Home » Modeling » UML2 » UML2 + CDO
| UML2 + CDO [message #478370] |
Fri, 08 May 2009 12:24  |
Eclipse User |
|
|
|
Originally posted by: maurice.gmail.com
Hi,
I recently learned that UML2 is currently incompatible with CDO. Are there
plans to enable UML2 for CDO in the near feature and if so, will it be in
the next milestone or further down the line?
Thanks!
Maurice
|
|
| | | |
| Re: UML2 + CDO [message #478374 is a reply to message #478373] |
Sat, 09 May 2009 14:23  |
 Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
This is a multi-part message in MIME format.
--------------080303090705090501010600
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Vlad,
Comments below.
Vlad Varnica wrote:
> Hi Maurice,
>
> You have many different ways to generate Jee & Mapping from a model.
I don't know, will you count the ways?
> What is important is to take a clean and standard XMI 2.1 or 2.2 and
> then use one of the commercial or open source tool.
Not sure how that relates to CDO?
>
> I tried the following tools which are really good to generate code
> from an UML Superstructure Model XMI. I didn't spend more time to
> investigate other tools but there could be more good tools/solutions
> on the market:
>
> - BLU AGE: an MDA compliant software workbench that enables 100%
> automated transformation of your UML 2.X diagrams into Java EE and
> .Net Enterprise business applications.
> - AndroMDA is open source http://galaxy.andromda.org/forum/
Maurice didn't ask about generating JEE, so how exactly is this relevant?
>
> CDO is a very nice integration for EMF.
Indeed.
> btw,I don't think it could be extended to generate what you need from
> UML2 because only related to EMF and not to the UML Superstructure Model.
CDO has nothing to do with generation, so I'm confused why you bring all
this up.
> It is important not to make a confusion between EMF and UML2.
I'm totally confused by your whole note...
> EMF is a great framework that we use inside Omondo and UML 2 is the
> XMI serialization of an EMF model but it doesn't mean that EMF is UML2.
Indeed, that's why one they're two different acronyms. Why are *you
*making confusion between them? And why are you confusing CDO with JEE
generation? Maybe the coffee hasn't kicked in yet.
>
> Vlad,
>
--------------080303090705090501010600
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Vlad,<br>
<br>
Comments below.<br>
<br>
Vlad Varnica wrote:
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite">Hi Maurice,
<br>
<br>
You have many different ways to generate Jee & Mapping from a model.</blockquote>
I don't know, will you count the ways?<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> What is important is to take a clean and standard XMI 2.1
or 2.2 and then use one of the commercial or open source tool.
<br>
</blockquote>
Not sure how that relates to CDO?<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"><br>
I tried the following tools which are really good to generate code from
an UML Superstructure Model XMI. I didn't spend more time to
investigate other tools but there could be more good tools/solutions on
the market:
<br>
<br>
- BLU AGE: an MDA compliant software workbench that enables 100%
automated transformation of your UML 2.X diagrams into Java EE and .Net
Enterprise business applications.
<br>
- AndroMDA is open source <a class="moz-txt-link-freetext" href="http://galaxy.andromda.org/forum/">http://galaxy.andromda.org/forum/</a>
<br>
</blockquote>
Maurice didn't ask about generating JEE, so how exactly is this
relevant?<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"><br>
CDO is a very nice integration for EMF.</blockquote>
Indeed.<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> btw,I don't think it could be extended to generate what
you need from UML2 because only related to EMF and not to the UML
Superstructure Model.</blockquote>
CDO has nothing to do with generation, so I'm confused why you bring
all this up.<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> It is important not to make a confusion between EMF and
UML2.</blockquote>
I'm totally confused by your whole note...<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> EMF is a great framework that we use inside Omondo and
UML 2 is the XMI serialization of an EMF model but it doesn't mean that
EMF is UML2.
<br>
</blockquote>
Indeed, that's why one they're two different acronyms.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| |
| Re: UML2 + CDO [message #478376 is a reply to message #478375] |
Sun, 10 May 2009 14:23 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Vlad,
Comments below.
Vlad Varnica wrote:
> Ed,
>
> I should take more coffee and stop smoking :-)
What have you been smoking? :-P
> CDO is a 3-tiers solution for distributed shared models that allow you
> to develop and use different mappers such as Hibernate.
> I therefore consider that a demand related to UML+CDO is dealing with
> persistence tiers generation from a model.
No, Maurice did not ask about generation. Maurice apparently wants UML
itself to be persisted using CDO. He's said nothing about what he's
doing with his UML instances other than wanting to persist them using CDO.
> It seems that Maurice is interested in UML and not EMF ?
Given his interest in CDO and the fact that it's a component of EMF, I'd
say you are mistaken.
> Sorry if I am wrong but this seems to me logical.
I'm quite sure you are wrong. He might well want to generate JEE, but
there's no indication of that.
>
> This is a very interesting project but Omondo never used it because it
> is still in the early stage and for our customers we prefer to
> recommend a more stable solution. Our recommendation for a successful
> project are:
> - use UML 2 with Blue Age, AndroMDA or MIA software for persistence
> tiers generation from an UML Superstructure Model - Use Agile
> incremental Hibernate, JPA annotations with EclipseUML 2008 and Oracle
> 11G. See more at:
> http://www.forum-omondo.com/documentation_eclipseuml_2008/Ec lipse_Database/Activate_JPA_Perspective/oracle_integration.h tml
>
Is it really necessary to use this newsgroup in this way?
>
> We will certainly consider CDO again with Eclipse 3.5.
I doubt it, because while you use EMF, you don't actually want any of
your customers using EMF.
>
> I think it is important to help Eclipse users to know what is ready
> for production and what is not because this is a waste of time at the
> current stage.
What is a waste of time? I'm starting to think your comments are a
waste of time and space.
> Open source is useful only if it is not more time consuming than the
> benefit it really gives.
Climbing onto your soapbox again are you?
> You need few months just to understand EMF and few others months to
> understand CDO.
Time well spent.
> Do you really think companies have time to do that ?
Apparently a great many people do indeed have time, make time, and find
it time well spent. How much time does it take to understand UML?
What about the UML API itself? What about learning how your products
work? And then, when your company goes belly up, how much time does it
take all your customers to find an alternative?
> I think it is more important to give this kind of answer than just
> support first level which makes users feel that they understand while
> the reality is different !!
I think your commentary is so confusing and so off the mark that it does
a disservice.
>
>
> Vlad,
>
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| | |
| Re: UML2 + CDO [message #478381 is a reply to message #478371] |
Mon, 11 May 2009 14:27 |
| Eclipse User |
|
|
|
Originally posted by: maurice.gmail.com
Hi Ed,
Thanks for your reply. I see the UML2 metamodel is extending from, for
example, EModelElement from the ECore metamodel. Is this an example of
what Eike meant when he talked about ECore extensions in UML2, in this
post?
http://www.eclipse.org/newsportal/article.php?id=41535&g roup=eclipse.tools.emf#41535
When I removed these extensions from UML2's metamodel it was then capable
of generating a CDO-enabled genmodel which would actually work. Can you
enlighten me as to why UML2 extends ECore in this fashion? Is this the
reason why CDO can't hook into these metamodels?
As a short-term fix, to get my UML2 model ready to persist in a CDO
environment, would it be wise to strip these extensions? We're only
interested in UML2 class diagrams and only looking to persist instances of
these.
Maurice
Ed Merks wrote:
> Maurice,
> I know that Eike and Kenn have been investigating this. I don't think
> something will be ready for this release though...
> Maurice wrote:
>> Hi,
>> I recently learned that UML2 is currently incompatible with CDO. Are
>> there plans to enable UML2 for CDO in the near feature and if so, will
>> it be in the next milestone or further down the line?
