Thursday, August 30, 2012

Modelitis, or, everything you always wanted to know about modeling

A couple of years ago I left the world of industrial software projects, and joined the academic world of computer science. I already had used models, I knew quite a few things about UML, even had programmed a graphical UML editor using GEF, in other words, I felt prepared to explore the universe of modeling. However, as my former boss and tutor Prof. Dr. Hans-Werner Six correctly diagnosed, I soon got infected with a terrible illness, widely spread in that area: modelitis. Actually, depending on how hard this disease hits you, you may call it meta-modelitis.

A special characteristic of this disease is that the patient himself is more or less unaware of it, actually, he or she often thinks it's great fun. So did I, and frankly, I doubt that I will ever be completely cured -- particularly since there are so many other victims of modelitis doing great and funny things. E.g.,  inventing one TLA (for non-modelitis afflicted readers: three letter acronym) after the other, such as MDA, MDD, DSL, MOF, ATL, EMF, ETL, GEF, GMF, OCL -- you name it. I was no better, I even tried certain variations, such as CAT, GEF3D, or Mitra (my own M2M-transformation language).

In the end, I wrote a book about my modelitis. It's title is

"Computerunterstützte Modelltransformationen. Modellierungstheorie, Konzeption und Visualisierung im Rahmen modellgetriebener Entwicklungsverfahren"

Yeah, it's a rather long title, and you can figure from its length that it's a real hard-core academic work (aka dissertation). Translated to english it's

"Computer-assisted transformations. Modeling theory, design and visualization in the context of model-driven development"

You may have noticed that "computer-assisted transformation" is abbreviated to CAT :-) This is the abstract of my thesis (in english, the thesis itself is written in german):

In the context of model driven development, fully automated model transformations are used to release developers from recurring and error-prone tasks, thus improving efficiency. However, these transformations cannot always be applied, or at least not without introducing new problems. By exploring semi-automated techniques, this work tries to eliminate some of these problems while preserving the positive effects. First, a modeling theory is developed for describing model driven approaches. This theory is used to analyze a typical problem of these approaches, the so called "semantic gap". The need to add semantics when transforming models is identified as a fundamental problem. Computer-assisted transformations are a semi-automated approach, which allow to combine manual transformation parts with automated parts. In order to define and execute this kind of transformations, a model-to-model transformation language called Mitra is designed. Due to the manual nature the approach, the models need to be visualized and a user- interface is to be provided. The basic idea of the proposed solution is the visualization of graphical models by means of two-dimensional diagrams projected onto planes in a three-dimensional scene. This permits not only the visualization of inter-model relations, but also semi-automated transformation of model elements with computer-assisted transformations employing the common drag-and-drop interaction pattern. With GEF3D, a framework is presented, which enables the combination of existing graphical editors in a three-dimensional scene. Three well-known transformation problems, among others the robustness analysis, are solved with the created tools in order to evaluate the proposed concepts.

If you are suffering modelitis as well, or if you are a otherwise seriously addicted to meta-modelling, you may download the PDF at the library of the FernUniversität in Hagen. If you want to look at some results, you may have a look at GEF3D, or at Mitra2. The latter is work in progress, as I'm currently migrating from Xtext 1.x to Xtext 2.x (the need for migrating from one version to another is one of the symptoms of modelitis, e.g., victims will enthusiastically tell you how they migrated their models from UML 1.x to UML 2.x ;-) ).

Since I doubt that too many of you will actually download and read the PDF, I will only repeat a snippet from my acknowledgement section:
 A big thank you to the Eclipse community! 
Without the great software, particularly in the modeling project, and without the great support (when I had problems with the great software ;-) ), I wouldn't had been able to actually implement software realizing my ideas and by that proof the concepts. And it wouldn't had been so much fun.

(All pictures taken from the dissertation)

Thursday, May 10, 2012

CfP: Methodical Development of Modeling Tools

Jens Gulden and I are organizing an international workshop, called "Methodical Development of Modeling Tools" (MeDMoT'12), which I assume to be particularly interesting for Eclipse developers, since modeling tools often leverage Eclipse (modeling) technologies (are there any modeling tools out there not using EMF? Just kidding ;-) ). It is held in conjunction with the 15th IEEE International Conference on Computational Science and Engineering (CSE 2012), 3-5 Oct 2012, Paphos, Cyprus. I'm especially happy about the great program committee, with experienced reviewers from both, academic and industry: Ulrich Frank, Holger Giese, Reinhard von Hanxleden, Jürgen Jung, Lutz Kirchner, Dimitris Kolovos, Paul McIntosh, Jan Mendling, Ed Merks, Erik Proper, Matti Rossi, Friedrich Steimann, and Stefan Strecker. A big thank you to all reviewers for joining the PC!

You can find more information at the official workshop website.

Snippet from the official call for papers:

Software tools for modeling are prerequisites for scientific and practical applications of modeling methods and modeling languages. To test and exemplify new modeling approaches, research prototypes of model editors are required, as well as tools for presenting and analyzing models. To be able to efficiently develop such modeling tools, it is desirable to methodically guide the development of modeling tools, and elaborate procedures to align their design with the conceptualization of new modeling languages and modeling methods.

