Hibernate version:2.1.4

Name and version of the database you are using:Oracle 8.1.7

Hi,

I have been reading a lot of documentation and faq's about the "inverse=true" settings in a bi-directional relationship. I have actually tried this out, and it works. But as we all know, it is not very convenient to add a child record, when you have a parent child relationship with 1000's of children. Problem is you have to load all children first (if you don't want to use a bag), which is of course unacceptable.

Now, in my situation, I have a parent and a child table. The child table has a foreign key (not-null) to the parent table. I am using a bi-directional relationship, but with inverse=false.
When I want to create a new child record, I simply do the following:

- load the parent
- create a new child, using the parent's id field to set the foreign key constraint (Child.setParentId(Parent.getId())
- save the child

When I look at the SQL Hibernate produces at child creation time, I see one simple insert statement, which succeeds. Now, I don't see how the "inverse=true" setting could optimize this? I am probably missing something, but for me everything works just fine with the default (inverse=false) setting??

Regards,
Johan

(1) you don't need a parentId field in HIbernate, just the parent reference will do fine.

(2) if all you do is call child.setParent(parent), you will certainly only see one harmless insert statement, regardless of the setting for inverse

(3) it is when you *also* call parent.getChildren().add(child) that you would see the problem (you need to do this to maintain integrity at the Java level, if you plan to keep working with the parent object, or especially if the collection is cached)

[quote="gavin
(3) it is when you *also* call parent.getChildren().add(child) that you would see the problem (you need to do this to maintain integrity at the Java level, if you plan to keep working with the parent object, or especially if the collection is cached)[/quote]

First of all, thank you for the quick answer!
Now, I can follow your way of thinking, but for the 3rd remark, I just don't see why I would want to call parent.getChildren().add(child), as this would load all children first (thousands in my case), and as a child can very easily be added by just mapping the parentId in the Child class, instead of referencing the parent class??

It wouldn't load the children if they are mapped as a bag (duplicates allowed). Think about it: A Set will always have to load them if you add an element, to guarantee uniqueness of the elements.

He is asking why does he need to add it at all. Which is a good question. Strictly, you don't, unless the collection is cached in the second-level cache.

When reading the forums, I still see a lot of questions regarding the proper use of "inverse=true" on bi-directional relationships. Also, in all (?) discussions and examples on this topic, I always see the parent.getChildren().add(child) appearing. Considering your answers, maybe it would be a good idea to add a very basic insert example like child.setParent(parent) on the FAQ pages, as this probably is sufficient in a lot of cases, and "solves" the annoying limitation of having to load all children first, before creating a child.