Sub-partitioning in PostgreSQL

Introduction

Sub-partitioning helps organize large tables into smaller, more manageable pieces by dividing data twice, making queries faster and maintenance easier.

When you have a big table with data that can be grouped by two different categories, like region and year.

When you want to speed up searches by narrowing down data to smaller parts.

When you need to archive or delete old data efficiently without affecting recent data.

When managing data that grows quickly and needs to be split for better performance.

When you want to improve backup and restore times by working on smaller data chunks.

Syntax

PostgreSQL

CREATE TABLE table_name (
  column_definitions
) PARTITION BY partition_method1 (column1);

CREATE TABLE partition_name PARTITION OF table_name
  FOR VALUES partition_values
  PARTITION BY partition_method2 (column2);

CREATE TABLE subpartition_name PARTITION OF partition_name
  FOR VALUES subpartition_values;

partition_method1 and partition_method2 can be RANGE, LIST, or HASH.

You first define the parent table partitioned by the first key, then define its partitions which are themselves partitioned by the second key.

Examples

This creates a sales table partitioned by region, then subpartitioned by sale date ranges.

PostgreSQL

CREATE TABLE sales (
  id SERIAL,
  region TEXT,
  sale_date DATE,
  amount NUMERIC
) PARTITION BY LIST (region);

Defines the 'East' region partition with subpartitioning by sale_date.

PostgreSQL

CREATE TABLE sales_east PARTITION OF sales
  FOR VALUES IN ('East')
  PARTITION BY RANGE (sale_date);

Defines a subpartition for sales in the East region for the year 2023.

PostgreSQL

CREATE TABLE sales_east_2023 PARTITION OF sales_east
  FOR VALUES FROM ('2023-01-01') TO ('2024-01-01');

Sample Program

This example creates an orders table partitioned by region and subpartitioned by order date. It inserts two orders in the North region for 2023 and then selects them.

PostgreSQL

CREATE TABLE orders (
  order_id SERIAL PRIMARY KEY,
  customer_id INT,
  order_region TEXT,
  order_date DATE,
  total NUMERIC
) PARTITION BY LIST (order_region);

CREATE TABLE orders_north PARTITION OF orders
  FOR VALUES IN ('North')
  PARTITION BY RANGE (order_date);

CREATE TABLE orders_north_2023 PARTITION OF orders_north
  FOR VALUES FROM ('2023-01-01') TO ('2024-01-01');

INSERT INTO orders (customer_id, order_region, order_date, total) VALUES
  (1, 'North', '2023-05-10', 100.00),
  (2, 'North', '2023-11-20', 250.50);

SELECT * FROM orders WHERE order_region = 'North' AND order_date >= '2023-01-01' AND order_date < '2024-01-01';

Important Notes

Sub-partitioning can improve query speed but adds complexity to table design.

Make sure your queries include the partition keys to benefit from partition pruning.

PostgreSQL supports multi-level partitioning (sub-partitioning) in version 10 or higher.

Summary

Sub-partitioning splits data twice for better organization and performance.

Use LIST, RANGE, or HASH methods for both partitions and subpartitions.

Define the parent table first, then create its partitions, then subpartitions of those.

Practice

1. What is the main purpose of sub-partitioning in PostgreSQL?

easy

A. To encrypt data within partitions

B. To create backups of partitions automatically

C. To merge multiple partitions into one

D. To split data twice for better organization and faster queries

Solution

1. Step 1: Understand partitioning basics

   Partitioning divides a table into parts to improve management and performance.

2. Step 2: Recognize sub-partitioning role

   Sub-partitioning splits each partition further, organizing data more finely and speeding up queries.

3. Final Answer:

   To split data twice for better organization and faster queries -> Option D

4. Quick Check:

   Sub-partitioning = double data split [OK]

Hint: Sub-partitioning means splitting partitions again [OK]

Common Mistakes:

• Thinking sub-partitioning creates backups
• Confusing sub-partitioning with encryption
• Believing it merges partitions

2. Which of the following is the correct syntax to create a sub-partitioned table in PostgreSQL?

easy

A. CREATE TABLE sales (id INT, region TEXT, month INT) SUBPARTITION BY RANGE (region) PARTITION BY LIST (month);

B. CREATE TABLE sales (id INT, region TEXT, month INT) PARTITION BY RANGE (region) PARTITION BY LIST (month);

C. CREATE TABLE sales (id INT, region TEXT, month INT) PARTITION BY RANGE (region) SUBPARTITION BY LIST (month);

D. CREATE TABLE sales (id INT, region TEXT, month INT) PARTITION BY LIST (region) SUBPARTITION BY HASH (month);

Solution

1. Step 1: Identify correct keywords for partitioning

   PostgreSQL uses PARTITION BY for main partition and SUBPARTITION BY for sub-partition.

