An Eclipse Project Set file (.psf) enables quick export and import of files from a repository like Git, SVN etc. Eclipse currently supports exporting files in a repository as a Project Set file. An editor for an existing PSF file is currently missing. The means updating an existing PSF file (for example, adding a new project, removing an existing project etc.) means editing PSF XML by hand!
The Project Set Editor provides a simple user interface to view and edit PSF files (similar to Manifest Editor or Target Definition Editor). The editor also allows directly importing the artifacts in a PSF file.
You can find the project at Eclipse Labs: http://code.google.com/a/eclipselabs.org/p/psfeditor/
Doors are open for testers and committers.
ReqIF is the new OMG standard for requirements interchange. RIF is the predecessor to it and already supported by many tools like DOORS. RMF delivers both the standards as Ecore models. However, there are quite a few differences between the two standards making them type incompatible.
The main differences lies in,
1. Element naming.
For example, the element RIF is now called REQIF, SpecHierarchyRoot is now Specification and so on.
2. More specification types.
If you know the RIF model already, you would know the concept of SpecTypes. This element allows creation of new types. In RIF and ReqIF, SpecObjects (the instance values) and SpecRelations (relations between instances) are typed. However, in RIF, both refer to the same SpecType element, where as ReqIF has different types called SpecObjectType for SpecObject, SpecRelationType for SpecRelation and some more new types.
3. More attribute definition and value types.
RIF had only a single attribute definition type called AttributeDefinitionSimple. ReqIF extends this for different the different data types as AttributeDefinitionInteger, AttributeDefinitionReal, AttributeDefinitionString etc. Same for values as AttributeValueInteger, AttributeValueReal, AttributeValueString etc.
4. RIF XML uses no XML attributes where as ReqIF uses a lot of them.
For example, IDENTIFIER is an XML element in RIF whereas it is a XML attribute in ReqIF.
This brings us to the question – how do I convert my existing RIF files to the new ReqIF format? Considering the fact that there are a lot of industry samples in RIF format, we thought of writing a RIF to ReqIF converter. The transformer takes in a RIF file and transforms it into a ReqIF, retaining all its contents. You could also do an in-memory transformation between the two models.
This is how a RIF source than we passed to our converter looks like after transformation.
Preparing to make the initial code contribution for RMF, we ran our RIF/ReqIF metamodels through several performance tests. To start with, we tested the load and save times of RIF files based on some industry samples. To get some comparison data, we generated XMI files using the same data held in RIF XML files and tested the load/save time against it. The results are quite promising.
Before we go into the details of the tests, its better to define two components involved in our test comparisons.
- The customized RIF XML loader (a.k.a RMF Loader) and serializer (a.k.a RMF Serializer) for loading/saving OMG RIF XML files into RIF Ecore metamodel (read more on the metamodel implementation here).
- The default RIF XMI loader and serializer for loading/saving RIF XMI files into RIF Ecore model (this is not in the scope of RMF. We use this only to get some comparison).
Here are some highlights from our tests.
- A 32MB RIF XML file is loaded in 14.4 seconds by RMF loader where as the same data in XMI format is loaded by the default EMF XMI loader in 22.2 seconds (and 70 mins(!!) without any optimizations to the XMI loader)
- The average time taken to load per MB of data from RIF XML is 0.5 seconds, whereas RIF XMI takes 1.63 seconds per MB. For save, average time taken per MB of data to RIF XML is 0.09 seconds, whereas RIF XMI takes 1.22 seconds per MB
- The load and save time for RIF XML files by RMF loader/serializer increases linearly with size
RIF/ReqIF is the new OMG standard for requirements interchange. RMF (currently in proposal phase) provides an EMF Ecore based metamodel implementation for the RIF/ReqIF XML format. The metamodel is a clean implementation of the format without any “XML noise”. The Ecore metamodel also conforms to the CMOF metamodel delivered by OMG (as it has been derived from it). The metamodel reads/writes RIF/ReqIF data in conformance with the RIF/ReqIF XML Schema.
The challenge in the Ecore based RIF/ReqIF implementation was the customization of loaders and serializers to make them RIF/ReqIF XML Schema conformant. EMF provides different ways of customizing the XML output.
1. Using ExtendedMetadata annotations.
Both approaches are quite tricky to implement when the expected XML output has structural differences compared to the Ecore metamodel. For example, the XML output has wrapper elements for lists (ELists) in the Ecore metamodel.
We went in for a third approach.
We imported RIF/ReqIF XML schema using EMF importer to create a RIF/ReqIF XML Ecore model (Anyone who has done this, knows it’s an “ugly” metamodel that EMF generates here). The next step was to create the real RIF/ReqIF metamodel, by importing RIF/ReqIF CMOF file. We then wrote a generic (of course reusable) Ecore XML to Ecore converter to do a model to model transformation in both directions. The whole processing is cleanly hidden in a new EMF Resource implementation such that the user hardly knows anything about these.
