Hibernate N+1 Demo (Spring Boot 3 / Hibernate 6)

A tiny, copy‑pastable project that reproduces the N+1 problem and shows 3 safe fixes:

• Baseline (trigger N+1): findAll() then access author.getBooks() in a loop.
• Fix #1: JOIN FETCH query
• Fix #2: @EntityGraph(attributePaths = "books")
• (Bonus) hibernate.default_batch_fetch_size is configured to help with to‑one proxies.

How to run

mvn -v          # Java 21 + Maven 3.9+ recommended
mvn test -q

The tests assert query counts using Hibernate Statistics:

• baselineShowsNPlusOne() expects >= 6 statements with our seed data (5 authors).
• joinFetchFixUsesSingleQuery() and entityGraphFixUsesSingleQuery() expect <= 2 statements.

Key files

• Author ↔ Book with @OneToMany / @ManyToOne (LAZY)
• AuthorRepository with:
  • findAllWithBooksJoinFetch() — select distinct a from Author a left join fetch a.books
  • findAllWithBooksGraph() — @EntityGraph(attributePaths="books")
  • IDs-then-fetch helpers to implement paging without N+1
• DemoService — methods that reproduce/fix N+1 and expose a helper to read the current prepareStatement count
• application.yml — open-in-view: false, SQL logging, and hibernate.default_batch_fetch_size: 50
• application-test.yml — hibernate.generate_statistics: true
• data.sql — seeds 5 authors × 3 books

Extending (pagination-safe two-step)

When you need pagination with a to-many association, use the two-step pattern:

1. Page parent IDs (Page<Long>)
2. Fetch the graph with join fetch for where id in :ids

See AuthorRepository.pageIdsByNamePrefix(...) and fetchGraphByIds(ids).

Why distinct?

Join fetching a collection duplicates parent rows in SQL. select distinct avoids exploding row counts and lets Hibernate deduplicate entities safely.

---

Happy debugging! If you want me to wire this into your existing service with concrete entity names, say the word.