PostgreSQL Performance & Query Optimization · レッスン

結合アルゴリズムを理解する

PostgreSQLがNested Loop、Hash Join、Merge Joinなどの結合方式を実行する仕組みを学びます。

「結合アルゴリズムを理解する」はCoddyKit上の無料PostgreSQL Performance & Query Optimizationレッスンです。これはレッスン1/4です。下記で完全なレッスンを無料で読むことができます。その後、ブラウザ内の組み込みコードエディタと24時間対応のAIチューターでハンズオン演習できます。これはPostgreSQL Performance & Query Optimization学習パスの一部であり、ウェブとCoddyKitアプリ全体で進捗が同期されます。 PostgreSQL Performance & Query Optimizationコースには全4レッスンが含まれています。

このレッスンの一部はまだ翻訳されておらず、英語で表示されています。

Joins: Connecting Data

Welcome to understanding PostgreSQL join algorithms! Joins are fundamental for combining data from multiple tables.

They allow you to retrieve related information that is spread across your database schema, forming a complete picture.

Beyond Basic Joins

When you write a JOIN clause, PostgreSQL doesn't just pick one way to execute it. It has several powerful algorithms at its disposal.

The database's query planner chooses the most efficient algorithm based on factors like table size, available indexes, and data distribution.

Nested Loop Join Basics

The Nested Loop Join (NLJ) is the simplest algorithm. It works like a nested 'for' loop:

For each row in the outer table...
It scans the inner table for matching rows.

NLJ is efficient for small datasets or when the inner table's join column is indexed, allowing quick lookups.

NLJ in Action

Consider joining a small users table with a user_details table. If user_details.user_id is indexed, NLJ can be very fast.

Try creating and joining these tables:

CREATE TABLE users (user_id INT PRIMARY KEY, name VARCHAR(50));
CREATE TABLE user_details (detail_id INT PRIMARY KEY, user_id INT, address VARCHAR(100));

INSERT INTO users VALUES (1, 'Alice'), (2, 'Bob');
INSERT INTO user_details VALUES (101, 1, '123 Main St'), (102, 2, '456 Oak Ave');

SELECT u.name, ud.address
FROM users u
JOIN user_details ud ON u.user_id = ud.user_id;

Hash Join: Faster Matches

Hash Join is often chosen for larger, unsorted tables, especially with equality (=) join conditions. It works in two phases:

Build Phase: PostgreSQL scans the smaller (or estimated smaller) table and builds an in-memory hash table using the join key.
Probe Phase: It scans the larger table, hashes each row's join key, and probes the hash table for matches.

This method is very effective when enough memory is available for the hash table.

Hash Join Scenario

Imagine joining two large tables, products and sales, on their product_id. If neither table is sorted or indexed on product_id, a Hash Join is a strong candidate.

The planner will likely choose Hash Join for this query:

CREATE TABLE products (product_id INT PRIMARY KEY, name VARCHAR(50));
CREATE TABLE sales (sale_id INT PRIMARY KEY, product_id INT, quantity INT);

INSERT INTO products VALUES (1, 'Laptop'), (2, 'Mouse');
INSERT INTO sales VALUES (1001, 1, 2), (1002, 2, 1), (1003, 1, 3);

SELECT p.name, s.quantity
FROM products p
JOIN sales s ON p.product_id = s.product_id;

Merge Join: Sorted Efficiency

The Merge Join is highly efficient when both tables are already sorted on their join keys, or can be sorted cheaply. It also works in phases:

Sort Phase: If not already sorted, both tables are sorted on their join columns.
Merge Phase: PostgreSQL simultaneously scans both sorted tables, merging matching rows. It's like merging two sorted lists.

This is beneficial for range joins or when data is retrieved in sorted order.

Merge Join Use Case

If you're joining two tables, employees and departments, and both are indexed (and thus often sorted) on their respective ID columns, or if your query involves an ORDER BY on the join key, a Merge Join can be optimal.

PostgreSQL might use Merge Join here:

CREATE TABLE employees (emp_id INT PRIMARY KEY, dept_id INT, name VARCHAR(50));
CREATE TABLE departments (dept_id INT PRIMARY KEY, dept_name VARCHAR(50));

INSERT INTO employees VALUES (1, 10, 'John'), (2, 20, 'Jane');
INSERT INTO departments VALUES (10, 'HR'), (20, 'IT');

SELECT e.name, d.dept_name
FROM employees e
JOIN departments d ON e.dept_id = d.dept_id
ORDER BY e.emp_id;

PostgreSQL's Decisions

The PostgreSQL query planner uses a cost-based optimizer to decide which join algorithm to use. It estimates the cost of each possible plan based on:

Table and index statistics
Available memory (work_mem)
Join condition type (e.g., equality, range)
Estimated row counts

Using EXPLAIN is crucial to see which algorithm the planner chose!

Algorithm Challenge

You need to join two very large tables, customers and orders, on customer_id. There are no indexes on customer_id in either table, and the data is unsorted. Which join algorithm is PostgreSQL most likely to choose for optimal performance?

Join Algorithms: Key Takeaways

In this lesson, you explored the three primary join algorithms PostgreSQL uses:

Nested Loop Join: Simple, good for small sets or indexed inner tables.
Hash Join: Efficient for large, unsorted tables with equality joins, using a hash table.
Merge Join: Best when tables are already sorted on join keys or can be sorted cheaply.

Understanding these helps you interpret EXPLAIN plans and write more performant queries. Next, we'll look at rewriting complex joins!

よくある質問

「結合アルゴリズムを理解する」レッスンは無料ですか？

はい。「結合アルゴリズムを理解する」の完全なテキストはこのウェブで無料で読めます。インタラクティブに演習し（組み込みコードエディタと24時間対応のAIチューター）、PostgreSQL Performance & Query Optimizationコースの残りをアンロックするには、CoddyKit PROにアップグレードしてください。 PostgreSQL Performance & Query Optimizationコースには全4レッスンが含まれています。

「結合アルゴリズムを理解する」で何を学びますか？

PostgreSQLがNested Loop、Hash Join、Merge Joinなどの結合方式を実行する仕組みを学びます。ブラウザで直接実行するハンズオンコードでPostgreSQL Performance & Query Optimizationを演習し、24時間対応のAIチューターがレッスンを進める中での質問に答えます。

PostgreSQL Performance & Query Optimizationを始めるのに経験は必要ですか？

事前経験は必要ありません。CoddyKitのPostgreSQL Performance & Query Optimizationは初級者から上級者向けに構成されているため、ここから始めるか最初から始めて、自分のペースで進むことができます。これはレッスン1/4です。

「結合アルゴリズムを理解する」レッスンにはどのくらい時間がかかりますか？

ほとんどのCoddyKitレッスンは約5～10分かかります。各レッスンはコンパクトでインタラクティブなので、着実に進歩し、ウェブとアプリ全体で正確に前回の場所から再開できます。

このPostgreSQL Performance & Query Optimizationレッスンでコードを書いて実行できますか？

はい。すべてのPostgreSQL Performance & Query Optimizationレッスンに組み込みコードエディタが含まれているため、ブラウザでリアルコードを書いて実行し、即座のAIフィードバックを取得できます。ローカル設定は不要です。

このコースのすべてのレッスン

← PostgreSQL Performance & Query Optimizationに戻る