>>
>> Thanks!
>>
>> Maurice
>>
|
|
|
| Re: UML2 + CDO [message #478382 is a reply to message #478381] |
Mon, 11 May 2009 14:40 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Maurice,
Yes, this is indeed the problem. UML extends EModelElement simply to
reuse Ecore's EAnnotation support. It was an unfortunate design
decision but given that it's part of the API, it's not easily changed.
Maurice wrote:
> Hi Ed,
> Thanks for your reply. I see the UML2 metamodel is extending from, for
> example, EModelElement from the ECore metamodel. Is this an example of
> what Eike meant when he talked about ECore extensions in UML2, in this
> post?
>
> http://www.eclipse.org/newsportal/article.php?id=41535&g roup=eclipse.tools.emf#41535
>
>
> When I removed these extensions from UML2's metamodel it was then
> capable of generating a CDO-enabled genmodel which would actually
> work. Can you enlighten me as to why UML2 extends ECore in this
> fashion? Is this the reason why CDO can't hook into these metamodels?
>
> As a short-term fix, to get my UML2 model ready to persist in a CDO
> environment, would it be wise to strip these extensions? We're only
> interested in UML2 class diagrams and only looking to persist
> instances of these.
>
> Maurice
>
> Ed Merks wrote:
>
>> Maurice,
>
>> I know that Eike and Kenn have been investigating this. I don't
>> think something will be ready for this release though...
>
>
>> Maurice wrote:
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so,
>>> will it be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: UML2 + CDO [message #478385 is a reply to message #478370] |
Tue, 12 May 2009 20:02 |
|
Maurice,
Yes, only today Kenn told me that he is currently working on a UML2
version that is ready for CDO. Indeed one of the main challenges is to
replace EMF's annotation support by something else. I must admit that
I'm right now not informed about how exactly Kenn will solve this issue
but we scheduled a discussion for tomorrow. I think we'll use the
opportunity to open a bugzilla so that all the interested can follow and
try out the patches that we come up with. I'll post a link here...
Cheers
/Eike
----
http://thegordian.blogspot.com
Maurice schrieb:
> Hi,
> I recently learned that UML2 is currently incompatible with CDO. Are
> there plans to enable UML2 for CDO in the near feature and if so, will
> it be in the next milestone or further down the line?
>
> Thanks!
>
> Maurice
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #478386 is a reply to message #478375] |
Tue, 12 May 2009 20:23 |
|
Vlad,
Comments below...
Vlad Varnica schrieb:
> Ed,
>
> I should take more coffee and stop smoking :-)
> CDO is a 3-tiers solution for distributed shared models that allow you
> to develop and use different mappers such as Hibernate.
So far so good.
> I therefore consider that a demand related to UML+CDO is dealing with
> persistence tiers generation from a model. It seems that Maurice is
> interested in UML and not EMF ? Sorry if I am wrong but this seems to
> me logical.
>
> This is a very interesting project but Omondo never used it because it
> is still in the early stage and for our customers we prefer to
> recommend a more stable solution.
What measures are you basing your statements about stability on?
> Our recommendation for a successful project are:
> - use UML 2 with Blue Age, AndroMDA or MIA software for persistence
> tiers generation from an UML Superstructure Model - Use Agile
> incremental Hibernate, JPA annotations with EclipseUML 2008 and Oracle
> 11G. See more at:
> http://www.forum-omondo.com/documentation_eclipseuml_2008/Ec lipse_Database/Activate_JPA_Perspective/oracle_integration.h tml
>
I don't know all of these products but are you really arguing that this
is *always* the best combination for *every* "successful project"?
>
> We will certainly consider CDO again with Eclipse 3.5.
Although it's not commercial?
>
> I think it is important to help Eclipse users to know what is ready
> for production and what is not
Good point! Are you volunteering to come up with more reliable help in
this regard? I guess those users would want to know where your expertise
is coming from.
> because this is a waste of time at the current stage.
Without proof this is defamation at best. It sounds like an abuse of
everybody who spent a lot of effort for making CDO what it is now. And
an abuse of all the users that are using CDO in production for many
years now, the NASA and the Canadian DoD for example, but many large
enterprises as well.
> Open source is useful only if it is not more time consuming than the
> benefit it really gives.
This is certainly not only valid for open source.
> You need few months just to understand EMF and few others months to
> understand CDO.
You should not deduce from your own experience.
Btw. shouldn't I remember at least a single newsgroup question of you
while you were spending months understanding CDO?
Could this be the reason why nobody else complained about so long
learning periods? They used the newsgroup to *ask questions* !
> Do you really think companies have time to do that ? I think it is
> more important to give this kind of answer than just support first
> level which makes users feel that they understand while the reality is
> different !!
I'm not sure I understand this one. Are you suggesting that we
deliberately give wrong advice here or do you think the users are just
too stupid to realize that we don't see the big picture?
Cheers
/Eike
----
http://thegordian.blogspot.com
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #478387 is a reply to message #478378] |
Tue, 12 May 2009 20:28 |
|
Patrick,
I appreciate your interest and I realize that I should have subscribed
to this newsgroup earlier ;-)
You might be interested in what Kenn is currently doing w.r.t. UML2/CDO.
If you have questions about CDO in the meantime please don't hesitate to
ask them...
Cheers
/Eike
----
http://thegordian.blogspot.com
Patrick Tessier schrieb:
> Hi,
> In the context of papyrus we are very interesting to use CDO in order to
> manage big models. For us managing gi models is a reality :-)
>
>
>
> "Ed Merks" <Ed.Merks@gmail.com> a écrit dans le message de news:
> gu1f9i$odt$1@build.eclipse.org...
>
>> Maurice,
>>
>> I know that Eike and Kenn have been investigating this. I don't think
>> something will be ready for this release though...
>>
>>
>> Maurice wrote:
>>
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so, will it
>>> be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>>>
>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
| |
| Re: UML2 + CDO [message #478389 is a reply to message #478388] |
Tue, 12 May 2009 21:53 |
|
Kenn Hussey schrieb:
> Eike,
>
> There already is a bug - see
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=267222.
Ah ;-)
> The plan to deal with
> the "problem" of extending EModelElement is to implement it as a mixin
> rather than using EModelElementImpl as a base class. If course, we will also
> need to provide a CDO-based implementation of EAnnotation...
>
In preparation for this exercise I tried to come up with a generic
CDOModelElementImpl that implements EModelElement but does not extend
from anything in Ecore (except EObjectImpl through CDOObjectImpl). I
fear it's not so easy because of all these featureIDs in EModelElement
that you can't just mixin to the other hierarchies when it comes to
notifications, etc. I'm curious how you'll manage this challenge.
Cheers
/Eike
----
http://thegordian.blogspot.com
> Kenn
>
> "Eike Stepper" <stepper@esc-net.de> wrote in message
> news:guckkk$rsr$1@build.eclipse.org...
>
>> Maurice,
>>
>> Yes, only today Kenn told me that he is currently working on a UML2
>> version that is ready for CDO. Indeed one of the main challenges is to
>> replace EMF's annotation support by something else. I must admit that
>> I'm right now not informed about how exactly Kenn will solve this issue
>> but we scheduled a discussion for tomorrow. I think we'll use the
>> opportunity to open a bugzilla so that all the interested can follow and
>> try out the patches that we come up with. I'll post a link here...
>>
>> Cheers
>> /Eike
>>
>> ----
>> http://thegordian.blogspot.com
>>
>>
>>
>> Maurice schrieb:
>>
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so, will
>>> it be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>>>
>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #478390 is a reply to message #478389] |
Wed, 13 May 2009 12:15 |
Ronan B
Messages: 273
Registered: July 2009 |
Senior Member |
|
|
Hi guys,
Modifying UML2 for CDO support would be fantastic. I must say I was very
surprised to find that UML2 extended Ecore in such a way as to break CDO
integration. That said I understand the reasons and appreciate the
difficulty in migrating to a more pure solution.
