Allow consumption of generated values beyond generated keys upon `INSERT`

[polyoxidonium]

I am migrating from h2 driver to mssql driver and have problem with returning the id after saving.
Problem happens also with writing foreign keys via custom write converters.
It first seemd like the id generation happens after r2dbc converters do their job but when i completely removed them and temporary adapted entities to not contain nested entities, I figured out that the generated ids are not returned at all.
How am I then supposed to gather generated id if not immediately after creation?
This all worked just fine with h2 with uuid() function in the flyway script
however with the MSSQL NEWID() function i keep getting null, however in the database values are persisted.

Flyway:

create table Foo
(
    foo_id                  uniqueidentifier default NEWID() PRIMARY KEY,
)

Repository: public interface FooRepository extends ReactiveCrudRepository<Foo, String>{}

Entity:

@NoArgsConstructor
@AllArgsConstructor
@Builder
@Getter
@Setter
@Table
public class Foo {

    @Id
    @Column(FooTable.ID) // foo_id
    private String id;

And so in the pipeline it looks like this:

    return fooRepository.save(foo)
        .doOnNext(f -> System.out.println(f.getId()); // prints null 

Not sure if it is a bug or I lack something.

[mp911de]

The SQL server driver returns only identifiers for IDENTITY columns. Default values are not supported.

[polyoxidonium]

The SQL server driver returns only identifiers for IDENTITY columns. Default values are not supported.

So in order to use UUIDs, one has to programatically generate them before persisting?

[mp911de]

Yes and beyond that, for every DEFAULT value/function that you want to consume, you need to fetch the persisted row. T-SQL supports an OUTPUT clause that can be used with INSERT but that has to happen from within the caller which is Spring Data R2DBC.

[polyoxidonium]

Yes and beyond that, for every DEFAULT value/function that you want to consume, you need to fetch the persisted row. T-SQL supports an OUTPUT clause that can be used with INSERT but that has to happen from within the caller which is Spring Data R2DBC.

I see, I guess to stay simple I will just avoid default values in the database itself,
thank you!

[mp911de]

Let's keep this issue open as other databases (e.g. Postgres) can automatically echo the entire inserted row without additional interaction. It would be neat if Spring Data could benefit from such behavior across multiple databases.