2. Step 2: Check syntax order and clauses

   CREATE TABLE sales (id INT, region TEXT, month INT) PARTITION BY RANGE (region) SUBPARTITION BY LIST (month); correctly uses PARTITION BY RANGE then SUBPARTITION BY LIST, matching PostgreSQL syntax.

3. Final Answer:

   CREATE TABLE sales (id INT, region TEXT, month INT) PARTITION BY RANGE (region) SUBPARTITION BY LIST (month); -> Option C

4. Quick Check:

   Use PARTITION BY then SUBPARTITION BY [OK]

Hint: Use PARTITION BY first, then SUBPARTITION BY [OK]

Common Mistakes:

• Using PARTITION BY twice instead of SUBPARTITION BY
• Swapping PARTITION BY and SUBPARTITION BY keywords
• Using SUBPARTITION BY before PARTITION BY

3. Given the following table and partitions:

CREATE TABLE orders (id INT, country TEXT, year INT) PARTITION BY LIST (country) SUBPARTITION BY RANGE (year);
CREATE TABLE orders_us PARTITION OF orders FOR VALUES IN ('US') SUBPARTITION BY RANGE (year);
CREATE TABLE orders_us_2022 PARTITION OF orders_us FOR VALUES FROM (2022) TO (2023);

What will be the result of SELECT * FROM orders WHERE country = 'US' AND year = 2022; if there are rows with country 'US' and year 2022?

medium

A. Rows with country 'US' and year 2022 will be returned

B. No rows will be returned because subpartition is missing

C. Syntax error due to incorrect partitioning

D. Rows with any country but year 2022 will be returned

Solution

1. Step 1: Understand partition and subpartition setup

   The table is partitioned by country (LIST) and subpartitioned by year (RANGE). The 'US' partition and 2022 subpartition exist.

2. Step 2: Query filters match partition and subpartition

   The query filters country='US' and year=2022, matching the defined partitions, so matching rows will be found.

3. Final Answer:

   Rows with country 'US' and year 2022 will be returned -> Option A

4. Quick Check:

   Partition + subpartition match = rows returned [OK]

Hint: Query matches partition and subpartition filters [OK]

Common Mistakes:

• Assuming no rows because subpartition is complex
• Thinking query causes syntax error
• Ignoring subpartition filtering

4. You wrote this code:

CREATE TABLE logs (id INT, region TEXT, day DATE) PARTITION BY RANGE (region) SUBPARTITION BY LIST (day);

What is the error in this statement?

medium

A. RANGE partitioning cannot be done on a TEXT column

B. Partitioning by RANGE requires a numeric or date type, not TEXT

C. Syntax error: SUBPARTITION BY must come before PARTITION BY

D. SUBPARTITION BY LIST cannot be used with RANGE partitioning

Solution

1. Step 1: Check partition column data type

   Partitioning by RANGE requires a column with an orderable type like numeric or date, not TEXT.

2. Step 2: Identify the error cause

   Here, region is TEXT, so RANGE partitioning on it is invalid.

3. Final Answer:

   Partitioning by RANGE requires a numeric or date type, not TEXT -> Option B

4. Quick Check:

   RANGE needs numeric/date, not TEXT [OK]

Hint: RANGE partition needs numeric or date column [OK]

Common Mistakes:

• Thinking TEXT can be used for RANGE partitioning
• Confusing order of PARTITION BY and SUBPARTITION BY
• Assuming SUBPARTITION BY LIST is invalid with RANGE

5. You want to create a sales table partitioned by region (LIST) and subpartitioned by sale_date (RANGE). Which approach correctly handles the subpartitioning to optimize query performance for recent sales?

hard

A. Partition by LIST on region, then subpartition by RANGE on sale_date with recent years as separate subpartitions

B. Partition by RANGE on sale_date, then subpartition by LIST on region with all regions in one subpartition

C. Partition by HASH on region, no subpartitioning needed for sale_date

D. Partition by LIST on sale_date, then subpartition by RANGE on region

Solution

1. Step 1: Match partitioning to data and query needs

   Partitioning by region (LIST) groups data by location, then subpartitioning by sale_date (RANGE) organizes by time.

2. Step 2: Optimize recent sales queries

   Using RANGE subpartitions for recent years allows fast access to recent data, improving query speed.

3. Final Answer:

   Partition by LIST on region, then subpartition by RANGE on sale_date with recent years as separate subpartitions -> Option A

4. Quick Check:

   LIST then RANGE for region and date [OK]

Hint: Partition by region LIST, subpartition by date RANGE [OK]

Common Mistakes:

• Reversing partition and subpartition order
• Using HASH partitioning without subpartitioning
• Partitioning sale_date by LIST instead of RANGE