To this end I have written a few ATL transformation to allow UML2 models
to be stored in CDO. This isn't ideal but it works until there is an
official solution. It goes as follows:
1) Transform the UML.ecore to remove EModelElement dependency and
operations that use Ecore types. The new UML.ecore uses a namespace such
as http://www.eclipse.org/uml2/3.0.0/UML_CDO so as to differentiate it
from the offical UML2 metamodel. (I used ATL for this)
2) Do the usual CDO genmodel stuff.
3) Register this new metamodel with CDO. It works fine with CDO 3.5M5
4) Transform UML2 models written using the offical metamodel to the new
CDO complaint metamodel. They will lose their eAnnotations and operations.
(I have used ATL for this)
I have passed this solution onto Maurice and it works nicely for his
needs. I'd be more than happy to send the temporary fix on to anyone else
who might need CDO for UML2 now, would this be useful? It is fine for his
requirements as he uses CDO as purely a store for further transformation
to non UML formats.
Of course there are some issues with this fix:
- UML2 models transformed by (4) above cannot be opened in UML2 editors
unless they are transformed back into UML2 proper. Of course this is
possible but the eAnnotations and operations will be blank.
I'm sure optimisations can be applied to the transformation to make this
approach quite useful.
Regards,
Ronan
Eike Stepper wrote:
> Kenn Hussey schrieb:
>> Eike,
>>
>> There already is a bug - see
>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=267222.
> Ah ;-)
>> The plan to deal with
>> the "problem" of extending EModelElement is to implement it as a mixin
>> rather than using EModelElementImpl as a base class. If course, we will
also
>> need to provide a CDO-based implementation of EAnnotation...
>>
> In preparation for this exercise I tried to come up with a generic
> CDOModelElementImpl that implements EModelElement but does not extend
> from anything in Ecore (except EObjectImpl through CDOObjectImpl). I
> fear it's not so easy because of all these featureIDs in EModelElement
> that you can't just mixin to the other hierarchies when it comes to
> notifications, etc. I'm curious how you'll manage this challenge.
> Cheers
> /Eike
> ----
> http://thegordian.blogspot.com
>> Kenn
>>
>> "Eike Stepper" <stepper@esc-net.de> wrote in message
>> news:guckkk$rsr$1@build.eclipse.org...
>>
>>> Maurice,
>>>
>>> Yes, only today Kenn told me that he is currently working on a UML2
>>> version that is ready for CDO. Indeed one of the main challenges is to
>>> replace EMF's annotation support by something else. I must admit that
>>> I'm right now not informed about how exactly Kenn will solve this issue
>>> but we scheduled a discussion for tomorrow. I think we'll use the
>>> opportunity to open a bugzilla so that all the interested can follow and
>>> try out the patches that we come up with. I'll post a link here...
>>>
>>> Cheers
>>> /Eike
>>>
>>> ----
>>> http://thegordian.blogspot.com
>>>
>>>
>>>
>>> Maurice schrieb:
>>>
>>>> Hi,
>>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>>> there plans to enable UML2 for CDO in the near feature and if so, will
>>>> it be in the next milestone or further down the line?
>>>>
>>>> Thanks!
>>>>
>>>> Maurice
>>>>
>>>>
>>
>>
>>
|
|
|
| Re: UML2 + CDO [message #478391 is a reply to message #478390] |
Wed, 13 May 2009 16:57 |
|
Ronan,
I'm impressed how active and involved you (all) are!!!
Even if I only get this by accident :P
Cheers
/Eike
----
http://thegordian.blogspot.com
Ronan schrieb:
> Hi guys,
> Modifying UML2 for CDO support would be fantastic. I must say I was
> very surprised to find that UML2 extended Ecore in such a way as to
> break CDO integration. That said I understand the reasons and
> appreciate the difficulty in migrating to a more pure solution.
>
> To this end I have written a few ATL transformation to allow UML2
> models to be stored in CDO. This isn't ideal but it works until there
> is an official solution. It goes as follows:
> 1) Transform the UML.ecore to remove EModelElement dependency and
> operations that use Ecore types. The new UML.ecore uses a namespace
> such as http://www.eclipse.org/uml2/3.0.0/UML_CDO so as to
> differentiate it from the offical UML2 metamodel. (I used ATL for this)
> 2) Do the usual CDO genmodel stuff.
> 3) Register this new metamodel with CDO. It works fine with CDO 3.5M5
> 4) Transform UML2 models written using the offical metamodel to the
> new CDO complaint metamodel. They will lose their eAnnotations and
> operations. (I have used ATL for this)
>
> I have passed this solution onto Maurice and it works nicely for his
> needs. I'd be more than happy to send the temporary fix on to anyone
> else who might need CDO for UML2 now, would this be useful? It is fine
> for his requirements as he uses CDO as purely a store for further
> transformation to non UML formats.
>
> Of course there are some issues with this fix:
> - UML2 models transformed by (4) above cannot be opened in UML2
> editors unless they are transformed back into UML2 proper. Of course
> this is possible but the eAnnotations and operations will be blank.
>
> I'm sure optimisations can be applied to the transformation to make
> this approach quite useful.
>
> Regards,
> Ronan
>
> Eike Stepper wrote:
>
>> Kenn Hussey schrieb:
>>> Eike,
>>>
>>> There already is a bug - see
>>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=267222.
>> Ah ;-)
>
>>> The plan to deal with the "problem" of extending EModelElement is to
>>> implement it as a mixin rather than using EModelElementImpl as a
>>> base class. If course, we will
> also
>>> need to provide a CDO-based implementation of EAnnotation...
>>>
>> In preparation for this exercise I tried to come up with a generic
>> CDOModelElementImpl that implements EModelElement but does not extend
>> from anything in Ecore (except EObjectImpl through CDOObjectImpl). I
>> fear it's not so easy because of all these featureIDs in EModelElement
>> that you can't just mixin to the other hierarchies when it comes to
>> notifications, etc. I'm curious how you'll manage this challenge.
>
>> Cheers
>> /Eike
>
>> ----
>> http://thegordian.blogspot.com
>
>
>>> Kenn
>>>
>>> "Eike Stepper" <stepper@esc-net.de> wrote in message
>>> news:guckkk$rsr$1@build.eclipse.org...
>>>
>>>> Maurice,
>>>>
>>>> Yes, only today Kenn told me that he is currently working on a UML2
>>>> version that is ready for CDO. Indeed one of the main challenges is to
>>>> replace EMF's annotation support by something else. I must admit that
>>>> I'm right now not informed about how exactly Kenn will solve this
>>>> issue
>>>> but we scheduled a discussion for tomorrow. I think we'll use the
>>>> opportunity to open a bugzilla so that all the interested can
>>>> follow and
>>>> try out the patches that we come up with. I'll post a link here...
>>>>
>>>> Cheers
>>>> /Eike
>>>>
>>>> ----
>>>> http://thegordian.blogspot.com
>>>>
>>>>
>>>>
>>>> Maurice schrieb:
>>>>
>>>>> Hi,
>>>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>>>> there plans to enable UML2 for CDO in the near feature and if so,
>>>>> will
>>>>> it be in the next milestone or further down the line?
>>>>>
>>>>> Thanks!
>>>>>
>>>>> Maurice
>>>>>
>>>>>
>>>
>>>
>>>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #478402 is a reply to message #478370] |
Sun, 17 May 2009 09:58 |
|
Hi,
I just finished a working version of UML2 for CDO. It's attached to
https://bugs.eclipse.org/267222
You have to checkout org.eclipse.uml2.common in addition. I replaced
EModelElement and EAnnotation by local concepts. This can lead to
problems with existing instance data (maybe Ronan can adjust his
transformations?) or existing client applications that refer to the
respective EMF API.