The following are the advantages we saw in the approach.
1. You don’t do customizations at a XML level, but at a higher EMF model API level. That is, you don’t hack some XML SAX events, but work with familiar EMF APIs.
2. The generic Ecore converter has enough hooks to plugin in the transformations.
3. Easy maintenance as the customizations are available in one place.
The drawback of this approach could be the processing involved in the model to model transformation. However, the highly optimized model to model transformation we implemented doesn’t make it all that slow. Our performance tests proves this (which would be the topic of my next blog post). It is infact upto 300 times faster than having the same data in the default XMI format (Yes I meant 300! Ed Merks is going to get back to me on this I hope ;) )
Ed did get back to me. With the performance optimizations applied to XMIResource loader to use cache for Intrinsic Ids and URIs, the performance improved dramatically. The XMI resource loader now takes only 22,2s to load a 32MB sample, compared to the earlier 70mins. I am glad that our RIF XMLResource however takes only 14,4s for the same contents.
Beginning of the year itemis opened its new location in Stuttgart. Now that we have settled down here, we made an official launch party. It was a nice little event attended by some of our customers and of course the itemikers. The event took place at “Stuttgarter Engineering Park (STEP)” where itemis has its new home.
- itemis was founded in 2003; 2006 in South Germany
- 10 locations in Europe now
- 150 employees in total, 18 at itemis Stuttgart
Markus Völter, then gave an excellent talk on “Model-Driven Development and DSLs“.During the one hour lively talk, he went on talking about the history of MDD and DSLs. What was even more interesting was when he displayed some DSLs he implemented for some real world applications like a “Pension Fund Specification” and “Fountains” (Fountain is no new buzzword, its the fountain you see at parks :) ).
The coffee break was a chance to get to know some new faces and welcome some familiar faces to our new location.
The second presentation of the day was from Andreas Graf and myself on “Requirements, Traceability & DSLs“. Andreas Graf’s “Presentation Zen style” slides appeared to catch everyone’s attention. He talked about the new activities at itemis on Requirements Engineering solutions. In between the talks, I managed to give a glimpse of the new tools with 3 demos.
The light dinner was again a chance to meet more people.
Thanks to all for attending. More images from the day can be found here.
Over the last few months some guys at itemis and Düsseldorf University have been working closely to bring out a solution to the open in an effort to lessen the big gap we have currently in Eclipse Ecosystem in the area of Requirements Management (RM). Requirements Modeling Framework (RMF) has been proposed to Eclipse Foundation as an open source project under Model Development Tools Project.
Scope of RMF
The scope of the project, as described in the proposal, is to provide an implementation for the OMG ReqIF standard (just as the Eclipse UML2 project provides an EMF based metamodel for OMG UML). Requirements management tools could then base their implementation on the provided ReqIF metamodel. It is also in the scope of the project to deliver a requirements authoring tool, again based on the ReqIF metamodel and optionally a generic traceability solution.
Significance of ReqIF
ReqIF (Requirements Interchange Format) is an open standard from OMG for requirements exchange (I had blogged a while ago on how we arrived at ReqIF (earlier called RIF)). Many tools like DOORS already support a snapshot export to this format. Having an EMF based metamodel for ReqIF, opens up the Eclipse framework for integration and new tool development in the area of Requirements Engineering.
ReqIF in a nutshell
ReqIF offers a generic metamodel to capture requirements (The generic nature is at times highly criticized. More on that later). The figure above shows a bird’s-eye view of the metamodel. The metamodel allows creation of requirement types (SpecType) with different attributes types and also instances for it (SpecObject). Since ReqIF also carries the metadata (in the form of SpecTypes), it makes it possible for any tool that understands ReqIF to process the data.
ReqIF also allows creating hierarchy for requirements, grouping them and controlling user access. To support rich text, the metamodel reuses XHTML.
Generic nature and “meta”ness of ReqIF
Anyone expecting a requirements metamodel might be surprised at first when they have a look at ReqIF. You hardly find a term called “requirement” in there. ReqIF is more at a higher meta-level, that is M1. It has been designed this way for a reason.
Requirements Management is an evolving field and until recently never crossed company boundaries. With tighter collaboration between partner companies the benefits of applying RM across company borders became known. The field being highly dominated by commercial products gave barely any chance for standardization. A group of companies in the automotive field, realized that with such a diversity existing hardly any unification is possible. They gave birth to this generic requirements interchange format which was later adopted by OMG and standardized.
The generic structure of ReqIF allows companies to use the tools of their choice to do requirements management and use ReqIF as the standard exchange mechanism and even do round trips. ReqIF is not the only standard to allow this generic nature in the field of Requirements Engineering. DOORS, for example, has an extensible database allowing users to add/delete attributes. The changes to the schema are often communicated to the partners by external means to replicate the changes in their database. Since ReqIF carries the meta-information with it, the changes are migrated automatically across tools/company boundaries. Transmitting the meta-information in ReqIF could however be controlled by tooling. This nevertheless would beat the purpose of ReqIF.
