Re: [Objectbridge-jdo-dev] Instrumentation/Enhancement pros and cons

[Thomas M. <tho...@ho...>]

Hi Jim,

Jim Hughes wrote:

> Hi.  My name is Jim Hughes, and I hope to be able to contribute in some
> way to this project.
> 
> Anyway, my understanding is that if we do enhancement, the way we do it
> is pretty well specified, if we want to comply with the spec, that is.
> Classes need to be modified to implement PersistenceCapable, etc.  The
> spec even states that enhanced classes must be portable between
> different JDO impls, which really constrains us.  E.g., we couldn't do
> this:
> 
> // way pseudo!
> package org.objectbridge.jdo;
> interface PersistenceCapable extends javax.jdo.PersistanceCapable
> {
> 	void persistYourself() ;
> }
> 
> public class PersistenceManager implements javax.jdo.PersistanceManager
> {
> 	void makePersistent(Object o)
> 	{
> 		PersistenceCapable pco =
> (org.objectbridge.jdo.PersistenceCapable)o;
> 		pco.persistYourself();
> 	}
> }
> 
> because this code would not work with PersistenceCapable class enhanced
> by other impls.


I totally agree with you !


> 
> I would be interested in working on the enhancer.  Is anyone working on
> it yet?


No, there is not much real work going on just now.
Christian Rath and I had some plans on writing a first prototype based 
on the JDO RI.

The RI has a clear layered architecture with a StoreManager interface 
that allows to abstract from implementations (as the default FOStore 
implementation).
Our idea was just to provide simply a OJB PersistenceBroker based 
StoreManager.

But we have been so busy with other things that we did not yet start on 
this !

cu,

Thomas


> Jim
> 
> 
> -----Original Message-----
> From: obj...@li...
> [mailto:obj...@li...] On Behalf Of g
> Sent: Monday, March 04, 02 10:48 AM
> To: obj...@li...
> Subject: [Objectbridge-jdo-dev] Instrumentation/Enhancement pros and
> cons
> 
> 
> Hi all,
> 	I have been reading up on the controversy over some of JDO's
> features, and have reaqd some of the postings on this list regarding
> JDO. I understand and agree with the concerns of some people who have
> found that classees that have been enhanced through byte code
> instrumentation may behave unpredictably, for example when debugging
> information is lost, mangled, or not properly inserted into an
> instrumented class file. The seemingly mysterious nature of this
> instrumenting process does evoke a sense of worry, but I have not seen
> any discussions of what the alternatives are.
> 
> Have we as a group had any discussion about how we might avoid choosing
> the abovementioned approach? Perhaps a compromise like inserting just
> enough code into the enhanced class to provide proper identity
> attributes, which would enable us to use metadata accessed from behind
> the facade of the PersistenceManager?
> 
> I have some experience with the BCEL byte code libraries, and have some
> code that traverses compiled code and can do analysis or make changes.
> This code uses BCEL. It's pretty straightforward.
> 
> Thanks,
> 
> George Coles
> 
> 
> _______________________________________________
> Objectbridge-jdo-dev mailing list
> Obj...@li...
> https://lists.sourceforge.net/lists/listinfo/objectbridge-jdo-dev
> 
> 
> _______________________________________________
> Objectbridge-jdo-dev mailing list
> Obj...@li...
> https://lists.sourceforge.net/lists/listinfo/objectbridge-jdo-dev
> 
> 
> 
>