After re-geneation I fixed all compile problems so that we have 0
errors/warnings now. All of these fixed places must be reviewed very
carefully by a UML2 implementation expert (Kenn?). The attached project
is connected to CVS so that diffs are easy (formatting could be an
issue). Personally I think it will be safer to start from scratch and
have my version as an example at hand...
I was able to create a very simple model with the generic CDO UI and
commit it to a CDO repository. The first commit also stores the
UMLPackage itself which takes quite a long time but the server console
logs out everything nicely. Other clients get updated about changes
without problems.
This is starting to become fun ;-)
Note, that we're still trying to get Ecore itself CDO'ified, which would
enable us to work directly with EAnnotations in UML...
Cheers
/Eike
----
http://thegordian.blogspot.com
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
| |
| Re: UML2 + CDO [message #478409 is a reply to message #478408] |
Mon, 18 May 2009 19:08 |
|
Ronan schrieb:
> Hi Eike,
> Excellent, that is a nice fix. I have attached to your bug report the
> ATL transformation to make UML models compatible with this new CDO
> complaint UML metamodel. The transformation is written for UML2
> version 3.0.0
Nice work, Ronan ;-)
> Can you test it out and report back?
No, I have no clue about ATL.
> I will also send this onto Maurice to test with his CDO setup.
>
> I have used the EModelCopyGenerator [
> http://ssel.vub.ac.be/viewvc/UML2CaseStudies/uml2cs-transfor mations/EModelCopyGenerator.atl?view=markup
> ] ATL script to build the transform so if it works nicely on the UML2
> 3.0.0 metamodel I can provide support for other UML2 versions.
Please keep in mind that our goal is to support the normal EAnnotations.
I suggest to keep your effort for transformations on a minimum ;-)
Cheers
/Eike
----
http://thegordian.blogspot.com
>
> Cheers,
> Ronan
>
> Eike Stepper wrote:
>
>> Hi,
>
>> I just finished a working version of UML2 for CDO. It's attached to
>> https://bugs.eclipse.org/267222
>
>> You have to checkout org.eclipse.uml2.common in addition. I replaced
>> EModelElement and EAnnotation by local concepts. This can lead to
>> problems with existing instance data (maybe Ronan can adjust his
>> transformations?) or existing client applications that refer to the
>> respective EMF API.
>
>> After re-geneation I fixed all compile problems so that we have 0
>> errors/warnings now. All of these fixed places must be reviewed very
>> carefully by a UML2 implementation expert (Kenn?). The attached project
>> is connected to CVS so that diffs are easy (formatting could be an
>> issue). Personally I think it will be safer to start from scratch and
>> have my version as an example at hand...
>
>> I was able to create a very simple model with the generic CDO UI and
>> commit it to a CDO repository. The first commit also stores the
>> UMLPackage itself which takes quite a long time but the server console
>> logs out everything nicely. Other clients get updated about changes
>> without problems.
>
>> This is starting to become fun ;-)
>
>> Note, that we're still trying to get Ecore itself CDO'ified, which would
>> enable us to work directly with EAnnotations in UML...
>
>> Cheers
>> /Eike
>
>> ----
>> http://thegordian.blogspot.com
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #478420 is a reply to message #478409] |
Tue, 19 May 2009 16:59 |
Ronan B
Messages: 273
Registered: July 2009 |
Senior Member |
|
|
Hey Eike,
You should learn ATL it is fantastic for these kind of tasks :) Maurice
tested the ATL script I appended to your bug and it all works nicely out
of the box. I'm still looking forward to the real Ecore/UML2 fix :)
Thanks,
Ronan
Eike Stepper wrote:
> Ronan schrieb:
>> Hi Eike,
>> Excellent, that is a nice fix. I have attached to your bug report the
>> ATL transformation to make UML models compatible with this new CDO
>> complaint UML metamodel. The transformation is written for UML2
>> version 3.0.0
> Nice work, Ronan ;-)
>> Can you test it out and report back?
> No, I have no clue about ATL.
>> I will also send this onto Maurice to test with his CDO setup.
>>
>> I have used the EModelCopyGenerator [
>>
http://ssel.vub.ac.be/viewvc/UML2CaseStudies/uml2cs-transfor mations/EModelCopyGenerator.atl?view=markup
>> ] ATL script to build the transform so if it works nicely on the UML2
>> 3.0.0 metamodel I can provide support for other UML2 versions.
> Please keep in mind that our goal is to support the normal EAnnotations.
> I suggest to keep your effort for transformations on a minimum ;-)
> Cheers
> /Eike
> ----
> http://thegordian.blogspot.com
>>
>> Cheers,
>> Ronan
>>
>> Eike Stepper wrote:
>>
>>> Hi,
>>
>>> I just finished a working version of UML2 for CDO. It's attached to
>>> https://bugs.eclipse.org/267222
>>
>>> You have to checkout org.eclipse.uml2.common in addition. I replaced
>>> EModelElement and EAnnotation by local concepts. This can lead to
>>> problems with existing instance data (maybe Ronan can adjust his
>>> transformations?) or existing client applications that refer to the
>>> respective EMF API.
>>
>>> After re-geneation I fixed all compile problems so that we have 0
>>> errors/warnings now. All of these fixed places must be reviewed very
>>> carefully by a UML2 implementation expert (Kenn?). The attached project
>>> is connected to CVS so that diffs are easy (formatting could be an
>>> issue). Personally I think it will be safer to start from scratch and
>>> have my version as an example at hand...
>>
>>> I was able to create a very simple model with the generic CDO UI and
>>> commit it to a CDO repository. The first commit also stores the
>>> UMLPackage itself which takes quite a long time but the server console
>>> logs out everything nicely. Other clients get updated about changes
>>> without problems.
>>
>>> This is starting to become fun ;-)
>>
>>> Note, that we're still trying to get Ecore itself CDO'ified, which would
>>> enable us to work directly with EAnnotations in UML...
>>
>>> Cheers
>>> /Eike
>>
>>> ----
>>> http://thegordian.blogspot.com
>>
>>
|
|
|
| Re: UML2 + CDO [message #478475 is a reply to message #478378] |
Tue, 02 June 2009 10:41 |
|
Patrick,
Are you committer on Papyrus?
I just talked with Vik, our CDO UI responsible, and he would like to
work with you on a CDO integration ;-)
Cheers
/Eike
----
http://thegordian.blogspot.com
http://twitter.com/eikestepper
Patrick Tessier schrieb:
> Hi,
> In the context of papyrus we are very interesting to use CDO in order to
> manage big models. For us managing gi models is a reality :-)
>
>
>
> "Ed Merks" <Ed.Merks@gmail.com> a écrit dans le message de news:
> gu1f9i$odt$1@build.eclipse.org...
>
>> Maurice,
>>
>> I know that Eike and Kenn have been investigating this. I don't think
>> something will be ready for this release though...
>>
>>
>> Maurice wrote:
>>
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so, will it
>>> be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>>>
>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #627557 is a reply to message #478370] |
Fri, 08 May 2009 14:23 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Maurice,
I know that Eike and Kenn have been investigating this. I don't think
something will be ready for this release though...
Maurice wrote:
> Hi,
> I recently learned that UML2 is currently incompatible with CDO. Are
> there plans to enable UML2 for CDO in the near feature and if so, will
> it be in the next milestone or further down the line?
>
> Thanks!
>
> Maurice
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| | |
| Re: UML2 + CDO [message #627560 is a reply to message #478373] |
Sat, 09 May 2009 14:23 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
This is a multi-part message in MIME format.
