Start with entity operations

A thin coroutine extension layer

The public functions in bluetape4k-spring-boot-r2dbc are Kotlin extensions on Spring Data R2DBC operation interfaces. They do not introduce a client or repository base class. Instead, they add read and write functions to an already configured R2dbcEntityOperations.

@Repository
class PostRepository(
    private val operations: R2dbcEntityOperations,
) {
    suspend fun findById(id: Long): Post? =
        operations.findOneByIdOrNullSuspending(id)

    fun findAll(): Flow<Post> =
        operations.selectAllSuspending()
}

Spring Data still owns entity metadata and row mapping. The extensions adapt Reactor results to coroutine types, so mapping annotations, naming strategies, converters, and transaction binding retain their Spring Data behavior.

Receiver types

Receiver	Operations added
R2dbcEntityOperations	ID lookup, conditional select, count, and exists
ReactiveInsertOperation	Entity insert
ReactiveUpdateOperation	Update with Query and Update
ReactiveDeleteOperation	Conditional or full delete

Spring’s R2dbcEntityTemplate normally implements these interfaces, allowing one bean to use all the extensions. The extensions do not own or close the receiver lifecycle.

ID lookup helpers

The ID helpers build a Criteria with the default name id and delegate to the general select functions.

val post: Post = operations.findOneByIdSuspending(1L)
val missing: Post? = operations.findOneByIdOrNullSuspending(-1L)

val first: LegacyPost =
    operations.findFirstByIdSuspending(1L, LegacyPost::postId.name)

idName is the name used in the query. When the property and physical column differ, confirm the Spring Data mapping rule and pass the intended name explicitly. The 1.11.0 tests cover both the default id and Post::id.name forms.

Exactly one versus first

findOneById* delegates to selectOne*, so duplicate rows are not accepted. findFirstById* consumes only the first result. A true unique ID usually benefits from the one contract because it exposes invalid data. Use first for a non-unique, explicitly sorted query that intentionally selects one leading result.

The nullable variants map only an absent row to null. They do not convert database, mapping, or connection failures to null.

What the module does not own

• Driver and ConnectionFactory selection
• Connection pool creation and shutdown
• Schema migration and seed data
• Transaction manager and transaction boundaries
• Raw SQL and custom row mapping

Those responsibilities remain with Spring Boot, Spring Data R2DBC, or bluetape4k-r2dbc. The artifact name is not evidence of auto-configuration.

Sources and tests

• R2dbcEntityOperationExtensions.kt
• ReactiveSelectOperationExtensions.kt
• PostRepository.kt
• PostRepositoryTest.kt

Next chapter

Continue to Flow and result cardinality.