Spring Data JPA repositories

Key points

• Extend JpaRepository<Book, Long> and Spring generates the implementation at start-up. There is no class for you to write.
• A method name is a query. findByAuthorNameAndPagesGreaterThan resolves against the entity model, so a typo fails at boot.
• Past three or four conditions the name stops reading like English. Move to @Query with JPQL and give the method a short name.
• save() on an entity with an id calls merge, which may SELECT first. On a new one it calls persist.
• Return Optional, a projection or a DTO. Handing entities to a controller drags lazy proxies into serialization.
• A repository is not worth mocking. Test it against a real database - the interesting bugs are in the SQL it generated.

Example

public interface BookRepository extends JpaRepository<Book, Long> {

    // derived: parsed from the name at start-up, no body needed
    Optional<Book> findByIsbn(String isbn);

    List<Book> findByAuthorNameAndPagesGreaterThan(String name, int pages);

    // the name would be unreadable, so say it in JPQL instead
    @Query("select b from Book b join fetch b.author " +
           "where b.publishedOn > :since order by b.publishedOn desc")
    List<Book> recentWithAuthor(@Param("since") LocalDate since);

    // read model: no entities, no lazy proxies, only two columns
    @Query("select new com.app.BookView(b.title, a.name) " +
           "from Book b join b.author a")
    List<BookView> catalogue();

    @Modifying
    @Query("update Book b set b.archived = true where b.publishedOn < :cut")
    int archiveOlderThan(@Param("cut") LocalDate cut);
}

The method name is the query. When the name turns ugly, write the JPQL.