1.2 sql server 2000.

Hi,

I was trying to delete a children element of a parent (one-to-many), and I got "deleted object would be re-saved by cascade", then i noticed in my parent mapping, I had a bag of children element with cascade="all", so I changed to "none" and the error is gone.

but this caused another problem. When I try to save the parent (and attached the children along with it), i get

"object references an unsaved transient instance - save the transient instance before flushing"

hm.. okay. So I changed the casecade="save-update" (and also tried "all-delete-orphan", and now saving parent(with children attached to it), agian it's fine, but back to delete children again, get the original error "deleted object would be re-saved by cascade".

I went back to documentation and now I'm a bit confused.. how do I delete the chidlren? and when should I use cascade="all" v.s "save", "delete" and etc?

From page 79 of the pdf reference guide

"Mapping an association (many-to-one, or collection) with cascade="all" marks the association as a parent/
child style relationship where save/update/deletion of the parent results in save/update/deletion of the
child(ren). Futhermore, a mere reference to a child from a persistent parent will result in save / update of the
child. The metaphor is incomplete, however. A child which becomes unreferenced by its parent is not automatically
deleted, except in the case of a <one-to-many> association mapped with cascade="all-delete-orphan"

here is my parent mapping file




here is my child mapping



and here is the code I used to delete the children




Thanks

okay.. looking at section "16.3. Cascading lifecycle"..

i did two thigns

1. change cascade to "all-delete-orphan" in the parent mapping



2. I first retrieved the parent object (JumptreeForum_Discussion) and then loop through all its children, find the child element, remove from the parent and then delete the child...



this worked.. but I'm just not sure if i'm doing it right.. to delete this child, from log, i see 4 sql exectued for this one operation

1. one sql select to get parent..
2. one sql select to get the children elements of the parent
3. one sql update to set the children element's parentid to null
4. one sql delete to delete the children element

is that right???? 4 sql to delete one record?

I've got the same problem here, I don't understand why I get an update query before the deletion, which at the same time, tries to "null" the parentid and the index column of my MAP collection.

That's expensive, and I get a "constrain error" because obviously, I don't want my index column to be "null".

1) If you want to retrieve the parent and child objects in a single statement, you have to add fetch="join" or fetch="subselect" to your bag mapping. But, then you can't use lazy fetching.

2) If you want to avoid the update, you have to add a parent mapping to your child mapping:

<many-to-one name="Discussion" column="discussionid" / >

and add inverse="true" to the bag in the parent class. The

See chapter 17 in the reference manual:



Hope that helps,

Wolfgang

Thank you, that kind of helps, but:

1- I am using a "map" which for makes more sense, since I want to be able to access specific objects easily, and maps can't have bi-directional relationships

2- I just found this in the Hibernate (not NHibernate) documentation:



It seems like the not-null attribute is what i need, but it also seems like it doesn't exist in NHibernate??

And also, I would like to be able to have the "index" key as not null.
I have already mapped them as "not-null" in the class mapping of the contained elements, but NHibernate doesn't seem to take it into consideration.

Afaik, the parent child mapping shoud work with maps exactly the way it does for bags or lists. Have you tried it ?



If you remove the child from the collection, hibernate has to set the column (key and/or index) to null otherwise the item would still be part of the collection.

If it does not work, can you post your mappings and the code you use for deleting the child.

1 - I haven't tried, you are right. I'll give it a try if I can't find any other way to solve the issue


2 - You are also right about setting the key and/or index to null, but I don't understand why it has to set them to null before deleting them. Wouldn't it make sense to just delete them without the "UPDATE" query first?

Have a look in chapter 17 of the reference, it's explained there quite well. Removing an object from a collection does not delete the object automatically. You have to delete the child explicitly or use cascade="all-delete-orphan" on the collection mapping. But I think, even then, you have to set the parent link in the child object to null (if you have one) before you commit your changes.

I have a "all-delete-orphans" cascade value.

Let me give you some idea about what code I have:


the container class contains a dictionary of contained object.

i just do this:

transaction = session.begintransaction()

container = session.get<containerclass>(id)

container.clear() //clears the dictionary

container.add(object) in a "loop" to repopulate the container with new objects.

transaction.commit()

and here, I get plenty of

updating with null values
inserting the new values
deleting the rows which had been updaetd just before.

I would like to know if it is possible to just automatically insert the new rows, delete the privious ones (knowing that they have unique ids).

yes, if you mark the collection mapping as inverse="true" and add a many-to-one mapping to your childs.

Ok, i'll do it right now and see what I get.

Thanks

It really seems like you can have the inverse relationship with the MAP

Wierd because I had read the opposite...

So thanks a lot, having the reverse relationship did fix my problem of UPDATE before DELETE, though it makes me think that NHibernate is not as transparent as I believed, since I had to change my classes design.