--------------080303090705090501010600
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 7bit
Vlad,
Comments below.
Vlad Varnica wrote:
> Hi Maurice,
>
> You have many different ways to generate Jee & Mapping from a model.
I don't know, will you count the ways?
> What is important is to take a clean and standard XMI 2.1 or 2.2 and
> then use one of the commercial or open source tool.
Not sure how that relates to CDO?
>
> I tried the following tools which are really good to generate code
> from an UML Superstructure Model XMI. I didn't spend more time to
> investigate other tools but there could be more good tools/solutions
> on the market:
>
> - BLU AGE: an MDA compliant software workbench that enables 100%
> automated transformation of your UML 2.X diagrams into Java EE and
> .Net Enterprise business applications.
> - AndroMDA is open source http://galaxy.andromda.org/forum/
Maurice didn't ask about generating JEE, so how exactly is this relevant?
>
> CDO is a very nice integration for EMF.
Indeed.
> btw,I don't think it could be extended to generate what you need from
> UML2 because only related to EMF and not to the UML Superstructure Model.
CDO has nothing to do with generation, so I'm confused why you bring all
this up.
> It is important not to make a confusion between EMF and UML2.
I'm totally confused by your whole note...
> EMF is a great framework that we use inside Omondo and UML 2 is the
> XMI serialization of an EMF model but it doesn't mean that EMF is UML2.
Indeed, that's why one they're two different acronyms. Why are *you
*making confusion between them? And why are you confusing CDO with JEE
generation? Maybe the coffee hasn't kicked in yet.
>
> Vlad,
>
--------------080303090705090501010600
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit
<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
<meta content="text/html;charset=ISO-8859-15"
http-equiv="Content-Type">
</head>
<body bgcolor="#ffffff" text="#000000">
Vlad,<br>
<br>
Comments below.<br>
<br>
Vlad Varnica wrote:
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite">Hi Maurice,
<br>
<br>
You have many different ways to generate Jee & Mapping from a model.</blockquote>
I don't know, will you count the ways?<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> What is important is to take a clean and standard XMI 2.1
or 2.2 and then use one of the commercial or open source tool.
<br>
</blockquote>
Not sure how that relates to CDO?<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"><br>
I tried the following tools which are really good to generate code from
an UML Superstructure Model XMI. I didn't spend more time to
investigate other tools but there could be more good tools/solutions on
the market:
<br>
<br>
- BLU AGE: an MDA compliant software workbench that enables 100%
automated transformation of your UML 2.X diagrams into Java EE and .Net
Enterprise business applications.
<br>
- AndroMDA is open source <a class="moz-txt-link-freetext" href="http://galaxy.andromda.org/forum/">http://galaxy.andromda.org/forum/</a>
<br>
</blockquote>
Maurice didn't ask about generating JEE, so how exactly is this
relevant?<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"><br>
CDO is a very nice integration for EMF.</blockquote>
Indeed.<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> btw,I don't think it could be extended to generate what
you need from UML2 because only related to EMF and not to the UML
Superstructure Model.</blockquote>
CDO has nothing to do with generation, so I'm confused why you bring
all this up.<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> It is important not to make a confusion between EMF and
UML2.</blockquote>
I'm totally confused by your whole note...<br>
<blockquote
cite="mid:5f3b98f84a2d8b6a92bdbe7128343764$1@www.eclipse.org"
type="cite"> EMF is a great framework that we use inside Omondo and
UML 2 is the XMI serialization of an EMF model but it doesn't mean that
EMF is UML2.
<br>
</blockquote>
Indeed, that's why one they're two different acronyms.
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| |
| Re: UML2 + CDO [message #627562 is a reply to message #478375] |
Sun, 10 May 2009 14:23 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Vlad,
Comments below.
Vlad Varnica wrote:
> Ed,
>
> I should take more coffee and stop smoking :-)
What have you been smoking? :-P
> CDO is a 3-tiers solution for distributed shared models that allow you
> to develop and use different mappers such as Hibernate.
> I therefore consider that a demand related to UML+CDO is dealing with
> persistence tiers generation from a model.
No, Maurice did not ask about generation. Maurice apparently wants UML
itself to be persisted using CDO. He's said nothing about what he's
doing with his UML instances other than wanting to persist them using CDO.
> It seems that Maurice is interested in UML and not EMF ?
Given his interest in CDO and the fact that it's a component of EMF, I'd
say you are mistaken.
> Sorry if I am wrong but this seems to me logical.
I'm quite sure you are wrong. He might well want to generate JEE, but
there's no indication of that.
>
> This is a very interesting project but Omondo never used it because it
> is still in the early stage and for our customers we prefer to
> recommend a more stable solution. Our recommendation for a successful
> project are:
> - use UML 2 with Blue Age, AndroMDA or MIA software for persistence
> tiers generation from an UML Superstructure Model - Use Agile
> incremental Hibernate, JPA annotations with EclipseUML 2008 and Oracle
> 11G. See more at:
> http://www.forum-omondo.com/documentation_eclipseuml_2008/Ec lipse_Database/Activate_JPA_Perspective/oracle_integration.h tml
>
Is it really necessary to use this newsgroup in this way?
>
> We will certainly consider CDO again with Eclipse 3.5.
I doubt it, because while you use EMF, you don't actually want any of
your customers using EMF.
>
> I think it is important to help Eclipse users to know what is ready
> for production and what is not because this is a waste of time at the
> current stage.
What is a waste of time? I'm starting to think your comments are a
waste of time and space.
> Open source is useful only if it is not more time consuming than the
> benefit it really gives.
Climbing onto your soapbox again are you?
> You need few months just to understand EMF and few others months to
> understand CDO.
Time well spent.
> Do you really think companies have time to do that ?
Apparently a great many people do indeed have time, make time, and find
it time well spent. How much time does it take to understand UML?
What about the UML API itself? What about learning how your products
work? And then, when your company goes belly up, how much time does it
take all your customers to find an alternative?
> I think it is more important to give this kind of answer than just
> support first level which makes users feel that they understand while
> the reality is different !!
I think your commentary is so confusing and so off the mark that it does
a disservice.
>
>
> Vlad,
>
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
| | |
| Re: UML2 + CDO [message #627568 is a reply to message #478371] |
Mon, 11 May 2009 14:27 |
Maurice O'Connor
Messages: 86
Registered: July 2009 |
Member |
|
|
Hi Ed,
Thanks for your reply. I see the UML2 metamodel is extending from, for
example, EModelElement from the ECore metamodel. Is this an example of
what Eike meant when he talked about ECore extensions in UML2, in this
post?
http://www.eclipse.org/newsportal/article.php?id=41535&g roup=eclipse.tools.emf#41535
When I removed these extensions from UML2's metamodel it was then capable
of generating a CDO-enabled genmodel which would actually work. Can you
enlighten me as to why UML2 extends ECore in this fashion? Is this the
reason why CDO can't hook into these metamodels?
As a short-term fix, to get my UML2 model ready to persist in a CDO
environment, would it be wise to strip these extensions? We're only
interested in UML2 class diagrams and only looking to persist instances of
these.
Maurice
Ed Merks wrote:
> Maurice,
> I know that Eike and Kenn have been investigating this. I don't think
> something will be ready for this release though...
> Maurice wrote:
>> Hi,
>> I recently learned that UML2 is currently incompatible with CDO. Are
>> there plans to enable UML2 for CDO in the near feature and if so, will
>> it be in the next milestone or further down the line?
>>
>> Thanks!
>>
>> Maurice
>>
|
|
|
| Re: UML2 + CDO [message #627569 is a reply to message #478381] |
Mon, 11 May 2009 14:40 |
Ed Merks
Messages: 33133
Registered: July 2009 |
Senior Member |
|
|
Maurice,
Yes, this is indeed the problem. UML extends EModelElement simply to
reuse Ecore's EAnnotation support. It was an unfortunate design
decision but given that it's part of the API, it's not easily changed.
