Hi,

Our datamodel requires us to write very strange hibernate mappings. Everything works so far. However, I'm curious about what the expected behaviour would be for the following mapping:



A Cell knows of its Col and Row -- easy. It manages the ternary Col-Cell-Row relation-ship.

A Col has a bag of Rows, which is defined via the Col-Cell-Row relation-ship in the CELL table. I don't want hibernate to execute any insert/update/delete for this bag, so it is mapped with inverse="true" (read-only).

Now the tricky part:

A Row has a map of Col to Cell via the same Col-Cell-Row relation-ship in the CELL table, which is why this map is also mapped with inverse="true" (read-only). This map is the only collection that can hold Cells, i.e. the life of a Cell is bound to this collection, which is why it is mapped with cascade="all-delete-orphan". In our application, it is OK to simply delete Rows (Hibernate is supposed to cascade-delete all referenced Cells).

Now, the question is, what will Hibernate do on a delete-orphan of a map-entry:

• Will it only delete the entry-value (i.e. the Cell) or
• will it also delete the entry-key (i.e. the Col).

Deleting the Col is an error, which will be punished with a ReferentialIntegrityException -- there still might be other Cells of different Rows that reference a Col.

In general: What is the expected behaviour of delete-orphan on a <map-key-many-to-many /> ?

PS: This is only my curiosity and not a real problem. I could easly refactor the map-of-Col-to-Cell to a bag-of-Cell and extract the Col from each Cell, but it's nice to have a map that is managed by Hibernate.