openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Pinaki Poddar" <ppod...@bea.com>
Subject RE: Persistence of Service Data Objects using OpenJPA
Date Fri, 27 Jul 2007 21:11:51 GMT
Hello Luciano,

>> I haven't looked into this with any great detail, 
  
  Thank you for your interest.

  The idea of Fluid is to keep it simple and use standard API contracts
-- so there is not much detail to look at:)
Just download (it is order of 100KB) and run. Or browse the test cases
to see how it is being used.

>> but I'd like to ask if would be feasible to add a DAS interface[1]
layer on top of Fluid [2].

  I have briefly looked at DAS and saw it using a Command Object
pattern. That kind of interface is generic enough to put on top of
almost anything -- so broadly speaking -- answer to your question is:
Yes. 

What worried me about DAS is its SQL-bias as expressed in methods like:
Command createCommand(String sql);

Notionally, JPA and object persistence layer prefers to operate on an
"Object View" -- the view being the Java domain model -- while actual
model is "database schema". The promise of JPA is to allow this
"updatable object view" of relational tables to the user. 
Shooting direct SQL from user application breaks that notion (though due
to public demand, JPA spec allows SQL to be executed on the side -- but
that is not kosher) 

  It is also important to note at this point that a robust persistence
mechanism that supports disconnect application programming model must
require the domain objects to carry two important bits of information:
identity & version. SDO Spec is silent about these aspects at this
moment -- but may be future versions will pay attention to such details.

  I have asked the question on identity scheme used by current DAS
implementation (but did not get an answer). Fluid
makes some convention which can be questioned but they are clearly spelt
out. Anyway, these restrictions are planned to relaxed. 


My comments against each of the DAS API method follow: 


interface DAS {
   // This will become merge() in JPA - very doable
   void applyChanges(DataObject root);

   // Possible to commandify JPA EntityManager API: Persist, Update,
Query, Find, Remove 
    Command getCommand(String name);

   // Not good: may be with some restriction
   Command createCommand(String sql);

   // Yes
   void releaseResources();
} 

public interface Command {
    // Depends what commands are served by DAS
    void execute();

    // Possible the result should be List<DataObject> rather than
DataObject so that one can iterate for large returns
    DataObject executeQuery();

    // Again very SQL centric and will work for restricted cases where
command is Query
    void setParameter(int index, Object value);
    Object getParameter(int index);

    // This will work with very restricted sense of identity. JPA has
more powerful feature of identity
    // which Fluid has only partially exposed 
    int getGeneratedKey();
}


Pinaki Poddar
972.834.2865

-----Original Message-----
From: Luciano Resende [mailto:luckbr1975@gmail.com] 
Sent: Friday, July 27, 2007 3:25 PM
To: tuscany-dev@ws.apache.org
Subject: Re: Persistence of Service Data Objects using OpenJPA

I haven't looked into this with any great detail, but I'd like to ask if
would be feasible to add a DAS interface[1] layer on top of Fluid [2].

Thoughts ?

[1]
https://svn.apache.org/repos/asf/incubator/tuscany/java/das/rdb/src/main
/java/org/apache/tuscany/das/rdb/DAS.java
[2] http://people.apache.org/~ppoddar/fluid/site/welcome.html

On 7/27/07, Frank Budinsky <frankb@ca.ibm.com> wrote:
> Hi Pinaki,
>
> I think your project has a lot of potential. Another related aspect is

> that there is interest in the SDO specification charter to discuss the

> possibility of converging/supporting JAXB with static SDO in version
3.
> JAXB2, as you may know, has a similar objective to support standard
POJOs.
> If that happens, then I see even more synergy with your approach. At 
> Tuscany, we'd also be very interested in the bytecode generation that 
> you mentioned you've prototyped in Fluid. Tuscany has done a little 
> bit of sandbox prototyping in that area as well, but nobody has had 
> enough time to really pursue it.
>
> Given this work, I think you should consider getting involved with the