Maurice wrote:
> Hi Ed,
> Thanks for your reply. I see the UML2 metamodel is extending from, for
> example, EModelElement from the ECore metamodel. Is this an example of
> what Eike meant when he talked about ECore extensions in UML2, in this
> post?
>
> http://www.eclipse.org/newsportal/article.php?id=41535&g roup=eclipse.tools.emf#41535
>
>
> When I removed these extensions from UML2's metamodel it was then
> capable of generating a CDO-enabled genmodel which would actually
> work. Can you enlighten me as to why UML2 extends ECore in this
> fashion? Is this the reason why CDO can't hook into these metamodels?
>
> As a short-term fix, to get my UML2 model ready to persist in a CDO
> environment, would it be wise to strip these extensions? We're only
> interested in UML2 class diagrams and only looking to persist
> instances of these.
>
> Maurice
>
> Ed Merks wrote:
>
>> Maurice,
>
>> I know that Eike and Kenn have been investigating this. I don't
>> think something will be ready for this release though...
>
>
>> Maurice wrote:
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so,
>>> will it be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>
>
Ed Merks
Professional Support: https://www.macromodeling.com/
|
|
|
| Re: UML2 + CDO [message #627572 is a reply to message #478370] |
Tue, 12 May 2009 20:02 |
|
Maurice,
Yes, only today Kenn told me that he is currently working on a UML2
version that is ready for CDO. Indeed one of the main challenges is to
replace EMF's annotation support by something else. I must admit that
I'm right now not informed about how exactly Kenn will solve this issue
but we scheduled a discussion for tomorrow. I think we'll use the
opportunity to open a bugzilla so that all the interested can follow and
try out the patches that we come up with. I'll post a link here...
Cheers
/Eike
----
http://thegordian.blogspot.com
Maurice schrieb:
> Hi,
> I recently learned that UML2 is currently incompatible with CDO. Are
> there plans to enable UML2 for CDO in the near feature and if so, will
> it be in the next milestone or further down the line?
>
> Thanks!
>
> Maurice
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #627573 is a reply to message #478375] |
Tue, 12 May 2009 20:23 |
|
Vlad,
Comments below...
Vlad Varnica schrieb:
> Ed,
>
> I should take more coffee and stop smoking :-)
> CDO is a 3-tiers solution for distributed shared models that allow you
> to develop and use different mappers such as Hibernate.
So far so good.
> I therefore consider that a demand related to UML+CDO is dealing with
> persistence tiers generation from a model. It seems that Maurice is
> interested in UML and not EMF ? Sorry if I am wrong but this seems to
> me logical.
>
> This is a very interesting project but Omondo never used it because it
> is still in the early stage and for our customers we prefer to
> recommend a more stable solution.
What measures are you basing your statements about stability on?
> Our recommendation for a successful project are:
> - use UML 2 with Blue Age, AndroMDA or MIA software for persistence
> tiers generation from an UML Superstructure Model - Use Agile
> incremental Hibernate, JPA annotations with EclipseUML 2008 and Oracle
> 11G. See more at:
> http://www.forum-omondo.com/documentation_eclipseuml_2008/Ec lipse_Database/Activate_JPA_Perspective/oracle_integration.h tml
>
I don't know all of these products but are you really arguing that this
is *always* the best combination for *every* "successful project"?
>
> We will certainly consider CDO again with Eclipse 3.5.
Although it's not commercial?
>
> I think it is important to help Eclipse users to know what is ready
> for production and what is not
Good point! Are you volunteering to come up with more reliable help in
this regard? I guess those users would want to know where your expertise
is coming from.
> because this is a waste of time at the current stage.
Without proof this is defamation at best. It sounds like an abuse of
everybody who spent a lot of effort for making CDO what it is now. And
an abuse of all the users that are using CDO in production for many
years now, the NASA and the Canadian DoD for example, but many large
enterprises as well.
> Open source is useful only if it is not more time consuming than the
> benefit it really gives.
This is certainly not only valid for open source.
> You need few months just to understand EMF and few others months to
> understand CDO.
You should not deduce from your own experience.
Btw. shouldn't I remember at least a single newsgroup question of you
while you were spending months understanding CDO?
Could this be the reason why nobody else complained about so long
learning periods? They used the newsgroup to *ask questions* !
> Do you really think companies have time to do that ? I think it is
> more important to give this kind of answer than just support first
> level which makes users feel that they understand while the reality is
> different !!
I'm not sure I understand this one. Are you suggesting that we
deliberately give wrong advice here or do you think the users are just
too stupid to realize that we don't see the big picture?
Cheers
/Eike
----
http://thegordian.blogspot.com
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #627574 is a reply to message #478378] |
Tue, 12 May 2009 20:28 |
|
Patrick,
I appreciate your interest and I realize that I should have subscribed
to this newsgroup earlier ;-)
You might be interested in what Kenn is currently doing w.r.t. UML2/CDO.
If you have questions about CDO in the meantime please don't hesitate to
ask them...
Cheers
/Eike
----
http://thegordian.blogspot.com
Patrick Tessier schrieb:
> Hi,
> In the context of papyrus we are very interesting to use CDO in order to
> manage big models. For us managing gi models is a reality :-)
>
>
>
> "Ed Merks" <Ed.Merks@gmail.com> a écrit dans le message de news:
> gu1f9i$odt$1@build.eclipse.org...
>
>> Maurice,
>>
>> I know that Eike and Kenn have been investigating this. I don't think
>> something will be ready for this release though...
>>
>>
>> Maurice wrote:
>>
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so, will it
>>> be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>>>
>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
| |
| Re: UML2 + CDO [message #627576 is a reply to message #478388] |
Tue, 12 May 2009 21:53 |
|
Kenn Hussey schrieb:
> Eike,
>
> There already is a bug - see
> https://bugs.eclipse.org/bugs/show_bug.cgi?id=267222
Ah ;-)
> The plan to deal with
> the "problem" of extending EModelElement is to implement it as a mixin
> rather than using EModelElementImpl as a base class. If course, we will also
> need to provide a CDO-based implementation of EAnnotation...
>
In preparation for this exercise I tried to come up with a generic
CDOModelElementImpl that implements EModelElement but does not extend
from anything in Ecore (except EObjectImpl through CDOObjectImpl). I
fear it's not so easy because of all these featureIDs in EModelElement
that you can't just mixin to the other hierarchies when it comes to
notifications, etc. I'm curious how you'll manage this challenge.
Cheers
/Eike
----
http://thegordian.blogspot.com
> Kenn
>
> "Eike Stepper" <stepper@esc-net.de> wrote in message
> news:guckkk$rsr$1@build.eclipse.org...
>
>> Maurice,
>>
>> Yes, only today Kenn told me that he is currently working on a UML2
>> version that is ready for CDO. Indeed one of the main challenges is to
>> replace EMF's annotation support by something else. I must admit that
>> I'm right now not informed about how exactly Kenn will solve this issue
>> but we scheduled a discussion for tomorrow. I think we'll use the
>> opportunity to open a bugzilla so that all the interested can follow and
>> try out the patches that we come up with. I'll post a link here...
>>
>> Cheers
>> /Eike
>>
>> ----
>> http://thegordian.blogspot.com
>>
>>
>>
>> Maurice schrieb:
>>
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so, will
>>> it be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>>>
>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #627577 is a reply to message #478389] |
Wed, 13 May 2009 12:15 |
Ronan B
Messages: 273
Registered: July 2009 |
Senior Member |
|
|
Hi guys,
Modifying UML2 for CDO support would be fantastic. I must say I was very
surprised to find that UML2 extended Ecore in such a way as to break CDO
integration. That said I understand the reasons and appreciate the
difficulty in migrating to a more pure solution.
