Verification of Model Transformations: A Case Study with BPEL

Abstract

Model transformations, like refinement or refactoring, have to respect the semantics of the models transformed. In the case of behavioural models this semantics can be specified by transformations, too, describing an abstract interpreter for the language. Both kinds of transformations, if given in a rule-based way, can formally be described as graph transformations.In this paper, we present executable business processes, their operational semantics and refactoring, as an example of this fact. Using results from graph transformation theory about critical pairs and local confluence, we show that our transformations preserve the semantics of processes. The analysis is performed by means of the graph transformation tool AGG.

Formal Specification and Testing of Model Transformations

“…In the first place we have the works that aim at fully validating the behaviour of the transformation and its associated properties (confluence of the rules, termination, etc.) using formal methods and their associated toolkits (see, e.g., [5][6][7][8][9][10][11]). The potential limitations of these proposals lie in their inherent computational complexity, which makes them inappropriate for fully specifying and testing large and complex model transformations.…”

Typing Model Transformations Using Tracts

“…Another group of works (see, e.g., [20][21][22][23][24]) also use a white-box approach to model-transformation specification and testing, aiming at fully validating the behaviour of the transformation (including other properties such as confluence of the rules, termination, etc.) using formal methods and their associated toolkitswhich include, e.g., Alloy, Maude, or graph rewriting techniques.…”

Tractable Model Transformation Testing

“…Other group of works (see, e.g., [5,6,7,8,10,11]) also use a white-box approach to model-transformation testing, aiming at fully validating the behaviour of the transformation (including other properties such as confluence of the rules, termination, etc.) using formal methods and their associated toolkits-which include, e.g., Alloy, Maude, or graph rewriting techniques.…”

“…In the first place we have the works that aim at fully validating the behaviour of the transformation and its associated properties (confluence of the rules, termination, etc.) using formal methods and their associated toolkits (see, e.g., [5,6,7,8,9,10,11]). The potential limitations with these proposals lies in their inherent computational complexity, which makes them inappropriate for fully testing large and complex model transformations.…”