Cascade delete not firing?

subdigital · **Posted:** Mon Sep 25, 2006 4:02 pm

I'm using NHibernate 1.0.2

(fyi: I'm sure this is a bug in my mapping, but I want to confirm here)

Say you have a hierarchy (like in the book HIA) where you have:

Categories
Items
Bids

A category has many Items
An Item has many bids

---------------------------------

now if you delete a category it should delete every item and bid underneath it... so in the mapping you'd specify something like this:

Code:

//Category.hbm.xml
<bag name="Items" ......  cascade="all-delete-orphan"> ... </bag>

//Item.hbm.xml
<bag name="Bids" ........ cascade="all-delete-orphan">....</bag>

Is this correct?

I currently have a relationship just like this, but when I call delete on the root object, it sets the FK to NULL on the bottom object (bids here).

so the action goes like this:

Delete issued in code
SQL -> UPDATE Bids SET ItemId = NULL
SQL -> DELETE Item WHERE ...... //ERROR! FK VIOLATION

Are my assumptions correct?

devonl · **Posted:** Mon Sep 25, 2006 5:22 pm

this can sometimes happen if you don't have inverse="true" set on the bag (assuming you have a relationship back from Item to Category and from Bid back to Item...

-devon

subdigital · **Posted:** Mon Sep 25, 2006 5:37 pm

I thought that too, but the associations are one-way...

devonl · **Posted:** Mon Sep 25, 2006 6:19 pm

The Item class is _not_ composed of a Category class (or a CategorizedItem class to quote HIA)? The Property might not be public, but you must have a member variable of the Item class that points in some way back to the Caetgory. Can you post the Item class?

Cheers,
-d

subdigital · **Posted:** Mon Sep 25, 2006 7:01 pm

In my object-model it is a string hierarchy like that.

The entities are:

Users -> Roles -> Constraints

if I remove a users roles, the constraints should be deleted as well, however they are not... my FK in the constraints table is being set to null by NHibernate.

This smells like a directionality issue / something that might be solved by inverse, however the Constraint has no reference back to Roles (is this my error?)

devonl · **Posted:** Mon Sep 25, 2006 10:15 pm

not entirely understanding your model, however, if a Constraint has a life cycle outside of a user's Role(s) then no you do not need an inverse relationship (however, I would make the case this is a M-M relationship). However, if a Constraint does not have a life cycle outside of a user's Role(s) then, yes, in my opinion, you need the inverse relationship from Constraint back to Role. This relationship does _not_ have to be navigable, however it must be there at least as a member variable. Something like:

Code:

class Role {
   private User user;
   private IList constraints;

   public Role (User u) {
      this.user = u;
      this.roles = new ArrayList();
   }

   public virtual User User {
      get { return this.user; }
   }

   public virtual IList Constraints {
      get { return this.constraints; }
   }
}

class Constraint {
   private Role role;
   ...

   public Constraint (Role r) {
      this.role = r;
   }
}

Note in the Constraint class there is no public accessor for the role member variable. The relationship exists but you can't navigate it. Apologize if you know this. Like I said, I may be misunderstanding your model.

-d

subdigital · **Posted:** Mon Sep 25, 2006 10:18 pm

I wanted to keep it simple so it wouldn't be polluted with extraneous details. What you mention looks like it will solve my problem completely.

I'll try it in the morning and let you know if it solves my issue. Thanks!

subdigital · **Posted:** Tue Sep 26, 2006 1:01 pm

Thanks for your suggestion. It had quite a few mapping ramifications, but once they all followed the same pattern all my tests are passing again.

Thanks!

devonl · **Posted:** Tue Sep 26, 2006 9:02 pm

great! glad it worked out.

cheers,

-d