To this end I have written a few ATL transformation to allow UML2 models
to be stored in CDO. This isn't ideal but it works until there is an
official solution. It goes as follows:
1) Transform the UML.ecore to remove EModelElement dependency and
operations that use Ecore types. The new UML.ecore uses a namespace such
as http://www.eclipse.org/uml2/3.0.0/UML_CDO so as to differentiate it
from the offical UML2 metamodel. (I used ATL for this)
2) Do the usual CDO genmodel stuff.
3) Register this new metamodel with CDO. It works fine with CDO 3.5M5
4) Transform UML2 models written using the offical metamodel to the new
CDO complaint metamodel. They will lose their eAnnotations and operations.
(I have used ATL for this)
I have passed this solution onto Maurice and it works nicely for his
needs. I'd be more than happy to send the temporary fix on to anyone else
who might need CDO for UML2 now, would this be useful? It is fine for his
requirements as he uses CDO as purely a store for further transformation
to non UML formats.
Of course there are some issues with this fix:
- UML2 models transformed by (4) above cannot be opened in UML2 editors
unless they are transformed back into UML2 proper. Of course this is
possible but the eAnnotations and operations will be blank.
I'm sure optimisations can be applied to the transformation to make this
approach quite useful.
Regards,
Ronan
Eike Stepper wrote:
> Kenn Hussey schrieb:
>> Eike,
>>
>> There already is a bug - see
>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=267222
> Ah ;-)
>> The plan to deal with
>> the "problem" of extending EModelElement is to implement it as a mixin
>> rather than using EModelElementImpl as a base class. If course, we will
also
>> need to provide a CDO-based implementation of EAnnotation...
>>
> In preparation for this exercise I tried to come up with a generic
> CDOModelElementImpl that implements EModelElement but does not extend
> from anything in Ecore (except EObjectImpl through CDOObjectImpl). I
> fear it's not so easy because of all these featureIDs in EModelElement
> that you can't just mixin to the other hierarchies when it comes to
> notifications, etc. I'm curious how you'll manage this challenge.
> Cheers
> /Eike
> ----
> http://thegordian.blogspot.com
>> Kenn
>>
>> "Eike Stepper" <stepper@esc-net.de> wrote in message
>> news:guckkk$rsr$1@build.eclipse.org...
>>
>>> Maurice,
>>>
>>> Yes, only today Kenn told me that he is currently working on a UML2
>>> version that is ready for CDO. Indeed one of the main challenges is to
>>> replace EMF's annotation support by something else. I must admit that
>>> I'm right now not informed about how exactly Kenn will solve this issue
>>> but we scheduled a discussion for tomorrow. I think we'll use the
>>> opportunity to open a bugzilla so that all the interested can follow and
>>> try out the patches that we come up with. I'll post a link here...
>>>
>>> Cheers
>>> /Eike
>>>
>>> ----
>>> http://thegordian.blogspot.com
>>>
>>>
>>>
>>> Maurice schrieb:
>>>
>>>> Hi,
>>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>>> there plans to enable UML2 for CDO in the near feature and if so, will
>>>> it be in the next milestone or further down the line?
>>>>
>>>> Thanks!
>>>>
>>>> Maurice
>>>>
>>>>
>>
>>
>>
|
|
|
| Re: UML2 + CDO [message #627578 is a reply to message #478390] |
Wed, 13 May 2009 16:57 |
|
Ronan,
I'm impressed how active and involved you (all) are!!!
Even if I only get this by accident :P
Cheers
/Eike
----
http://thegordian.blogspot.com
Ronan schrieb:
> Hi guys,
> Modifying UML2 for CDO support would be fantastic. I must say I was
> very surprised to find that UML2 extended Ecore in such a way as to
> break CDO integration. That said I understand the reasons and
> appreciate the difficulty in migrating to a more pure solution.
>
> To this end I have written a few ATL transformation to allow UML2
> models to be stored in CDO. This isn't ideal but it works until there
> is an official solution. It goes as follows:
> 1) Transform the UML.ecore to remove EModelElement dependency and
> operations that use Ecore types. The new UML.ecore uses a namespace
> such as http://www.eclipse.org/uml2/3.0.0/UML_CDO so as to
> differentiate it from the offical UML2 metamodel. (I used ATL for this)
> 2) Do the usual CDO genmodel stuff.
> 3) Register this new metamodel with CDO. It works fine with CDO 3.5M5
> 4) Transform UML2 models written using the offical metamodel to the
> new CDO complaint metamodel. They will lose their eAnnotations and
> operations. (I have used ATL for this)
>
> I have passed this solution onto Maurice and it works nicely for his
> needs. I'd be more than happy to send the temporary fix on to anyone
> else who might need CDO for UML2 now, would this be useful? It is fine
> for his requirements as he uses CDO as purely a store for further
> transformation to non UML formats.
>
> Of course there are some issues with this fix:
> - UML2 models transformed by (4) above cannot be opened in UML2
> editors unless they are transformed back into UML2 proper. Of course
> this is possible but the eAnnotations and operations will be blank.
>
> I'm sure optimisations can be applied to the transformation to make
> this approach quite useful.
>
> Regards,
> Ronan
>
> Eike Stepper wrote:
>
>> Kenn Hussey schrieb:
>>> Eike,
>>>
>>> There already is a bug - see
>>> https://bugs.eclipse.org/bugs/show_bug.cgi?id=267222
>> Ah ;-)
>
>>> The plan to deal with the "problem" of extending EModelElement is to
>>> implement it as a mixin rather than using EModelElementImpl as a
>>> base class. If course, we will
> also
>>> need to provide a CDO-based implementation of EAnnotation...
>>>
>> In preparation for this exercise I tried to come up with a generic
>> CDOModelElementImpl that implements EModelElement but does not extend
>> from anything in Ecore (except EObjectImpl through CDOObjectImpl). I
>> fear it's not so easy because of all these featureIDs in EModelElement
>> that you can't just mixin to the other hierarchies when it comes to
>> notifications, etc. I'm curious how you'll manage this challenge.
>
>> Cheers
>> /Eike
>
>> ----
>> http://thegordian.blogspot.com
>
>
>>> Kenn
>>>
>>> "Eike Stepper" <stepper@esc-net.de> wrote in message
>>> news:guckkk$rsr$1@build.eclipse.org...
>>>
>>>> Maurice,
>>>>
>>>> Yes, only today Kenn told me that he is currently working on a UML2
>>>> version that is ready for CDO. Indeed one of the main challenges is to
>>>> replace EMF's annotation support by something else. I must admit that
>>>> I'm right now not informed about how exactly Kenn will solve this
>>>> issue
>>>> but we scheduled a discussion for tomorrow. I think we'll use the
>>>> opportunity to open a bugzilla so that all the interested can
>>>> follow and
>>>> try out the patches that we come up with. I'll post a link here...
>>>>
>>>> Cheers
>>>> /Eike
>>>>
>>>> ----
>>>> http://thegordian.blogspot.com
>>>>
>>>>
>>>>
>>>> Maurice schrieb:
>>>>
>>>>> Hi,
>>>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>>>> there plans to enable UML2 for CDO in the near feature and if so,
>>>>> will
>>>>> it be in the next milestone or further down the line?
>>>>>
>>>>> Thanks!
>>>>>
>>>>> Maurice
>>>>>
>>>>>
>>>
>>>
>>>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #627589 is a reply to message #478370] |
Sun, 17 May 2009 09:58 |
|
Hi,
I just finished a working version of UML2 for CDO. It's attached to
https://bugs.eclipse.org/267222
You have to checkout org.eclipse.uml2.common in addition. I replaced
EModelElement and EAnnotation by local concepts. This can lead to
problems with existing instance data (maybe Ronan can adjust his
transformations?) or existing client applications that refer to the
respective EMF API.
