[jira] Commented: (OPENJPA-291) Entity Relation as identity for self-referential relation breaks enhancer

["Krzysztof (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/OPENJPA-291?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12740601#action_12740601 ] 

Krzysztof commented on OPENJPA-291:
-----------------------------------

Hello,
This is quite common scenario in our project. Any chance to have it fixed in a foreseeable future? Quite surprised it has low priority..
Best regards
Chris

> Entity Relation as identity for self-referential relation breaks enhancer
> -------------------------------------------------------------------------
>
>                 Key: OPENJPA-291
>                 URL: https://issues.apache.org/jira/browse/OPENJPA-291
>             Project: OpenJPA
>          Issue Type: Bug
>            Reporter: Pinaki Poddar
>            Priority: Minor
>         Attachments: Node.jad, Node.java, TestTree.java
>
>
> Use Case:
> A Node of a tree has a reference to parent node and a collection of child nodes.
> The complexity is how the node is identified. Instead of a simple primary identity,
> the node is using an identity that is composed of a primitive and its parent.
> OpenJPA allows parent (a uni-cardinarity relation) to be used as part of compound primary key. 
> But the entire process of enhancement and metadata resolution do not fully cover when the
> relation is self-recursive. 
> The use case of defining such a key however does not seem to be common.
> Anyway, consider Node of a tree defined for persistence as follows:
> @Entity
> @IdClass(Node.SelfId.class)
> public class Node {
>     @Id 
>     private long id;
> 	
>     @Id 
>     @ManyToOne private Node parent;
> 	
>     @OneToMany(mappedBy="parent", cascade=CascadeType.ALL)
>     private List<Node> children;
>     public class SelfId {
> 	public long id;
> 	public SelfId parent;
> }
> 1. Enhancer breaks on Node.class
>    Because metadata resolution goes to infinite recursive loop and a StackOverflow
>    The codelines that breaks are in MetaDataRepository and the 
>    tentative patch shows why infinite recursion is happening by just stopping it. 
>   The impact of cutting it short is not clear/investigated.
>     
> Index: openjpa-kernel/src/main/java/org/apache/openjpa/meta/MetaDataRepository.java
> ===================================================================
> --- openjpa-kernel/src/main/java/org/apache/openjpa/meta/MetaDataRepository.java        (revision 558795)
> +++ openjpa-kernel/src/main/java/org/apache/openjpa/meta/MetaDataRepository.java        (working copy)
> @@ -552,11 +552,15 @@
>          // resolve relation primary key fields for mapping dependencies
>          FieldMetaData[] fmds = meta.getDeclaredFields();
> -        for (int i = 0; i < fmds.length; i++)
> -            if (fmds[i].isPrimaryKey())
> +        for (int i = 0; i < fmds.length; i++) {
> +            if (fmds[i].isPrimaryKey()) {
> +               if (fmds[i].getObjectIdFieldType().
> +                       isAssignableFrom(meta.getDescribedType()))
> +                       continue;
>                  getMetaData(fmds[i].getDeclaredType(),
>                      meta.getEnvClassLoader(), false);
> -
> +            }
> +        }    
> 2. This patch stops the recursion but does not solve the problem.
>    The enhanced Node class fails to create a ApplicationId. For a simple test case such as (the test case is attached)
> 	public void testPersistTree() {
> 		OpenJPAEntityManager em = getEM();
> 		em.begin();
> 		Node root = new Node(1000L,null);
> 		createTree(root, new int[]{1,2,3}, 2);// 3-levels, fanout/node=2 
> 		em.persist(root);
> 		em.commit();
> 	}
>  
> <1.0.0-SNAPSHOT-SNAPSHOT nonfatal general error> org.apache.openjpa.persistence.PersistenceException: org.apache.openjpa.util.ObjectId
> 	at org.apache.openjpa.kernel.BrokerImpl.persist(BrokerImpl.java:2437)
> 	at org.apache.openjpa.kernel.BrokerImpl.persist(BrokerImpl.java:2333)
> 	at org.apache.openjpa.kernel.BrokerImpl.persistAll(BrokerImpl.java:2279)
> 	at org.apache.openjpa.kernel.SingleFieldManager.persist(SingleFieldManager.java:264)
> 	at org.apache.openjpa.kernel.StateManagerImpl.cascadePersist(StateManagerImpl.java:2741)
> 	at org.apache.openjpa.kernel.BrokerImpl.persist(BrokerImpl.java:2432)
> 	at org.apache.openjpa.kernel.BrokerImpl.persist(BrokerImpl.java:2251)
> 	at org.apache.openjpa.kernel.DelegatingBroker.persist(DelegatingBroker.java:1010)
> 	at org.apache.openjpa.persistence.EntityManagerImpl.persist(EntityManagerImpl.java:541)
> 	at tree.TestTree.testPersistTree(TestTree.java:22)
> 	
> 	
> Caused by: java.lang.ClassCastException: org.apache.openjpa.util.ObjectId
> 	at tree.Node.pcCopyKeyFieldsToObjectId(Node.java)
> 	at org.apache.openjpa.util.ApplicationIds.create(ApplicationIds.java:386)
> 	at org.apache.openjpa.kernel.BrokerImpl.persist(BrokerImpl.java:2405)
> 	... 27 more
> 3. How does the enhanced Node class looks like?
>    Firstly, Java decomiler (jad) itself is confused about the same method.
>    $ jad classes\tree\Node.class
>      Parsing classes\tree\Node.class...Parsing inner class classes\tree\Node$SelfId.class...Overwrite Node.jad [y/n/a/s] ? y
>      Generating Node.jad
>      Couldn't fully decompile method pcCopyKeyFieldsToObjectId
>    The decompiled file is attached.
> 4. The enhanced code is generated in
>    PCEnhancer.addExtractObjectIdFieldValueCode()
>    But I found the debugging the piece hard.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.