HHH-15722 : @OneToMany mappedBy with a @Any

[boutss]

https://hibernate.atlassian.net/browse/HHH-15722
https://discourse.hibernate.org/t/onetomany-with-interface-type-return-variable-role-column-any/7050

Since the collection, which is mapped by the return variable (mappedBy) of the referenced objects, knows its type, the identifying column of the foreign key is sufficient, without having to use the discrimination column.

@startuml

class Library {
    -Long id
    -Set<Book> books
    +void setId(Long)
    +Long getId()
    +Set<Book> getBooks()
    +void setBooks(Set<Book>)
}

class Book {
    -Long id
    -Store store
    +void setId(Long)
    +Long getId()
    +Store getStore()
    +void setStore(Store)
}

interface Store {
}

Library --|> Store : implements
Library "1" -- "*" Book : contains

@enduml

[beikov]

the identifying column of the foreign key is sufficient, without having to use the discrimination column.

I'm not so sure about this. In simple cases like your test case this might work out, but there is no guarantee that an FK with value X always points to a certain subtype.

Imagine the following test data:

Book(id: 1, store_id: 1, store_role: 'library'), 
Book(id: 2, store_id: 1, store_role: 'shop'),
Library(id: 1),
Shop(id: 1) 

Now, the way the mapping is constructed, the collection in a library would contain both books, which is wrong.

You have to add an additional restriction to the collection e.g. store_role = 'library'. I tried annotating @SQLRestriction("store_role = 'library'") on the collection mapping of your test entity, and it produced the desired effect. Now you just need to ensure that you infer the correct restriction in the boot model of the collection. Also, please add a test based on the test data that I posted to ensure the correct behavior.

[boutss]

Thank you very much, I'll do this as quickly as possible!