Automating Model Migration with EMF Edapt

Session Type: 
Standard [25 minutes]
Presenter: 
Markus Herrmannsdoerfer

Like other software artifacts, Ecore models are subject to evolution. When an Ecore model evolves, existing instances may need to be migrated. Manual migration of these instances is tedious and error-prone, and thus instance migration needs to be automated. However, building an automated migrator is a non-trivial task, as it needs to preserve the meaning of possibly infinite number of instances.

To ease the development of an automated migrator, EMF Edapt records the instance migration together with the Ecore model evolution in an explicit history model. The history model specifies the sequence of coupled operations that have been performed. A coupled operation encapsulates both Ecore model evolution and instance migration. Instance migration can be further automated by reusing recurring coupled operations. In order not to disturb EMF users in their habits, EMF Edapt seamlessly integrates into the Ecore model editor.

EMF Edapt results from the research project COPE whose code has been successfully transferred to Eclipse. After the transfer, a lot of significant improvements have been made, e.g. Groovy has been replaced by Java as language to specify custom model migrations. This talk introduces the concepts underlying EMF Edapt and gives a hands-on demonstration of EMF Edapt’s features.

Comments

Hi Markus, Cope/EMF Edapt

Submitted by mariot.chauvin@... on Thu, 2011-09-08 11:08.

Hi Markus,

Cope/EMF Edapt have already been presented to the following eclipse conventions (at least) :

  • Eclipse Summit Europe 2008, during modeling symposium
  • Eclipse Summit Europe 2009, short talk
  • EclipseCon 2010, during painless metamodel evolution talk
  • Eclipse Summit Europe 2010, during modeling symposium
  • EclipseCon 2011, during EMF mega tutorial

And I did not count the posters.
The proposal was created I think in mars 2009, and since :

  • There have not been any official release, and the project plan is empty.
  • There have been only 2 questions on the newsgroup (I did not count the one asking for the project status)

See http://www.eclipse.org/projects/project_summary.php?projectid=modeling.e... for more details.

Further more In your abstract you did not give us any details on the improvements you made.

So based on this facts I think you will agree that Edapt has been presented several times to Eclipse events, and that its audience could be quite limited for this talk.

We had to make a choice, and as Christian has written, there was a lot of modeling submissions.

We hope to see a presentation of those Edapt improvements in a BoF modeling symposium.

Hi Christian, I do not

Submitted by herrmama@in.tum.de on Thu, 2011-09-08 09:12.

Hi Christian,

I do not completely agree with your fear. COPE/EMF Edapt was always only presented as part of the Eclipse Modeling Symposium, and only once got accepted as a short talk. For the short talk, I remember that the room was packed with people, some people even standing. And for the transition of COPE to EMF Edapt, a lot of changes have been performed in which people might be interested. Now that COPE is available through the EMF Edapt project, I get more and more requests of people interested in it. Of course, I will accept your decision, but maybe you want to rethink it due to this information.
Cheers,

Markus

We did not have enough

Submitted by christian.campo... on Thu, 2011-09-08 08:07.

We did not have enough confidence that EMF Edapt topic would attract a lot of people. Its around for a number of years, has a limited audience, and has been presented several times to Eclipse conventions. We recommend to go for a BoF at EclipseCon if you like to interest more people. Blog articles on Planet eclipse are also a good way to extend your community.