jcr mapping question

[Padraic Hannon <pi...@wasabicowboy.com>]

We are using the repository pattern for our object retrieval 
(http://www.martinfowler.com/eaaCatalog/repository.html) and I am trying 
to migrate from a typical ORM system (toplink, hibernate, etc) backed by 
an RDBMS to jackrabbit (actually CRX) using the jcr-mapping project.

It appears that for child object the jcr project does what I would 
assume (NTCollectionConverterImpl) and stores the objects that haven't 
already been stored. However, if one does not one a referenced object to 
be a child but rather just a property with a reference then things start 
to diverge from what I would assume would happen. Perhaps, the error is 
in my modeling (entirely possible!), however, I would think that the 
Reference converters for both beans and collections would do more than 
just persist the UUIDs. Would it be horrible if these where updated to 
persist the referenced object is a UUID is not present prior to 
persisting the referenced object? Of course I think it should probably 
update the referenced objects properties if the object already exists. 
Does this make sense? It adds complexity, but it seems like that would 
be the expected functionality.

-paddy

Re: jcr mapping question

[Christophe Lombart <ch...@gmail.com>]

Hi,

Right now, I don't see strong use cases where it is interesting to store
"more" than the UUID with BeanReferenceCollectionConverterImpl &
ReferenceBeanConverterImpl. Is it not possible to have 2 steps : 1/ create
the referenceable object 2/ add the reference into another object ? If not,
keep your converter and we will see if others want to have the same
requirement. It is too early to modify thoses converters. I would like to
wait. You can also add a new Jira issue. By this way, others can vote on it.


br,
Christophe


On 8/16/07, Padraic Hannon <pi...@wasabicowboy.com> wrote:
>
> We are using the repository pattern for our object retrieval
> (http://www.martinfowler.com/eaaCatalog/repository.html) and I am trying
> to migrate from a typical ORM system (toplink, hibernate, etc) backed by
> an RDBMS to jackrabbit (actually CRX) using the jcr-mapping project.
>
> It appears that for child object the jcr project does what I would
> assume (NTCollectionConverterImpl) and stores the objects that haven't
> already been stored. However, if one does not one a referenced object to
> be a child but rather just a property with a reference then things start
> to diverge from what I would assume would happen. Perhaps, the error is
> in my modeling (entirely possible!), however, I would think that the
> Reference converters for both beans and collections would do more than
> just persist the UUIDs.






Would it be horrible if these where updated to
> persist the referenced object is a UUID is not present prior to
> persisting the referenced object? Of course I think it should probably
> update the referenced objects properties if the object already exists.
> Does this make sense? It adds complexity, but it seems like that would
> be the expected functionality.
>
> -paddy
>