persistence-graph-extraction-jdbc

Demonstrates how to assemble an object graph from a relational database using plain JDBC (JdbcClient) without ORM or pagination.

What it shows

• Assembling a three-level object graph (Owner → Pet → Visit) from flat SQL JOIN results
• Using package-private classes as an encapsulation boundary inside the infrastructure layer
• Separating persistence projections from application-level read models via ViewMapper
• Two query strategies:
  • full graph — Owner with Pets and Visits (OwnerView)
  • flat list — Owner with pet names only (OwnerListView)

Stack

• Java 25
• Spring Boot
• Spring JDBC (JdbcClient)
• H2 (in-memory database)

Structure

application/
    OwnerReadRepository       # port (interface)
    view/                     # read models: OwnerView, OwnerListView, PetView, VisitView

infrastructure/
    JdbcOwnerReadRepository   # JDBC implementation (package-private)
    OwnerProjectionExtractor  # assembles graph from ResultSet (package-private)
    ViewMapper                # maps projections to read models (package-private)
    OwnerProjection           # mutable accumulator (package-private)
    PetProjection             # mutable accumulator (package-private)
    VisitProjection           # mutable accumulator (package-private)
    OwnerListProjection       # mutable accumulator (package-private)

Key idea

A LEFT JOIN across three tables returns flat rows with duplicates:

owner_id | owner_name | pet_id | pet_name | visit_id | visit_date
---------|------------|--------|----------|----------|------------
1        | jack       | 1      | buddy1   | 1        | 2026-01-10
1        | jack       | 1      | buddy1   | 2        | 2026-02-15
1        | jack       | 2      | buddy2   | 3        | 2026-03-01
3        | bob        | 4      | hew1     | null     | null

OwnerProjectionExtractor reassembles them into an object graph in a single pass:

• computeIfAbsent — get or create Owner
• getOrCreatePet — get or create Pet under that Owner
• getOrCreateVisit — add Visit if visit_id is not null

Owner(jack)
  └── Pet(buddy1)
        ├── Visit(2026-01-10)
        └── Visit(2026-02-15)
  └── Pet(buddy2)
        └── Visit(2026-03-01)

Compare with the jOOQ and JPA + Blaze approaches:

	JDBC	jOOQ	JPA + Blaze
SQL result	flat rows (cartesian JOIN)	nested collections	optimized per-view query
Deduplication	LinkedHashMap in projections	not needed	not needed
Intermediate classes	OwnerProjection, PetProjection, etc.	none	@EntityView interfaces
Mapping	ViewMapper	Records.mapping()	ViewMapper
N+1 problem	no (single JOIN query)	no (single query)	no (Blaze optimizes automatically)

No ORM, no JSON parsing, no nested queries — just plain JDBC and a single pass.

All internal classes are package-private by design; only OwnerReadRepository (the interface) is exposed outside the infrastructure package.

Tests

Integration and unit tests in src/test/java cover graph assembly, flat list extraction, and view mapping.

Related

• persistence-graph-extraction-jooq — same approach with jOOQ MULTISET
• persistence-graph-extraction-jpa — same approach with JPA + Blaze Persistence
• persistence-graph-pagination-jdbc — same approach with pagination support

Run

./mvnw spring-boot:run