> DAS specification effort, if you haven't already.
>
> Frank.
>
>
>
>
> "Pinaki Poddar" <ppoddar@bea.com>
> 07/27/2007 12:32 PM
> Please respond to
> tuscany-dev@ws.apache.org
>
>
> To
> <tuscany-dev@ws.apache.org>
> cc
> <dev@openjpa.apache.org>
> Subject
> Persistence of Service Data Objects using OpenJPA
>
>
>
>
>
>
> Hi Frank,
>
>    Thank you for your interest. Before I answer the specific questions

> let me say few words to explain why Fluid is attempted at first place.
>
>    A) there is a natural synergy between SDO model and the POJO model 
> where JPA excels.
>    B) Object Persistence (be it strongly-typed POJO or loosely-typed
> SDO) is complex problem by itself. Throw in quirk of multiple database

> vendors -- and one really got a hard problem in hand. OpenJPA, 
> Hibernate, TopLink had worked over many years to solve this problem to

> a reasonable degree. SDO persistence should leverage that 
> knowledge/effort instead of reinventing a costly wheel.
>
>
>  1)  Yes -- you're right. The views expressed in
>
> http://www.osoa.org/display/Main/SDO+and+Java+Persistence+Architecture
> +%
> 28JPA%29 is in agreement with what promoted me to start the lab 
> (actually I checked this page while researching and mentioned in in my

> blog). According to this wiki page: "Work on the specification of a 
> JPA data access service by the OSOA collaboration is envisaged 
> sometime in the future but has not started yet."
>    So Fluid can be considered as a prototype/exploration of similar 
> ideas.
>
>  2) Yes, I am aware of SDO's generation of Java and I have played with

> org.apache.tuscany.sdo.generate.XSD2JavaGenerator briefly.  I decided 
> to write yet another code generator for Fluid because
>
>     a) SDO2POJOGenerator handles JPA annotations as it sources SDO 
> Type information defined by XSDHelper.define().
> The annotations in the class will be made optional as we proceed -- 
> because JPA allows a separate mapping file Orm.xml thereby retaining 
> POJO-ness of the persistent domain classes.
>
>     b) The generated Java classes is "domain-ready" i.e. they are 
> self-consistent and independent. Proof: they can be compiled without 
> any other package other that java.util.*. Adding annotation, however, 
> makes them compile-dependent on jpa.jar, but that dependency will be 
> made optional as mapping information can be externalized in orm.xml.
>
>     C) as far as SDO metamodel to Java metamodel mapping goes, it is 
> essentially isomorphic (at least for this prototype).
> SDO2POJO does not introduce any extra artifact (see below). I have not

> played enough with org.apache.tuscany.sdo.generate.XSD2JavaGenerator 
> to make a fair judgement -- but most probably it is generating Java 
> classes which has more dependency on framework classes in terms of
> (inheriting/implementing) as well as package imports.
>
>    d) One mapping element is introduced in SDO-Java conversion
process.
> It is about "Container" SDO Types. I describe the details in Fluid 
> website. The reason for defining a Container Type abstraction is 
> better alignment for mapping to relational database (save extra joins 
> and allow navigational query across multi-cardinality relation paths).
>
>
>   3) Besides the source code generation route, another possibilities 
> is dynamic Java bytecode generation for SDO Types. Fluid prototyped 
> that approach too.
>
>   Regards --
>
>
> Pinaki Poddar
> 972.834.2865
>
> -----Original Message-----
> From: Frank Budinsky [mailto:frankb@ca.ibm.com]
> Sent: Friday, July 27, 2007 10:39 AM
> To: tuscany-dev@ws.apache.org
> Subject: RE: How does xsd:ID property type is distinguished from 
> xsd:string
>
> Hi Pinaki,
>
> This looks very cool, especially that you've got a working prototype! 
> I have a couple of questions.
>
> 1) This seems to overlap with the DAS plan described here:
>
>
> http://www.osoa.org/display/Main/SDO+and+Java+Persistence+Architecture
> +%
> 28JPA%29
>
> I'm not very familiar with the DAS work myself, but there seems to be 
> some overlap with your work.
>
> 2) I'm also wondering if you were aware that SDO also defines a 
> mapping to static Java interfaces/classes. See section 5 of the SDO
spec:
>
>
> http://www.osoa.org/download/attachments/36/Java-SDO-Spec-v2.1.0-FINAL
> .p
> df?version=1
>
> Does your SDO2POJOGenerator conform to that mapping? Have you tried 
> the Tuscany static SDO generator 
> (org.apache.tuscany.sdo.generate.XSD2JavaGenerator)? Is there any 
> relashionship?
>
> Frank.
>
> "Pinaki Poddar" <ppoddar@bea.com> wrote on 07/26/2007 10:41:27 PM:
>
> > Hello,
> >   I have been asking newbie questions to this forum. And people have

