[hibernate-issues] [Hibernate-JIRA] Commented: (HBX-793) Inherited meta attributes are totally broken
Rob Jellinghaus (JIRA)
noreply at atlassian.com
Mon Oct 30 00:57:04 EST 2006
[ http://opensource.atlassian.com/projects/hibernate/browse/HBX-793?page=comments#action_25014 ]
Rob Jellinghaus commented on HBX-793:
-------------------------------------
OK, it seems that the behavior of "inherit" changed dramatically between hibernate2 and hibernate3.
I uploaded an example of the same mapping, with build targets for both the hibernate2 and hibernate3 versions of hbm2java. The same mapping file works great under hibernate2 but produces corrupt output (as described above) under hibernate3.
Adding inherit="false" to the parent meta attributes fixes the problems with Hibernate3, for both generated-class and scope-get/set.
I'm pretty sure that the root problem is that when you redeclare an inherited meta atttribute in Hibernate2, the redeclaration overrides the parent meta declaration. Which I would strongly argue is the correct behavior. But in Hibernate3, the redeclared meta attribute gets *concatenated* to the parent (inherited) attribute. Is there ever any situation where that is the right behavior??? I can't imagine any right now...?
Also, the current Hibernate3 documentation on meta inheritance -- http://www.hibernate.org/hib_docs/tools/reference/en/html_single/#d0e1630 -- doesn't describe what the behavior of overriding an inherited meta attribute should be.
If I can make the time, I will dig into the source code of each generator to see exactly what Hibernate2 did differently, but I've been converting all our production code to Hibernate3 and have gotten hung up on this for too long already, so I'll have to come back to it later (I'll have to just fix the mappings for now).
The frustrating part is that there really isn't a specification, and without a clear spec, bugginess is in the eye of the beholder. If people think that the Hibernate3 behavior is a bug (and I would like to hear a good argument as to why it isn't), then please vote for this issue.
> Inherited meta attributes are totally broken
> --------------------------------------------
>
> Key: HBX-793
> URL: http://opensource.atlassian.com/projects/hibernate/browse/HBX-793
> Project: Hibernate Tools
> Type: Bug
> Components: hbm2java
> Versions: 3.2beta8
> Environment: Hibernate 3.2.0cr8, Hibernate Tools 3.2.0cr8
> Reporter: Rob Jellinghaus
> Attachments: hbm2java_meta_testcase.zip
>
> Original Estimate: 2 days
> Remaining: 2 days
>
> Inherited / overridden meta attributes are totally broken.
> By "totally broken" I mean that they are essentially concatenated in the output template, rather than being overridden properly.
> There is already an existing bug http://opensource.atlassian.com/projects/hibernate/browse/HBX-621 that is one manifestation of this underlying problem for <meta attribute="scope-get"> and <meta attribute="scope-get">. The exact same bug reappears for <meta attribute="generated-class">.
> Here is an example that shows the bug appearing in both attributes. The following mapping (simplified):
> <?xml version="1.0" encoding="UTF-8"?>
> <!DOCTYPE hibernate-mapping PUBLIC "-//Hibernate/Hibernate Mapping DTD//EN" "http://hibernate.sourceforge.net/hibernate-mapping-3.0.dtd">
> <hibernate-mapping default-lazy="false">
> <class name="com.nimblefish.core.domain.useraccesscontrol.Permission" table="permission">
> <meta attribute="generated-class">com.nimblefish.core.domain.generated.PermissionGen</meta>
> <id column="id" name="id" type="long" unsaved-value="null">
> <meta attribute="use-in-tostring">true</meta>
> <generator class="native"/>
> </id>
> <!-- other properties deleted -->
> <joined-subclass name="com.nimblefish.core.domain.useraccesscontrol.Role"table="role">
> <meta attribute="generated-class">com.nimblefish.core.domain.generated.RoleGen</meta>
> <meta attribute="scope-set">protected</meta>
> <key column="id" foreign-key="fk_role_sub_permission"/>
> <set name="grantedPermissions" inverse="true" table="role_to_granted_permission">
> <meta attribute="scope-set">public</meta>
> <key column="role_id" foreign-key="fk_granted_permission_role"/>
> <many-to-many class="com.nimblefish.core.domain.useraccesscontrol.Permission" column="permission_id" foreign-key="fk_role_granted_permission"/>
> </set>
> <!-- other properties deleted -->
> </joined-subclass>
> </class>
> </hibernate-mapping>
> Produces the following class, in the directory <out-dir>/PermissionGencom/nimblefish/core/domain/generated/RoleGen.java (that is NOT a typo):
> package com.nimblefish.core.domain.generated.PermissionGencom.nimblefish.core.domain.generated;
> // Generated Oct 27, 2006 9:58:57 PM by Hibernate Tools 3.2.0.beta8
> import com.nimblefish.core.domain.useraccesscontrol.TargetObjectContainer;
> import java.util.HashSet;
> import java.util.Set;
> /**
> * RoleGen generated by hbm2java
> */
> public class RoleGen extends com.nimblefish.core.domain.useraccesscontrol.Permission implements java.io.Serializable {
> // Fields
> // fields deleted
> protectedpublic void setGrantedPermissions(Set grantedPermissions) {
> this.grantedPermissions = grantedPermissions;
> }
> }
> You can see that:
> 1) The filename contains the CONCATENATION of the meta attributes and superclass name.
> 2) The package declaration contains the exact same conctatenation.
> 3) The scope of the setGrantedPermissions method contains the CONCATENATION of the scope-set values.
> I strongly believe that the root bug is related to the following method in org.hibernate.tool.hbm2x.MetaAttributeHelper.java:
> public static String getMetaAsString(org.hibernate.mapping.MetaAttribute c) {
> return c==null?"":getMetaAsString(c.getValues() );
> }
> I don't know exactly what the collection of values of a MetaAttribute is, but I am fairly convinced that it is the sequence of values defined up the inheritance chain. And I am *quite* sure that getMetaAsString(Collection meta, String separator) concatenates all the meta values together, which is exactly the behavior that is causing all these bugs.
> Hibernate Tools can't go to 3.2.0GA until these issues are fixed. I will be getting the code from Subversion and patching it myself (hackishly) on Sunday; I'll upload details of my fix then, along with a better test case in a properly runnable self-contained zip.
--
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators:
http://opensource.atlassian.com/projects/hibernate/secure/Administrators.jspa
-
For more information on JIRA, see:
http://www.atlassian.com/software/jira
More information about the hibernate-issues
mailing list