save() method on repository not working correctly

[kylooh]

Device: MacOS with M1 chip
Language: Kotlin
Spring R2dbc Version: spring-boot-starter-data-r2dbc:2.6.3
Driver Version: r2dbc-mysql:0.8.2.REALEASE

Here is my codes.

@Repository
interface ProductR2Repository : CoroutineCrudRepository<ProductEntity, String> {

}

@Table(value = "product")
data class ProductEntity(
    @Id val code: String,
    val name: String,
    val expectedReturn: Double,
    val expectedRisk: Double,
    val creator: String,
    val createTime: String,
    val updateTime: String
)

It throws Exception in thread "main" ApplicationErrorException (0x201): Failed to update table [product]. Row with Id [testing-0] does not exist.

And I found the problem is, save() method will pre-check entity's isNew() status, then insert() if new, update() if not. And PersistentEntity's isNew() method doing following behaviors.

1. find entity's id field which annotation by @Id. If id is null, returns true. Which means entity without @Id is always new.
2. check id is null or a primitive type. If it is not null or primitive, returns false, which means using update. Which means entity without any id field is always new.
3. if id is not null, and also a number value, and also equals 0, returning true. Which means entity id equals 0 is always new.

I used JPA before, so I'm hardly understand why exists check using that way, instead of do check exists like using existsById() method. And also I removed @Id from entity, but it will throws exception

Exception in thread "main" ApplicationErrorException (0x201): Required identifier property not found for class com.betalpha.w4.infra.r2dbc.product.info.simulate.ProductEntity!

And If I using null as id's value, it throws

nested exception is io.r2dbc.spi.R2dbcDataIntegrityViolationException: [23502] [23502] NULL not allowed for column "CODE";

Anyone has any idea for this problems? Help, please

[schauder]

You have multiple options dealing with Inserts and IDs in R2DBC.

If your entity has a non-null id, but you know it is new, you can perform an insert using the R2dbcEntityTemplate.insert() operation.

Or you can leave the id null until the isNew check was performed and then set the id to your preferred value using a callback

Or you could add a field annotated with @Version for optimistic locking, which then would also be used for the is-new check.

Or you could let your entity implement Persistable and thereby control what is considered new and what isn't.

[roshan1892]

@schauder
I think this needs to be supported by Spring Data r2dbc by default just like how Spring Data JPA does, isn't it? The things that you suggested are just workarounds. This more looks like a bug.

[schauder]

These are not workarounds, but the intended behavior. Just because JPA does something doesn't mean it is a good idea.

[roshan1892]

@schauder I think a lot of people are already familiar with JPA and are just starting to try out R2DBC, so keeping the workflow similar to JPA would make things easier unless there's a really good reason not to.