After re-geneation I fixed all compile problems so that we have 0
errors/warnings now. All of these fixed places must be reviewed very
carefully by a UML2 implementation expert (Kenn?). The attached project
is connected to CVS so that diffs are easy (formatting could be an
issue). Personally I think it will be safer to start from scratch and
have my version as an example at hand...
I was able to create a very simple model with the generic CDO UI and
commit it to a CDO repository. The first commit also stores the
UMLPackage itself which takes quite a long time but the server console
logs out everything nicely. Other clients get updated about changes
without problems.
This is starting to become fun ;-)
Note, that we're still trying to get Ecore itself CDO'ified, which would
enable us to work directly with EAnnotations in UML...
Cheers
/Eike
----
http://thegordian.blogspot.com
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
| |
| Re: UML2 + CDO [message #627594 is a reply to message #478408] |
Mon, 18 May 2009 19:08 |
|
Ronan schrieb:
> Hi Eike,
> Excellent, that is a nice fix. I have attached to your bug report the
> ATL transformation to make UML models compatible with this new CDO
> complaint UML metamodel. The transformation is written for UML2
> version 3.0.0
Nice work, Ronan ;-)
> Can you test it out and report back?
No, I have no clue about ATL.
> I will also send this onto Maurice to test with his CDO setup.
>
> I have used the EModelCopyGenerator [
> http://ssel.vub.ac.be/viewvc/UML2CaseStudies/uml2cs-transfor mations/EModelCopyGenerator.atl?view=markup
> ] ATL script to build the transform so if it works nicely on the UML2
> 3.0.0 metamodel I can provide support for other UML2 versions.
Please keep in mind that our goal is to support the normal EAnnotations.
I suggest to keep your effort for transformations on a minimum ;-)
Cheers
/Eike
----
http://thegordian.blogspot.com
>
> Cheers,
> Ronan
>
> Eike Stepper wrote:
>
>> Hi,
>
>> I just finished a working version of UML2 for CDO. It's attached to
>> https://bugs.eclipse.org/267222
>
>> You have to checkout org.eclipse.uml2.common in addition. I replaced
>> EModelElement and EAnnotation by local concepts. This can lead to
>> problems with existing instance data (maybe Ronan can adjust his
>> transformations?) or existing client applications that refer to the
>> respective EMF API.
>
>> After re-geneation I fixed all compile problems so that we have 0
>> errors/warnings now. All of these fixed places must be reviewed very
>> carefully by a UML2 implementation expert (Kenn?). The attached project
>> is connected to CVS so that diffs are easy (formatting could be an
>> issue). Personally I think it will be safer to start from scratch and
>> have my version as an example at hand...
>
>> I was able to create a very simple model with the generic CDO UI and
>> commit it to a CDO repository. The first commit also stores the
>> UMLPackage itself which takes quite a long time but the server console
>> logs out everything nicely. Other clients get updated about changes
>> without problems.
>
>> This is starting to become fun ;-)
>
>> Note, that we're still trying to get Ecore itself CDO'ified, which would
>> enable us to work directly with EAnnotations in UML...
>
>> Cheers
>> /Eike
>
>> ----
>> http://thegordian.blogspot.com
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
| Re: UML2 + CDO [message #627605 is a reply to message #478409] |
Tue, 19 May 2009 16:59 |
Ronan B
Messages: 273
Registered: July 2009 |
Senior Member |
|
|
Hey Eike,
You should learn ATL it is fantastic for these kind of tasks :) Maurice
tested the ATL script I appended to your bug and it all works nicely out
of the box. I'm still looking forward to the real Ecore/UML2 fix :)
Thanks,
Ronan
Eike Stepper wrote:
> Ronan schrieb:
>> Hi Eike,
>> Excellent, that is a nice fix. I have attached to your bug report the
>> ATL transformation to make UML models compatible with this new CDO
>> complaint UML metamodel. The transformation is written for UML2
>> version 3.0.0
> Nice work, Ronan ;-)
>> Can you test it out and report back?
> No, I have no clue about ATL.
>> I will also send this onto Maurice to test with his CDO setup.
>>
>> I have used the EModelCopyGenerator [
>>
http://ssel.vub.ac.be/viewvc/UML2CaseStudies/uml2cs-transfor mations/EModelCopyGenerator.atl?view=markup
>> ] ATL script to build the transform so if it works nicely on the UML2
>> 3.0.0 metamodel I can provide support for other UML2 versions.
> Please keep in mind that our goal is to support the normal EAnnotations.
> I suggest to keep your effort for transformations on a minimum ;-)
> Cheers
> /Eike
> ----
> http://thegordian.blogspot.com
>>
>> Cheers,
>> Ronan
>>
>> Eike Stepper wrote:
>>
>>> Hi,
>>
>>> I just finished a working version of UML2 for CDO. It's attached to
>>> https://bugs.eclipse.org/267222
>>
>>> You have to checkout org.eclipse.uml2.common in addition. I replaced
>>> EModelElement and EAnnotation by local concepts. This can lead to
>>> problems with existing instance data (maybe Ronan can adjust his
>>> transformations?) or existing client applications that refer to the
>>> respective EMF API.
>>
>>> After re-geneation I fixed all compile problems so that we have 0
>>> errors/warnings now. All of these fixed places must be reviewed very
>>> carefully by a UML2 implementation expert (Kenn?). The attached project
>>> is connected to CVS so that diffs are easy (formatting could be an
>>> issue). Personally I think it will be safer to start from scratch and
>>> have my version as an example at hand...
>>
>>> I was able to create a very simple model with the generic CDO UI and
>>> commit it to a CDO repository. The first commit also stores the
>>> UMLPackage itself which takes quite a long time but the server console
>>> logs out everything nicely. Other clients get updated about changes
>>> without problems.
>>
>>> This is starting to become fun ;-)
>>
>>> Note, that we're still trying to get Ecore itself CDO'ified, which would
>>> enable us to work directly with EAnnotations in UML...
>>
>>> Cheers
>>> /Eike
>>
>>> ----
>>> http://thegordian.blogspot.com
>>
>>
|
|
|
| Re: UML2 + CDO [message #627653 is a reply to message #478378] |
Tue, 02 June 2009 10:41 |
|
Patrick,
Are you committer on Papyrus?
I just talked with Vik, our CDO UI responsible, and he would like to
work with you on a CDO integration ;-)
Cheers
/Eike
----
http://thegordian.blogspot.com
http://twitter.com/eikestepper
Patrick Tessier schrieb:
> Hi,
> In the context of papyrus we are very interesting to use CDO in order to
> manage big models. For us managing gi models is a reality :-)
>
>
>
> "Ed Merks" <Ed.Merks@gmail.com> a écrit dans le message de news:
> gu1f9i$odt$1@build.eclipse.org...
>
>> Maurice,
>>
>> I know that Eike and Kenn have been investigating this. I don't think
>> something will be ready for this release though...
>>
>>
>> Maurice wrote:
>>
>>> Hi,
>>> I recently learned that UML2 is currently incompatible with CDO. Are
>>> there plans to enable UML2 for CDO in the near feature and if so, will it
>>> be in the next milestone or further down the line?
>>>
>>> Thanks!
>>>
>>> Maurice
>>>
>>>
>
>
>
Cheers
/Eike
----
http://www.esc-net.de
http://thegordian.blogspot.com
http://twitter.com/eikestepper
|
|
|
Goto Forum:
Current Time: Tue Apr 16 21:21:56 GMT 2024
Powered by FUDForum. Page generated in 2.44906 seconds |
] 
Search
Help
Register
Login
Home
