tudose-spring-data-hibernate

Keep all associations LAZY globally, then fix n+1 per use case with a join fetch JPQL query, an entity graph, @BatchSize(N), or @Fetch(SUBSELECT). Do not flip mappings to EAGER, and never eager-fetch two collections at once, which causes a Cartesian product.

Use TABLE_PER_CLASS when no polymorphism is needed, SINGLE_TABLE when subclasses differ mainly in behavior (accepting nullable columns), and JOINED when subclasses hold significant data requiring NOT NULL constraints. Benchmark JOINED versus TABLE_PER_CLASS unions with realistic data.

The default recipe is READ_COMMITTED isolation plus optimistic locking with a numeric @Version field, giving first-commit-wins semantics without holding locks. Escalate to OPTIMISTIC_FORCE_INCREMENT for aggregate roots or PESSIMISTIC_WRITE only for hot rows.

Choose Spring Data JPA for full JPA power, derived queries, and productivity. Choose Spring Data JDBC for simple aggregates when you want predictable SQL without a persistence context — no lazy loading, dirty checking, caching, or JPQL — using its own annotations, never JPA's.

The generated id is null before flush, so id-based equality breaks for transient instances, and default identity equality breaks across persistence contexts. Implement equals and hashCode on an immutable unique business key using instanceof and getters for proxy safety.

The skill covers only the book's content: Spring Data and Hibernate persistence patterns as taught by Tudose. Topics beyond the book, such as reactive R2DBC or Hibernate Reactive, are out of scope and require other resources.