It is also questionable, why such a generic model is required when we have metamodels like UML or EMF already available. The reason for this is very much the same as why EMF co-exists with UML. ReqIF is not a general purpose modeling language like UML or EMF, but more focused on requirements domain.
RMF provides ReqIF as an EMF based metamodel. This could be used by tool vendors as the internal tool model or an export model by means of model to model transformation. The provided loaders and serializers make sure that the ReqIF file is read/written according to the ReqIF XML Schema.
Using ReqIF natively brings all the advantages of ReqIF to the tooling as well. For example, ReqIF based tools could model the requirements domain of the company within the tooling and share it with partners along with the instance data.
The initial code contribution is planned by early August 2011 and will be made available as soon as the IP review at Eclipse is complete. If you have any suggestions/comments about RMF, would like to contribute or like to be listed as an interested party, please provide it here.Images from FreeDigitalPhotos.net
1. Treating EMF generated code as only an initial code base
EMF is a good startup kit for introducing MDSD (Model-Driven Software Development) in your project. Anyone who has been introduced to EMF and had their first code generated using EMF will for sure be impressed by the short turn around time needed to have an initial version of your model based application up and running. For many, the thrill ends as soon as you need to change the behaviour of generated code and have to dig into it and understand it. It is complex, but only complex as any other similar sized framework. Many projects I have come across make a mistake at this point. They start treating the EMF generated code as an initial code base and start making it dirty by modifying it by hand. Adding @generated NOT tags in the beginning and not anymore later. At this point you part ways with “EMF way of doing things” and also with MDSD.
The customizations mostly start with the generated editor. This is quite acceptable as the editor is intended as an initial “template” and it expects you to customize it. However, this is not the case with model and edit projects. They have to be in sync with your model and this needs making some strong decisions, specifically – “I will not touch the generated code”. Follow EMF recommended way of changing generated code or use Dependency Injection.
2. Model modification without EMF Commands
In EMF you deal a lot with model objects. Almost any UI selection operation hands over a model object to you. It is so easy to fall into the trap of changing the state of the model objects directly once you have access to it, either using the generated model API or using reflective methods. Problems due to such model updates are detected only later, mostly when Undo/Redo operations stops working as expected.
Use EMF Commands! Modify your model objects only using commands. Either directly using EMF commands, extending them or creating your own commands. When you need to club many operations as a single command use CompoundCommands. Use ChangeCommand when you are up to making a lot of model changes in a transactional way.
3. Not reacting to notifications/Overuse of notifications
EMF notification is one the most powerful features of the framework. Any change to model objects are notified to anyone who is interested in knowing about it. Many projects decide to ignore listening to model changes and react to model changes by traversing the model again to look for changes or making assumptions about model changes and combining the behaviour into UI code.
When you want to react to a model change, don’t assume that you are the only one going to originate that change. Always listen to model changes and react accordingly. Don’t forget the handy EContentAdapter class.
On the contrary, some projects add a lot of listeners to listen to model changes without applying proper filters. This could greatly slow down your application.
4. Forgetting EcoreUtils
Before writing your own utility functions, keep an eye on EcoreUtils class. Mostly, the generic function you are trying to implement is already there.
5. Leaving Namespace URI as default
When you create an Ecore model, EMF generates a default Prefix and Namespace URI for you. Mostly this is left as it is without realizing that this is the most important ID for your new model. Only when newer versions of the model needs to be released later, people start looking out for meaningful URIs.
Change the default to a descriptive one before generating code. For example, including some version information of the model.
6. UI getting out of sync with model
The EMF generated editor makes use of Jface viewers. These viewers are always kept in sync using the generated ItemProviders. However these viewers cannot always satisfy the different UI requirements. You might have to use other SWT components or Eclipse forms. At this juncture, many implementors mix UI and model concerns together and in turn lose the sync between the model and UI.
Although, you could go ahead and implement your custom viewers, the easier way would be to use EMF Databinding.
7. Relying on default identification mechanism
Every EMF object is identified by a Xpath like URI fragment. For example, //@firstname.lastname@example.org. This is the default referencing mechanism within an EMF model or across models. If you look closely you could see that the default mechanism is based on feature names and indexes. The default implementation could turn dangerous if the index order changes and the references are not updated accordingly.
Extend this to uniquely identify your model object or simply assign an intrinsic ID to your model object. EMF will then use these IDs to reference your objects.
8. Forgetting reflective behavior
Instead of depending fully on generated APIs, think about creating generic reusable functions which makes use of the meta information every EObject carries. Have a look at the implementation of EcoreUtil.copy function to understand what I mean.
9. Standalone? Why do I bother
While developing applications based on EMF you should design your applications such that the core is not tied to Eclipse. This would allow your applications to be run even in a non OSGi server-side environment.