Für eine korrekte Darstellung dieser Seite benötigen Sie einen XHTML-standardkonformen Browser, der die Darstellung von CSS-Dateien zulässt.

. .
Suche
 

The SiDiff Model Generator

The SiDiff Model Generator

Description

The SiDiff Model Generator (SMG) is a highly configurable prototype tool for generating sequences of artificial models with almost real characteristics which are conforming to a meta-model. The produced models are syntactically correct and can contain complex structures. Additionally the generation process is managed by a configurable statistical controller to simulate realistic evolutions of models.

We presented our tool at ASE 2011. We also published a full conference paper @ SE 2012 in Berlin.

The SMG ist available in a virtual machine in the download section here.

Motivation

The necessity for the SMG arose in the context of the SiDiff Project. SiDiff is a highly configurable model comparison tool that can be adapted to any Ecore based modeling domain. We wanted to research the quality of model comparison algorithms by comparing the results computed by the algorithms with the actual changes that happened to the models. The problem was that for many domains either not enough models are available for testing or not enough is known about the modifications of available test models. Therefore the need for synthetic test models arose, but current state-of-the-art test model generators produced models of insufficient quality, because they violated one or more constraints of our application context:

  • Not only the produced models must be correct with regard to their meta models but also should be compliant with OCL constraints
  • The creation of complex structures is as important as creating simple structures (e.g. one should be able to create an association between two classes in a UML class diagram)
  • The statistical properties of the generated models should be adjustable and the creation process should be done in a controlled manner
  • When sets of related models are generated (i.e. a sequence of evolving models), then the change characteristics between subsequent models should be also under control
To solve this problem the SMG was devised within a project group at our chair. The SMG is highly configurable, can create new models of modify existing ones, can apply complex operations on model and allows validation of the correctness of the created models.

Application Scenarios

Our Application Scenario

As briefly described earlier, the motivation of creating our SMG prototype tool arouse in the context of the SiDiff and the QuDiMo projects. The main goal is to test the existing algorithms that are used in detection of changes that happen to a software model. Each of these algorithms have some characteristics and in order to validate them and to test their quality and their performance one should use some standard input models. The problem is that for many domains no test data are available and in the case of existence of some test data in some domains, they do not have some standard useful features. For an example how created models are used, click here.

Other Application Domains

By using the SMG one is able to create syntactically valid models that have almost real characteristics in a controlled stochastic way. These kind of models can be used in our application scenario or any other scenarios that there is a need for synthetic benchmarks, like model design and representation tools, model repositories, model search and queries, model storage and retrieval etc.

Supported Model Types

Currently the following model types are supported. We are working on support for other model types as well.

  • SiDiff for UML
  • Library Example Model
  • SiDiff JavaAST Model (reverse engineered models)
  • more to come...

Benefits and Advantages of the SMG:

  • Capability of creating almost real models
  • Highly configurable in order to simulate the behavior of model evolution
  • It is able to produce not only one artificial model but also a sequence of models
  • The production process is additionally equipped with the log of model transformations
  • Validation and consistency checks for the generated results
  • The Stochastic Controller of the SMG makes it possible to produce different scenarios in model evolutions by it's fine tunable configuration file and it is also being extended by adding more functionality
  • It can use synonym words by employing WordNet to reflect common updates of names
  • It can support Simple Edit Operations as well as Complex Edit Operations which are defined on different model types
  • Easily adaptable to support any EMF based model type

Team Members

Pit Pietsch
Hamed Shariat Yazdi

Student members of the development team (in alphabetical order):

Andre Bertels
Thi Minh Hoa Nguyen
Michaela Rindt
Petrissa Roth
Tim Sollbach
Davy Franck Wanmeni


This website uses Google Analytics Datenschutzerklärung | Disclaimer