> > been very helpful.
> >
> >   Those newbie questions were for a Apache Lab project named Fluid 
> > -- to explore possibilities of SDO persistence using JPA API.
> >
> >   With all your help, I could put together an initial proptotype 
> > that creates/updates/queries SDO using JPA API. I used Tuscany and 
> > OpenJPA for SDO and JPA implementation respectively.
> >
> >    Further details of this project (with user documentation) can be 
> > found at
> >
> >      http://people.apache.org/~ppoddar/fluid/site/welcome.html
> >
> >    Your comments/suggestions are most welcome --
> >
> >
> > Pinaki Poddar
> > 972.834.2865
> >
> > -----Original Message-----
> > From: Frank Budinsky [mailto:frankb@ca.ibm.com]
> > Sent: Tuesday, July 24, 2007 4:59 PM
> > To: tuscany-dev@ws.apache.org
> > Subject: RE: How does xsd:ID property type is distinguished from 
> > xsd:string
> >
> > EAttribute.isID() will only be true if the type is xsd:ID.
> >
> > Frank.
> >
> > "Pinaki Poddar" <ppoddar@bea.com> wrote on 07/24/2007 05:31:09 PM:
> >
> > > Hi Frank,
> > >    > Database IDs (e.g., primary and foreign keys) are more 
> > > related to
> >
> > > xsd:key/xsd:keyref, then xsd:ID, but fortunately SDO 3 is planning

> > > to address all of them :-)
> > >
> > >   Thanks for telling me this.
> > >
> > >   Now is ((property.getType().isDataType()) &&
> > > ((EAttribute)property).isID())) == true for a property p that is 
> > > declared as xsd:key or xsd:keyref?
> > >
> > >   Or broadly, what is the semantics of Eattribute.isID()?
> > >
> > >
> > > Pinaki Poddar
> > > 972.834.2865
> > >
> > > -----Original Message-----
> > > From: Frank Budinsky [mailto:frankb@ca.ibm.com]
> > > Sent: Tuesday, July 24, 2007 3:01 PM
> > > To: tuscany-dev@ws.apache.org
> > > Subject: RE: How does xsd:ID property type is distinguished from 
> > > xsd:string
> > >
> > > Hi Pinaki,
> > >
> > > Identity support is also in the SDO 3 charter: Support for a 
> > > concept
>
> > > of identity in SDO, and its relationship to other technologies.
> > >
> > > Database IDs (e.g., primary and foreign keys) are more related to 
> > > xsd:key/xsd:keyref, then xsd:ID, but fortunately SDO 3 is planning

> > > to address all of them :-)
> > >
> > > Frank.
> > >
> > > "Pinaki Poddar" <ppoddar@bea.com> wrote on 07/24/2007 11:02:21 AM:
> > >
> > > > Hi Frank,
> > > >   Thanks.
> > > >
> > > > > SDO (SDO 3) is planning to provide an api for accessing 
> > > > > extended
>
> > > > > XSD
> > > > metadata
> > > >
> > > >   That is good news. However, identity mechanics should appear 
> > > > more distinctly on API surface e.g.
> > > >    boolean Proprty.isIdentifier();
> > > >    List<Property> Type.getIdentifiers();
> > > >
> > > > I will qualify absence of any identity semantics from SDO a 
> > > > major drawback. Especially, if it comes to any persistence 
> > > > operation on SDO DataObject/DataGraph. Hopefully some of the SDO

