backend-spring-data

name	backend-spring-data
description	Use when reviewing Spring Data JPA code — `@Transactional` placement and propagation, self-invocation, JPQL fetch strategies, `@Modifying` queries, repository patterns, equals/hashCode on entities. Triggers when JPA repositories or `@Transactional` methods are in scope and Spring Boot Data is detected.
spring_boot_versions	["3.x"]

Spring Data Review

Overview

Review Spring Data JPA code for transactional correctness, query patterns, and entity design. Catches Spring-specific issues that the agnostic database-review skill doesn't flag (which is schema/index/N+1-focused).

Triggers when spring-boot-starter-data-* is detected and JPA repositories or @Transactional methods are in scope.

Version Compatibility

Supported: Spring Boot 3.x (declared via spring_boot_versions: ["3.x"] in frontmatter).

Spring Boot 2.x — degraded coverage:

• All checks (P1–P10) apply identically; package names differ (jakarta.persistence.* in SB3 vs javax.persistence.* in SB2)
• @Transactional annotations live in org.springframework.transaction.annotation in both versions — no change

To add full SB2 coverage, follow Pattern A (broaden the rubric) per shield/skills/backend/EXTENDING-VERSIONS.md.

When to Use

• Reviewing @Repository interfaces (Spring Data, custom)
• Reviewing @Transactional placement and propagation
• Reviewing JPA entities for equals/hashCode and lifecycle annotations
• Reviewing JPQL/native queries with @Query/@Modifying

When NOT to Use

• Pure schema design (FKs, indexes, normalization) — use database-review
• N+1 query detection at the schema level — use database-review (D6)
• Non-JPA persistence (Spring Data MongoDB, JDBC, R2DBC) — patterns differ; this skill is JPA-focused

Review Process

1. Inventory: @Transactional annotations, @Repository interfaces, JPA entity classes, @Query methods
2. For each, apply Evaluation Points P1–P10
3. Cross-cut: trace a write path; flag every transactional boundary that is silently missing or misplaced

Evaluation Points

#	Check	What to Look For	Severity
P1	@Transactional on private method	Spring AOP proxies don't intercept private methods. @Transactional on private is a no-op	High
P2	Self-invocation problem	Calling a @Transactional method from within the same bean bypasses the proxy. Refactor to separate bean OR use AopContext.currentProxy() (rarely the right answer)	High
P3	readOnly = true on read methods	Read methods marked @Transactional(readOnly = true) — enables JPA flush optimization, signals intent	Medium
P4	Propagation correctness	REQUIRES_NEW requires explicit justification (e.g., audit log that commits independently). Default REQUIRED is correct most of the time	Medium
P5	@Modifying for mutating @Query	UPDATE/DELETE in @Query requires @Modifying. Without it, Spring treats it as SELECT and silently ignores the change	High
P6	Mutating queries also need @Transactional	@Modifying queries fail without an active transaction. Annotate the caller (or use @Transactional on the repo method)	High
P7	equals/hashCode on entities	Auto-generated ID-based equals breaks before persistence (HashSet bugs). Use natural keys or business identifier; document the choice	Medium
P8	Fetch strategy via @EntityGraph	When findAll/custom queries need to load children, use @EntityGraph(attributePaths=...) rather than EAGER on the entity	Medium
P9	@Repository annotation	Optional on Spring Data interfaces extending JpaRepository (auto-detected), but harmless. Required for plain JPA repos with

Critical Checks

• @Transactional annotation on a private method (always a bug)
• A method calling another @Transactional method on this (self-invocation)
• A @Query containing UPDATE or DELETE without @Modifying
• A @Modifying query on a method outside any @Transactional boundary
• An entity with default Object-identity equals/hashCode used in a HashSet/HashMap

Severity Guide

Severity	When
High	Silent bug — code looks like it works but doesn't (no-op @Transactional, ignored update, lost data)
Medium	Performance/correctness friction (missing readOnly, wrong propagation, equals bugs in collections)
Low	Style — annotation hygiene

Common Mistakes

Mistake	Fix
Demanding @Transactional on every public method	Many service methods are read-only and don't need explicit @Transactional (auto-tx in Spring Data repos handles reads)
Flagging Propagation.REQUIRES_NEW always	Legitimate when the caller's tx must not affect the inner work (audit log, retry logic). The smell is REQUIRES_NEW WITHOUT a comment justifying it
Insisting on @Repository on Spring Data interfaces	Spring Data auto-detects them. Adding @Repository is harmless but not required
Penalizing all Cascade.ALL	Acceptable on parent-owned children where deleting the parent SHOULD delete children (e.g., line items of an invoice)
Treating ID-based equals as always wrong	OK when entities are always persisted before going into collections. The smell is using them in HashSet pre-persistence

Related Skills

• For schema design (FKs, indexes, migrations) → backend-database-review
• For underlying code structure → backend-code-quality-review
• For multi-instance state safety on shared caches → backend-deployment-safety-review