Yet another CMP question....

[Rick McGuire]

Found another thing I'm trying to understand.  In the mapClass1x()
method of CmpJpaConversion, there's the following code:

                pkClass = classLoader.loadClass(bean.getPrimKeyClass());
                MappedSuperclass superclass = null;
                for (java.lang.reflect.Field pkField :
pkClass.getFields()) {
                    String fieldName = pkField.getName();
                    int modifiers = pkField.getModifiers();
                    // the primary key fields must be public,
non-static, must be defined as a CMP field,
                    // AND must also exist in the class hierarchy (not
enforced by mapFields());
                    if (Modifier.isPublic(modifiers) &&
!Modifier.isStatic(modifiers) && allFields.contains(fieldName)) {
                        superclass = superclassByField.get(fieldName);
                        if (superclass == null) {
                            throw new IllegalStateException("Primary key
field " + fieldName + " is not defined in class " + ejbClassName + " or
any super classes");
                        }
                        superclass.addField(new Id(fieldName));
                        mapping.addField(new AttributeOverride(fieldName));
                        primaryKeyFields.add(fieldName);
                    }
                }
                if (superclass != null) {
                    superclass.setIdClass(new
IdClass(bean.getPrimKeyClass()));
                }

I have a question about the last couple of lines.  If I'm reading what
this (and previously executed) is doing correctly, it's setting the
IdClass() attribute on the class that happens to define the last of the
primary key fields.  However, if I'm reading things properly, the
primary key fields might not all be implemented at the same level of the
class hierarchy.  Is there potentially a problem here if that's the
case?  Would defining this attribute too early in the class hierarchy be
a problem?  That is, it's possible that it might be defined at a level
where not all of the required fields exist yet.

Rick

[Dain Sundstrom]

On Jul 9, 2008, at 9:08 AM, Rick McGuire wrote:

> Found another thing I'm trying to understand.  In the mapClass1x()  
> method of CmpJpaConversion, there's the following code:
>
>               pkClass = classLoader.loadClass(bean.getPrimKeyClass());
>               MappedSuperclass superclass = null;
>               for (java.lang.reflect.Field pkField :  
> pkClass.getFields()) {
>                   String fieldName = pkField.getName();
>                   int modifiers = pkField.getModifiers();
>                   // the primary key fields must be public, non-
> static, must be defined as a CMP field,
>                   // AND must also exist in the class hierarchy (not  
> enforced by mapFields());
>                   if (Modifier.isPublic(modifiers) && !
> Modifier.isStatic(modifiers) && allFields.contains(fieldName)) {
>                       superclass = superclassByField.get(fieldName);
>                       if (superclass == null) {
>                           throw new IllegalStateException("Primary  
> key field " + fieldName + " is not defined in class " + ejbClassName  
> + " or any super classes");
>                       }
>                       superclass.addField(new Id(fieldName));
>                       mapping.addField(new  
> AttributeOverride(fieldName));
>                       primaryKeyFields.add(fieldName);
>                   }
>               }
>               if (superclass != null) {
>                   superclass.setIdClass(new  
> IdClass(bean.getPrimKeyClass()));
>               }
>
> I have a question about the last couple of lines.  If I'm reading  
> what this (and previously executed) is doing correctly, it's setting  
> the IdClass() attribute on the class that happens to define the last  
> of the primary key fields.  However, if I'm reading things properly,  
> the primary key fields might not all be implemented at the same  
> level of the class hierarchy.  Is there potentially a problem here  
> if that's the case?  Would defining this attribute too early in the  
> class hierarchy be a problem?  That is, it's possible that it might  
> be defined at a level where not all of the required fields exist yet.

Ya, that seems weird to me.  I don't think any of the OpenEJB itests  
or TCK tests have code where the fields of the primary key class are  
split over the inheritance hierarchy.  I believe it is an error if you  
set the IdClass is set in the wrong mapped superclass.  I wonder if it  
is possible to declare the IdClass on the last bean in the hierarchy.  
Anyway, you will have to check the spec on this one, and may have to  
experiment (see JpaTest).

-dain

[KMalhi]

you can also look at this

http://openjpa.apache.org/docs/latest/manual/manual.html#jpa_overview_pc_identity_hierarchy

On Wed, Jul 9, 2008 at 1:19 PM, Dain Sundstrom <dain@...> wrote:

> On Jul 9, 2008, at 9:08 AM, Rick McGuire wrote:
>
>  Found another thing I'm trying to understand.  In the mapClass1x() method
>> of CmpJpaConversion, there's the following code:
>>
>>              pkClass = classLoader.loadClass(bean.getPrimKeyClass());
>>              MappedSuperclass superclass = null;
>>              for (java.lang.reflect.Field pkField : pkClass.getFields()) {
>>                  String fieldName = pkField.getName();
>>                  int modifiers = pkField.getModifiers();
>>                  // the primary key fields must be public, non-static,
>> must be defined as a CMP field,
>>                  // AND must also exist in the class hierarchy (not
>> enforced by mapFields());
>>                  if (Modifier.isPublic(modifiers) &&
>> !Modifier.isStatic(modifiers) && allFields.contains(fieldName)) {
>>                      superclass = superclassByField.get(fieldName);
>>                      if (superclass == null) {
>>                          throw new IllegalStateException("Primary key
>> field " + fieldName + " is not defined in class " + ejbClassName + " or any
>> super classes");
>>                      }
>>                      superclass.addField(new Id(fieldName));
>>                      mapping.addField(new AttributeOverride(fieldName));
>>                      primaryKeyFields.add(fieldName);
>>                  }
>>              }
>>              if (superclass != null) {
>>                  superclass.setIdClass(new
>> IdClass(bean.getPrimKeyClass()));
>>              }
>>
>> I have a question about the last couple of lines.  If I'm reading what
>> this (and previously executed) is doing correctly, it's setting the
>> IdClass() attribute on the class that happens to define the last of the
>> primary key fields.  However, if I'm reading things properly, the primary
>> key fields might not all be implemented at the same level of the class
>> hierarchy.  Is there potentially a problem here if that's the case?  Would
>> defining this attribute too early in the class hierarchy be a problem?  That
>> is, it's possible that it might be defined at a level where not all of the
>> required fields exist yet.
>>
>
> Ya, that seems weird to me.  I don't think any of the OpenEJB itests or TCK
> tests have code where the fields of the primary key class are split over the
> inheritance hierarchy.  I believe it is an error if you set the IdClass is
> set in the wrong mapped superclass.  I wonder if it is possible to declare
> the IdClass on the last bean in the hierarchy.  Anyway, you will have to
> check the spec on this one, and may have to experiment (see JpaTest).
>
> -dain
>

--
Karan Singh Malhi