> > > > spec writers would notice this omission and add it to future
spec version.
> > > >
> > > > After a quick pick in current DAS implementation, it appeared 
> > > > that
>
> > > > 'primary key' identification is based on existing database 
> > > > column name
> > >
> > > > "ID" (yes, hardcoded) -- but I may be wrong and am ready to 
> > > > learn how DAS is handling identity issue.
> > > >
> > > > > SDO (SDO 3) is planning to provide an api for accessing 
> > > > > extended
>
> > > > > XSD
> > > > metadata
> > > > That is a good decision. Wrapping should always provide access 
> > > > to what
> > >
> > > > is being wrapped.
> > > >
> > > > > downcasting to the EMF implementation class
> > > >
> > > > Thanks for this info. I will do this for now. But I heed your 
> > > > advice
> >
> > > > and have already a scheme in place that programs against *only* 
> > > > commonj.sdo API but can access underlying implementaion, if 
> > > > available,
> > >
> > > > without any compile-time binding.
> > > > Slightly costly -- but works for, say, extracting package name 
> > > > from Types.
> > > >
> > > >
> > > >
> > > > Pinaki Poddar
> > > > 972.834.2865
> > > >
> > > > -----Original Message-----
> > > > From: Frank Budinsky [mailto:frankb@ca.ibm.com]
> > > > Sent: Tuesday, July 24, 2007 9:16 AM
> > > > To: tuscany-dev@ws.apache.org
> > > > Subject: Re: How does xsd:ID property type is distinguished from

> > > > xsd:string
> > > >
> > > > Hi Pinaki,
> > > >
> > > > They can't be distinguished in the current version of SDO 
> > > > metadata, you need to look at the original XSD. The next version

> > > > of SDO (SDO
> > > > 3) is planning to provide an api for accessing extended XSD 
> > > > metadata. In Tuscany, you can currently determine this by 
> > > > downcasting to the EMF implementation class, although we don't
> > recommend people do that:
> > > >
> > > >       System.out.println("  Property isID: " +
> > > > ((property.getType().isDataType()) && 
> > > > ((EAttribute)property).isID()));
> > > >
> > > > Frank.
> > > >
> > > > "Pinaki Poddar" <ppoddar@bea.com> wrote on 07/24/2007 01:00:03
AM:
> > > >
> > > > >
> > > > > Hi,
> > > > >
> > > > > A newbie question:
> > > > >    How does two Property: one defined as xsd:string and other 
> > > > > as
>
> > > > > xsd:ID can be distiguished?
> > > > >
> > > > > Assume:
> > > > >  1. we have a simple XML schema defining a Person SDO Type 
> > > > > with two properties as follows:
> > > > >     <xsd:complexType name="Person">
> > > > >          <xsd:attribute name="firstName" type="xsd:string"/>
> > > > >          <xsd:attribute name="id"        type="xsd:ID"/>
> > > > >     </xsd:complexType>
> > > > >
> > > > >  2. TypeHelper.INSTANCE.define()
> > > > >     defines SDO Type with two commonj.sdo.Property, p1 (for
> > > > > firstName)
> > > >
> > > > > and p2 (for id)
> > > > >
> > > > >  3. both p1.getType().getInstanceClass() and
> > > > > p2.getType().getInstanceClass() return java.lang.String
> > > > >     both p1.getType().isDataType() and 
> > > > > p2.getType().isDataType()
>
> > > > > return true
> > > > >
> > > > > The question is, what can be done to identify p2 as a property

> > > > > that were defined as xsd:ID?
> > > > >
> > > > >
> > > > > Thanks for your help --
> > > > >
> > > > >
> > > > >
> > > > >
> > > > > Pinaki Poddar
> > > > > 972.834.2865
> > > > >
> > > > > Notice:  This email message, together with any attachments, 
> > > > > may contain information  of  BEA Systems,  Inc.,  its 
> > > > > subsidiaries and affiliated entities,  that may be 
> > > > > confidential, proprietary, copyrighted  and/or legally 
> > > > > privileged, and is intended solely for
> >
> > > > > the
> > > >
> > > > > use of the individual or entity named in this message. If you 
> > > > > are not the intended recipient, and have received this message

