Thanks for all your suggestions guys,
I think I've figured out what is going on here, and it does seem to be
linked to the fact that Hibernate is using CGLIB enhancements for
Classes that are lazily loaded.
During the Validate phase of the Marshalling, the method
MappingLoader.getDescriptor(Class type) is called, this uses the
Hashtable _clsDescs to find the ClassDescriptor which are keyed by the
pojo's classes.
However, when the marshaller gets the class of a lazily loaded field
(with it's CGLIB enhancements), it doesn't match the classes within
MappingLoader._clsDescs, so the Marshaller treats the classes as not
mapped, and introspects them.
I don't know if this is the only place this is a problem, or whether I'm
going to encounter similar issues deeper within the code - can anyone
tell me?
I figure it must be possible to fix this problem, and as the alternative
is to drop Castor XML from the project I'll have a go.
Unless there is some other way of geting around it?
Any ideas?
Thanks
Andrew
Werner Guttmann wrote:
Stephen,
we'd be more than happy to provide you with some help when it comes down
to your personal learning experience. What do you need ?
To, for example, model a 1:M relation from A to B, in A you'll need a
property named 'bs' of type e.g. List, and in the mapping file a field
mapping similar to
<field name="bs" type="some.domain.B" collection="arraylist">
<sql many-key=" ..." />
</field>
And in terms of the SQL statements to create A and B, one could use the
following statements as a starting point:
create table domain_a (
id int not null PRIMARY KEY,
name varchar(20)
);
create table domain_B 8
id int not null PRIMARY KEY,
descr varchar(20),
a_id int
);
I hope this keeps you going ... ;-).
Werner
Stephen Bash wrote:
Well gentlemen, since I've wanted to learn more about the JDO world
for a while now, I took this as a chance to read up more on Castor
JDO. And honestly, most of it makes sense though I have very little
experience with O/R mapping. Unfortunately, when it comes to actually
creating the database schema or correctly modeling the 1:m relations,
I'm out of my league, so I'm probably going to have to leave this one
for someone else.
Thanks,
Stephen
On 2/13/06, Ralf Joachim <[EMAIL PROTECTED]> wrote:
Hi Andrew, Hi Stephen,
according to my knowledge hibernate returns cglib-proxies instead of the
original ones as they havily us lazy loading of data from database. In
the past a few users had problems with the constelation castor xml
together with hibernate according to this cglib-proxies.
Just in case you are not aware of it
Ralf
Stephen Bash schrieb:
Andrew-
So I tried to recreate your problem with a simpler set of classes:
public class ParentFoo {
protected int parentId;
protected Set children;
public int getParentId() { return parentId; }
public void setParentId( int parentId ) { this.parentId = parentId; }
public Set getChildren() { return children; }
public void setChildren( Set children ) { this.children = children; }
}
public class ChildFoo {
protected int childId;
protected Set parents;
public int getChildId() { return childId; }
public void setChildId( int childId ) { this.childId = childId; }
public Set getParents() { return parents; }
public void setParents( Set parents ) { this.parents = parents; }
public String toString() { return( String.valueOf( childId ) ); }
}
And the following mapping file:
<mapping>
<class name="ParentFoo">
<map-to xml="parent-foo" />
<field name="parentId" type="integer">
<bind-xml name="id" node="attribute" />
</field>
<field name="children" collection="set" type="ChildFoo">
<bind-xml name="child" />
</field>
</class>
<class name="ChildFoo">
<map-to xml="child-foo" />
<field name="childId" type="integer">
<bind-xml name="id" node="attribute" />
</field>
<!--
<field name="parents" collection="set"
type="castor.mason.one.ParentFoo">
<bind-xml name="parent" />
</field>
-->
</class>
</mapping>
I've attempted to recreate the 1:m m:1 relationship you had. You can
see in my mapping file I have the connection between the children and
the parents commented out. I then created a series of parents and
children, mixing and matching them. Using this mapping file as is to
marshal my first parent object I get:
<parent-foo id="1">
<child id="5"/>
<child id="4"/>
</parent-foo>
If I uncomment the child->parent relationship, I get:
<parent-foo id="1">
<child id="5">
<parent id="2">
<child id="6">
<parent id="3"/>
</child>
<child id="4"/>
</parent>
</child>
<child id="4">
<parent id="2">
<child id="6">
<parent id="3"/>
</child>
<child id="5"/>
</parent>
</child>
</parent-foo>
I'm somewhat surprised to find that I'm not getting the deep recursion
you mention. I tested this with Castor 0.9.9 and 1.0M2. What version
of Castor are you running? If you can run this example on your system
and agree or disagree with the output, that would also be helpful.
Do you see anything grossly different between your setup and mine? I
don't have the intermediate classes, so that might change things, but
I wanted to start simple.
Let me know your thoughts on this example and its comparison to your
code. I can't directly answer your question because from what I can
tell, you're doing what I'm doing...
