, "running_sum": null}, {"date": "2024-01-02", "amount": 200, "running_sum": 300}, {"date": "2024-01-03", "amount": 300, "running_sum": 500}]

Solution

1. Step 1: Understand frame range

   The frame includes the current row and 1 row before it, so for each row sum amounts of current and previous row.

2. Step 2: Calculate running sums

   Row 1: only 100 (no previous row), Row 2: 100+200=300, Row 3: 200+300=500.

3. Final Answer:

   [{"date": "2024-01-01", "amount": 100, "running_sum": 100}, {"date": "2024-01-02", "amount": 200, "running_sum": 300}, {"date": "2024-01-03", "amount": 300, "running_sum": 500}] -> Option C

4. Quick Check:

   Sum current + 1 preceding = A [OK]

Hint: Sum includes current and 1 previous row amounts [OK]

Common Mistakes:

• Including 2 rows before instead of 1
• Ignoring the order of rows
• Assuming NULL for first row sum

4. Identify the error in this SQL window function:

SELECT id, value, AVG(value) OVER (ORDER BY id ROWS BETWEEN CURRENT ROW AND 2 PRECEDING) AS avg_val FROM data;

medium

A. ORDER BY cannot be used with ROWS BETWEEN

B. The frame boundaries are reversed; start must be before end

C. Missing PARTITION BY clause

D. AVG cannot be used as a window function

Solution

1. Step 1: Check frame boundaries order

   The frame must start with the earlier row and end with the later row. Here, CURRENT ROW is after 2 PRECEDING, so order is reversed.

2. Step 2: Identify correct syntax

   The correct syntax is ROWS BETWEEN 2 PRECEDING AND CURRENT ROW, not the other way around.

3. Final Answer:

   The frame boundaries are reversed; start must be before end -> Option B

4. Quick Check:

   Frame start must be before end = C [OK]

Hint: Start boundary must be earlier than end boundary [OK]

Common Mistakes:

• Swapping start and end boundaries
• Assuming PARTITION BY is mandatory
• Thinking ORDER BY disallows ROWS BETWEEN

5. You want to calculate a moving average of the last 3 sales amounts including the current row, but your data has gaps in dates. Which window frame specification ensures exactly 3 rows are included regardless of date gaps?

hard

A. RANGE BETWEEN INTERVAL '2' DAY PRECEDING AND CURRENT ROW

B. RANGE BETWEEN 2 PRECEDING AND CURRENT ROW

C. ROWS BETWEEN CURRENT ROW AND 2 FOLLOWING

D. ROWS BETWEEN 2 PRECEDING AND CURRENT ROW

Solution

1. Step 1: Understand difference between ROWS and RANGE

   ROWS counts physical rows, while RANGE counts logical values like dates, which can skip rows if dates have gaps.

2. Step 2: Choose frame to include 3 rows

   Using ROWS BETWEEN 2 PRECEDING AND CURRENT ROW always includes exactly 3 rows regardless of date gaps.

3. Final Answer:

   ROWS BETWEEN 2 PRECEDING AND CURRENT ROW -> Option D

4. Quick Check:

   ROWS counts rows, RANGE counts values = B [OK]

Hint: Use ROWS to count exact rows, not RANGE for gaps [OK]

Common Mistakes:

• Using RANGE which depends on value gaps
• Confusing FOLLOWING with PRECEDING
• Assuming RANGE counts rows