Does anyone know why when I do an update on a cross reference table defined by @ManyToMany, hibernate has to delete all related records in the cross reference table and re-insert them again.. Even I don't change anything, it will still fetch all data from the table, delete them and re-insert them again. The ids of those records always change.

How can I just change one of them instead of affecting all of them? Hibernate should know the id of that record already exists and it should only perform an update, but not an insert after a delete.

Thank you.

If you do not change anything in the relation Hibernate doesn't do anything. A common mistake is to not return exactly the same object from the getter method that Hibernate used when calling the setter method. Eg. user something like that below. No code should never be allowed to change 'foo' to point to some other Set object.



The reason is that Hibernate uses a special Set implementation that keeps track of what you add and remove from the Set so that Hibernate can only submit changes to the database. If the 'foo' is changed to point to some other object Hibernate looses track of changes and as a safety measure deletes everything and re-inserts new rows.

Thank you for the explanation. I think I got your point.

The problem is that I use lazy loading and collection will only be initialized when it displays in view. So when user clicks on the update button without changing anything (I think the collection is already there), the id of each element in the collection is passed to the back end. I need to add them individually to the collection by using another find (database call) to avoid checking if the element is newly added..

Because of the ManyToMany, I have to pass an object to the collection instead of an id.

So I guess you are actually doing changes to the many-to-many relation. It will be kind of hard to help you unless you post some code, mappings and other relevant information.

I am integrating Spring with Hibernate and I take advantage of initBinder method to register a custom editor to convert a collection of ids from view to a collection of objects. Because of ManyToMany, I have to save a collection of objects instead of collection of Long ids. I checked the api for registerCustomEditor and it says that "the PropertyEditor has to create an array or Collection value". I guess that's why I always have a new collection which tells hibernate to delete all related record first and re-insert them even there is nothing changed.

Does that mean that Spring is calling setFoo() with a different collection instead of getFoo().add()/getFoo().remove()? If that is the case then the behavior you see with Hibernate is quite expected.

I have no experience with Spring, but I guess this is more a Spring-related problem than a Hibernate-related problem.

I have tried to modify the collection in one session as following..



But the id of a record the cross reference table was still changed. Could it be related to the oracle database trigger I defined?

Does the above code really works?

According to the API documenation for the Session.update() method an exception should be thrown if it is called with an argument that is already attached to the session. In any case, there is no need to call Session.update().

I am not sure that the call to tx.commit() should be placed inside the finally clause. If there is an error you will end up calling both rollback() and then commit().

Regarding the trigger, I have no idea.

Thank you for the suggestion, but even I got rid of those, hibernate still tried to delete first and then did a insert instead of just delete the proper record only...

Hibernate: delete from A_B where A_ID=?
Hibernate: insert into A_B (A_ID, B_ID) values (?, ?)