Stephen
On 2/13/06, Andrew Mason <[EMAIL PROTECTED]> wrote:
OK,
Here is the castor mapping file:
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE mapping PUBLIC "-//EXOLAB/Castor Object Mapping DTD Version
1.0//EN"
"http://castor.org/mapping.dtd";>
<mapping>
<class name="uk.gov.forestry.harpps.model.Species" auto-complete="false">
<map-to xml="species"/>
<field name="scientificName">
<bind-xml name="scientific-name" node="attribute"/>
</field>
<field name="protections" collection="set"
type="uk.gov.forestry.harpps.model.Protection">
</field>
</class>
<class name="uk.gov.forestry.harpps.model.Protection"
auto-complete="false">
<map-to xml="protection"/>
<field name="lookupProtection"
type="uk.gov.forestry.harpps.model.LookupProtection">
<bind-xml name="lookup-protection" node="element"/>
</field>
</class>
<class name="uk.gov.forestry.harpps.model.LookupProtection"
auto-complete="false">
<map-to xml="protection-lookup"/>
<field name="protectionName">
<bind-xml name="protection-name" node="attribute"/>
</field>
<field name="lookupProtectionType"
type="uk.gov.forestry.harpps.model.LookupProtectionType">
<bind-xml name="protection-type" node="element"/>
</field>
</class>
<class name="uk.gov.forestry.harpps.model.LookupProtectionType"
auto-complete="false">
<map-to xml="protection-type"/>
<field name="protectionTypeId">
<bind-xml name="protection-type-id" node="attribute"/>
</field>
</class>
</mapping>
A Species has a Hashset of Protection objects. It also has many other
Objects associated with it in a similar way, but I'm not interested in
generating the XML for those (yet).
A Protection has a LookupProtection object, and a protectionName String.
A LookupProtection object has a LookupProtectionType object.
A LookupProtectionType has a protectionTypeId (long)
Below is the java code of the LookupProtectionType class, you will see
it has a relationship back to the LookupProtection object, but I don't
want to show this within the XML. Most of the Objects also have similar
2 way relationships.
public class LookupProtectionType implements java.io.Serializable {
// Fields
private long protectionTypeId;
private String protectionType;
private Set lookupProtections = new HashSet(0);
// Constructors
/** default constructor */
public LookupProtectionType() {
}
/** minimal constructor */
public LookupProtectionType(long protectionTypeId) {
this.protectionTypeId = protectionTypeId;
}
/** full constructor */
public LookupProtectionType(long protectionTypeId, String
protectionType, Set lookupProtections) {
this.protectionTypeId = protectionTypeId;
this.protectionType = protectionType;
this.lookupProtections = lookupProtections;
}
// Property accessors
public long getProtectionTypeId() {
return this.protectionTypeId;
}
public void setProtectionTypeId(long protectionTypeId) {
this.protectionTypeId = protectionTypeId;
}
public String getProtectionType() {
return this.protectionType;
}
public void setProtectionType(String protectionType) {
this.protectionType = protectionType;
}
public Set getLookupProtections() {
return this.lookupProtections;
}
public void setLookupProtections(Set lookupProtections) {
this.lookupProtections = lookupProtections;
}
}
Within the Hibernate mapping I made all the relationships Lazy, so they
should only load when their get method is called.
When I test the marshalling, I can see all the Non-mapped objects being
queried from the database.
I have tried mapping all the fields and setting them as Trasient, but
that seemed to have no effect.
Werner Guttmann wrote:
Andrew,
Can you please show us the relevant mapping file section you are using
to 'disable' some of the relations ?
Thanks
Werner
-----Original Message-----
From: Andrew Mason [mailto:[EMAIL PROTECTED]
Sent: Montag, 13. Februar 2006 11:56
To: user@castor.codehaus.org
Subject: [castor-user] Castor marshalling Hibernate Pojos
Hello,
I apologise if this topic has been covered before - if so a
pointer to the relevent posts will be appreciated.
I am using Hibernate as my object persistence layer (sorry
but I can't use Castor JDO due to political reasons).
I wish to use Castor XML to marshall the pojos into xml documents.
Many of the objects have relationships with objects that are
required from a database perspective, but are not required
within the XML.
For example
Species has a one-to-many relationship with Predator.
Predator has a one-to-many relationship with Species.
When I marshal a Species object I want to include it's
predators, however, I do not want to include all the Species
the predator preys upon within the Species XML document.
I assumed that by writing a mapping document for Predator,
that did not include rerlationship with Species, Castor would
not bother calling the getPreySpecies method.
What I am finding is that despite the mapping not including
this relationship, the method is still being called, causing
more species to be instantiated, thus causing more predators
to be instantiated and so on until my machine runs out of memory!
Is there any way of ensuring that the Marshaller pays strict
attention to the Map, and doesn't call methods that are not defined?
Thanks
--
*************************************************
Andrew Mason
Environ-IT Ltd.
Tel.: 01899 880200
Mob.: 07919 967494
E-mail: [EMAIL PROTECTED]
*************************************************
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
--
*************************************************
Andrew Mason
Environ-IT Ltd.
Tel.: 01899 880200
Mob.: 07919 967494
E-mail: [EMAIL PROTECTED]
*************************************************
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
--
*************************************************
Andrew Mason
Environ-IT Ltd.
Tel.: 01899 880200
Mob.: 07919 967494
E-mail: [EMAIL PROTECTED]
*************************************************
-------------------------------------------------
If you wish to unsubscribe from this list, please
send an empty message to the following address:
[EMAIL PROTECTED]
-------------------------------------------------