> > > > > in error,
> > >
> > > > > please immediately return this by email and then delete it.
> > > > >
> > > > > --------------------------------------------------------------
> > > > > --
> > > > > --
> > > > > --
> > > > > - To unsubscribe, e-mail: 
> > > > > tuscany-dev-unsubscribe@ws.apache.org
> > > > > For additional commands, e-mail: 
> > > > > tuscany-dev-help@ws.apache.org
> > > > >
> > > >
> > > >
> > > > ----------------------------------------------------------------
> > > > --
> > > > --
> > > > - To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> > > > For additional commands, e-mail: tuscany-dev-help@ws.apache.org
> > > >
> > > >
> > > > Notice:  This email message, together with any attachments, may 
> > > > contain information  of  BEA Systems,  Inc.,  its subsidiaries 
> > > > and affiliated entities,  that may be confidential,  
> > > > proprietary, copyrighted  and/or legally privileged, and is 
> > > > intended solely for
>
> > > > the
> > >
> > > > use of the individual or entity named in this message. If you 
> > > > are not the intended recipient, and have received this message 
> > > > in error,
> >
> > > > please immediately return this by email and then delete it.
> > > >
> > > > ----------------------------------------------------------------
> > > > --
> > > > --
> > > > - To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> > > > For additional commands, e-mail: tuscany-dev-help@ws.apache.org
> > > >
> > >
> > >
> > > ------------------------------------------------------------------
> > > --
> > > - To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> > > For additional commands, e-mail: tuscany-dev-help@ws.apache.org
> > >
> > >
> > > Notice:  This email message, together with any attachments, may 
> > > contain information  of  BEA Systems,  Inc.,  its subsidiaries  
> > > and affiliated entities,  that may be confidential,  proprietary, 
> > > copyrighted  and/or legally privileged, and is intended solely for

> > > the
> >
> > > use of the individual or entity named in this message. If you are 
> > > not the intended recipient, and have received this message in 
> > > error,
>
> > > please immediately return this by email and then delete it.
> > >
> > > ------------------------------------------------------------------
> > > --
> > > - To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> > > For additional commands, e-mail: tuscany-dev-help@ws.apache.org
> > >
> >
> >
> > --------------------------------------------------------------------
> > - To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> > For additional commands, e-mail: tuscany-dev-help@ws.apache.org
> >
> >
> > Notice:  This email message, together with any attachments, may 
> > contain information  of  BEA Systems,  Inc.,  its subsidiaries  and 
> > affiliated entities,  that may be confidential,  proprietary, 
> > copyrighted  and/or legally privileged, and is intended solely for 
> > the
>
> > use of the individual or entity named in this message. If you are 
> > not the intended recipient, and have received this message in error,

> > please immediately return this by email and then delete it.
> >
> > --------------------------------------------------------------------
> > - To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> > For additional commands, e-mail: tuscany-dev-help@ws.apache.org
> >
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> For additional commands, e-mail: tuscany-dev-help@ws.apache.org
>
>
> Notice:  This email message, together with any attachments, may 
> contain information  of  BEA Systems,  Inc.,  its subsidiaries  and  
> affiliated entities,  that may be confidential,  proprietary,  
> copyrighted  and/or legally privileged, and is intended solely for the

> use of the individual or entity named in this message. If you are not 
> the intended recipient, and have received this message in error, 
> please immediately return this by email and then delete it.
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> For additional commands, e-mail: tuscany-dev-help@ws.apache.org
>
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
> For additional commands, e-mail: tuscany-dev-help@ws.apache.org
>
>


--
Luciano Resende
Apache Tuscany Committer
http://people.apache.org/~lresende
http://lresende.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: tuscany-dev-unsubscribe@ws.apache.org
For additional commands, e-mail: tuscany-dev-help@ws.apache.org


Notice:  This email message, together with any attachments, may contain information  of  BEA
Systems,  Inc.,  its subsidiaries  and  affiliated entities,  that may be confidential,  proprietary,
 copyrighted  and/or legally privileged, and is intended solely for the use of the individual
or entity named in this message. If you are not the intended recipient, and have received
this message in error, please immediately return this by email and then delete it.

Mime
View raw message