Hello!,

I'm trying to construct a mapping to legacy hierarchical data. The data uses a parent_id link to navigate from any node to the root in the tree
(navigating the tree from top-bottom is not required only from bottom to top).
Below is my current mapping, using hbm2ddl I'm able to generate an identical table structure as in our legacy, so I'm almost convinced this mapping is correct.

The problem is at runtime, when creating the session Factory (probably when it reads the hibernate.cfg.xml) that I get a
"Repeated column in mapping for collection" error, I've read some postings suggesting to upgrade to hb3.

Is that the only choice? or is there something that can be changed in this mapping or even the legacy table structure to make it work.
(our preference is not to change the legacy table schema)

Here is some sample data:

PKID NODENAME PARENT_ID
1 A 0 // this is the root
2 A1 1 // second level in tree, parent is #1, first child in this level
3 A11 2 // third level in tree, parent is #2
4 A2 1 // second level in tree, parent is #1, second child in this level

Appreciate any help you can provide in this issue.



Hibernate version: 2.1.8

Mapping documents:

<hibernate-mapping package="persistencetier.tree">
<class name="Tree" table="TREE">
<id name="id" type="integer" column="PKID">
<generator class="increment"/>
</id>
<property name="nodeName" column="NODENAME" not-null="false"
type="string" length="200"/>

<list name="parentList" table="TREE" lazy ="true" >
<key column="PKID" />
<index column="PKID" type="integer" />
<element type="integer" column="PARENT_ID" not-null="true"/>
</list>
</class>
</hibernate-mapping>

Code between sessionFactory.openSession() and session.close():

session = sessionFactory.openSession();
Query query = session
.createQuery("from Tree tree where tree.id = 3");
List result = query.list();

session.close();


Full stack trace of any exception that occurs:

14:49:41,527 INFO SettingsFactory:114 - Use scrollable result sets: true
14:49:41,537 INFO SettingsFactory:117 - Use JDBC3 getGeneratedKeys(): false
14:49:41,537 INFO SettingsFactory:120 - Optimize cache for minimal puts: false
14:49:41,547 INFO SettingsFactory:126 - echoing all SQL to stdout
14:49:41,547 INFO SettingsFactory:129 - Query language substitutions: {}
14:49:41,557 INFO SettingsFactory:140 - cache provider: net.sf.hibernate.cache.EhCacheProvider
14:49:41,627 DEBUG SettingsFactory:173 - Wrap result sets enabled? : false
14:49:41,637 INFO Configuration:1130 - instantiating and configuring caches
14:49:41,767 WARN Configurator:125 - No configuration found. Configuring ehcache from ehcache-failsafe.xml found in the classpath: jar:file:/C:/jakarta-tomcat-4.1.31/webapps/myapp/WEB-INF/lib/ehcache-0.9.jar!/ehcache-failsafe.xml
14:49:42,158 INFO SessionFactoryImpl:119 - building session factory
14:49:42,168 DEBUG SessionFactoryImpl:125 - instantiating session factory with properties:
net.sf.hibernate.MappingException: Repeated column in mapping for collection: myapp.Tree.parentList column: PKID
at net.sf.hibernate.collection.AbstractCollectionPersister.checkColumnDuplication(AbstractCollectionPersister.java:654)
at net.sf.hibernate.collection.AbstractCollectionPersister.<init>(AbstractCollectionPersister.java:210)
at net.sf.hibernate.collection.BasicCollectionPersister.<init>(BasicCollectionPersister.java:35)
at net.sf.hibernate.persister.PersisterFactory.createCollectionPersister(PersisterFactory.java:55)
at net.sf.hibernate.impl.SessionFactoryImpl.<init>(SessionFactoryImpl.java:148)
at net.sf.hibernate.cfg.Configuration.buildSessionFactory(Configuration.java:805)
at myapp.Test.myTest(Test.java:120)


Name and version of the database you are using: Oracle 9

The generated SQL (show_sql=true):

no sql generated, fails when reading hibernate.cfg.xml

Debug level Hibernate log excerpt:

all set to DEBUG

You're using <index> but your table does not have an index column (see http://www.hibernate.org/hib_docs/v3/reference/en/html/collections.html#collections-indexed).

Whoops, the top half of my post got lost...

You should rather use a <set> instead of a <list>.

Aside from that, why not directly map real associations instead of integers?

Hi Mickey,

Thanks for your response.
The reason to use list instead of set is because I need to be able to
store duplicate elements.

PKID NODENAME PARENT_ID
1 A 0 // this is the root
2 A1 1 // 2nd level in tree, parent is #1, first child
3 A11 2 // 3rd level in tree, parent is #2
4 A2 1 // 2nd level in tree, parent is #1, 2nd child
5 A12 2

In this case the tree is something like this:

A
/ \
A1 A2
/ \
A11 A12


Please notice I'm currently using Hibernate 2.1.8 and the schema (table structure) is a legacy without too many options to be changed.

What do you mean by "map real associations instead of integers?"
It may be that I have to use that because is a legacy table struct.

Thanks for your help !

Yeyo

Are you saying a parent can have a child more than one time? As in:



Like that?




Yes, but that doesn't prevent you from having real Parent objects in the parentList collection instead of their FK's.

Hi Mickey,

There are no repeated child nodes but repeated instances of parents when navigating from bottom to up.
For instance for the following tree if we start from A11 I should be able to retrieve A11,A1,A or if I'm in node A12 I should get A12,A1,A
I've tried a Set but then it creates the constraint where PARENT_ID must be unique (which can't in this tree)
This sample contains only nodes with 2 childs but the actual data contains nodes with multiple childs.

Thanks !

Just to add more info to the previous posting.

Would it be possible to load this type of structure in an efficient manner, without really knowing the number of parents (assuming your are navigating the tree from the leaves to the branch) ?

look at subselect and batchfetcing

Thanks for your info.

Now, how does hibernate know when it's the last node in the recursive list ? By finding a null reference? further more in my case due to the legacy design/data would it be possible to specify different values in this column to indicate the end of the list?

Many thanks

i dont get why hibernate should care about when you reach the "end"...the end is when there isn't more references to follow.

regarding have special marker values then look at usertypes.

Maybe I'm not reading this right, but if you are always going to be navigating from child to parent, just have the child reference the parent:



Is that what you need?