public enum SharedJoinTableUniqueness
extends java.lang.Enum

How a join table that several collections of one owner share keeps the rows of each collection unique, by database.

Such a table has a key column per collection and one element column, or one key column and an element column per collection, and every row leaves the columns of the other collections null. A row is unique by the key column of its own collection and the element, so each collection needs a constraint of its own over those two; one constraint over all key columns would hold a null in every row and reject nothing on a database that treats nulls as distinct. What that per-collection constraint has to look like depends on how the unique index of the database treats a null in its key (UniqueIndexNulls):

• KEY_PER_COLLECTION, for UniqueIndexNulls.DISTINCT: a unique key over the key column and the element. The rows of the other collections, null in this key column, never collide.
• ONE_KEY_FOR_ALL_COLLECTIONS, for UniqueIndexNulls.COMPARED: Oracle compares the non-null columns of a partially null row, so a unique key per collection would reject the rows of the other collections, null in that collection's columns and equal in the ones they have in common. One unique key over the key columns of every collection and the element columns is right there: a row collides only with a row that is non-null in the same columns, that is a row of its own collection, for the same element. That holds for collections that share the element column and have a key column each, and for collections that share the key column and have an element column each. Hibernate renders that key as a plain constraint on Oracle, where it is what is wanted; on a database that treats nulls as distinct the same key would hold a null in every row and reject nothing, which is why it is for COMPARED alone.

Since:

8.0.1