Description Fabian 2010-05-27 05:24:07 EDT

Hello,

we faced the following problem using Postgresql 8.4:

Having a class BankAccount with two sets (A and B) both referencing class BookingLine, the generated SQL will be as follows:

- Table BankAccount
  * id
  * ...


- Table BookingLine
  * id
  * ...

- N:M Table BankAccount_BookingLine
  * bankAccount_id NOT NULL
  * setA_bookingLine_id NOT NULL
  * setB_bookingLine_id NOT NULL


See the problem? If I add a new BookingLine to a BankAccount in one set (let's assume set A), the generated SQL will look as follows:

INSERT INTO BankAccount_BookingLine VALUES (1, 23, null)

which leads to a constraint violation because of the NOT NULL constraints.


A workaround: Use @JoinTable to specify a different join table for each set.
I don't know what is the best behaviour in this case but I believe that EclipseLink should not generate a schema that cannot be used properly with the corresponding data model.

Hope it helps,

Fabian