The workshop focuses on procedures and architectural principles related to the creation of software for presenting, editing, transforming or analyzing models. This covers special constellations related to the use of models and modeling languages, for example, the ability to automatically derive model editor functionality from formal specifications of modeling languages (meta-models, grammars, etc.). In addition, design principles and implementation options for requirements towards modeling tools, such as wizard support for interactive modeling, or alternative ways to display graphical models, are discussed. All kinds of models are included in this discussion, either domain-specific or general-purpose models, as well as graphical models and textual models with their related software tooling support. Intended audience are scientists and practitioners, who apply modeling techniques and model-driven procedures, and develop their own modeling tools.

Submissions may document research work on newly created modeling tools, either for prototype or production purposes, as well as work on methodology and architecture of modeling tool development. This may cover model editor software in a narrow sense, as well as any approach for handling, transforming or analyzing models in a wider sense.

Important dates

2012-06-10Paper submission (with blinded author names)
2012-07-15Notification of acceptance
2012-07-31Camera-ready article

This workshop is the (international) successor of a german workshop series called "MEMWe". It was held in conjunction with the "Informatik" conference (the annual conference of the "Gesellschaft für Informatik") in 2011, 2010, and 2009.

Friday, April 27, 2012

Hosted Continuous Integration

Chris blogged about Travis CI a while ago, a site which provides free CI for github projects. Unfortunately, Travis is for github projects only. So, I was wondering if there are other hosted CI solutions available. Here are the results a short google session, some tools are also suggested at stackoverflow:
SiteCI systemBuild systemVCSTriggerNotificationPricingDeploymentComment
Travis CI Travis Maven, Ant, and others git (github only) github: Travis hook email free for OS no still alpha
CloudBees, FOSS free Jenkins Maven git, SVN, and others github: Jenkins hook ? free for OS ?
FaZend Hudson Maven, Ant, and others SVN ? email free ? still beta
CodeBetter TeamCity Maven git, SVN, and others github: TeamCity hook email, Jabber, and others free for OS ?
Shining Panda Jenkins Maven git, SVN, and others github: Jenkins hook email €0.36 / hour ? Focused on Python
Jenkins Hosting (Bubbleware Technology) Jenkins ? git, SVN ? ? free for OS (+ Amazon EC2 fees) ? ? ? ? ? ? free for OS (+ Amazon EC2 fees) ?
I focused on support for Java projects and Maven/Ant support. Some remarks:
  • Trigger: This means how the build is triggered. This is of course not a feature of the CI system itself, but of the source code repository. Github or Bitbucket provide some hooks, unfortunately, EclipseLabs (GoogleCode) doesn't, or does it?
  • Notification: This means how the system does notify the user of the build result. After a successful build, one probably want to somehow deploy the created artifacts.
  • Deployment: Actually, I'm looking for a solution of copying the artifacts, which actually are a P2 repository, to some download or update site. Of course, this could be achieved by Maven plugins or Ant tasks, however, the crucial point is how to pass some credentials (user name, pass word) to the build system. Especially if you are building an open source project, you do not want to write this information into the maven script. Seems as if this is a common problem. If this problem were solved by Travis, it would be the best solution, at least for github projects.

Actually, it is very hard to determine the features of the different solutions. Many projects seem to be in a very early stage, and some projects announced are already gone (e.g., Mike CI). Most projects are cloud based, some use Amazon EC2 and users need their own Amazon EC2 account. For open source projects, it is often required to fill out some request form, and I don't know whether all projects are accepted.

Leave me a comment if you have any experience with a solution listed above, or if you know of other hosted CI systems.

Wednesday, February 1, 2012

Dreaming of an Eclipse Plugin-Store...

or how to sell little tools for little money...

A while ago, I wrote a little tool for exporting UML-like diagrams for Java classes and packages to OmniGraffle. I blogged about that tool and, from my statistics, it got downloaded over 100 times. I also announced that the tool will stop working in 2012, which it actually did. As I wrote in the announcement of the tool, based on the feedback I wanted to decide whether to continue developing the tool or not.

So, what is the feedback after almost six months? There were a few comments, and about 150 downloads according to my logs. Although I installed a donate and Flatter button on my blog, I received no money at all. The natural consequence would be to stop developing (and providing) the tool.

Today, someone posted a comment as the tool has stopped working (just as announced). The commenter also wrote that he needs the tool to create some diagrams for him. Hmm... As I'm trying to be a good guy, I published an update of the tool working until June 2012.

I won't complain about people not giving any feedback or money voluntary. Instead I'm wondering how to make the tool available for a small amount of money. Actually, I don't know about any Eclipse tools to be sold for a couple of dollars/euros, except Log4E. Most tools are either freely available, or they are really expensive. Log4E comes in two versions: A free community edition and a so called "Pro" version for only €7,50. I have purchased that tool a long time ago, rather to support the author than urgently needing the additional features of the Pro version. However, due to the lack of Eclipse supporting this kind of "business model", it is rather complicated to install the license key (and keep it up to date with new Eclipse installations), and I figure managing licenses and payments to be time-consuming for the author as well.

Apple, and also Google, have successfully created systems enabling authors of software to make (little) money by selling there products very easily. Although I don't like the Apple way of approving software, I'm wondering if some kind of "Eclipse PluginStore" would be a good idea. People spend a lot of money on "apps", including a lot of small games. If buying a commercial Eclipse plugin would be as simple as purchasing an iPhone/Android app, would people do that? And how many programmers would publish their tools then? Maybe combining that kind of store with a BugStore (see "Should We Pay for Eclipse Bug Fixes?" for a summary of a discussion taken place in April 2010) would be a good idea as well...

As long as there is no Eclipse PluginStore available: How do you sell your tools?