Model repair and transformation with Echo

Abstract

Abstract-Models are paramount in model-driven engineering. In a software project many models may coexist, capturing different views of the system or different levels of abstraction. A key and arduous task in this development method is to keep all such models consistent, both with their meta-models (and the respective constraints) and among themselves. This paper describes Echo, a tool that aims at simplifying this task by automating inconsistency detection and repair using a solver based engine. Consistency between different models can be specified by bidirectional model transformations, and is guaranteed to be recovered by minimal updates on the inconsistent models. The tool is freely available as an Eclipse plugin, developed on top of the popular EMF framework, and supports constraints and transformations specified in the OMG standard languages OCL and QVT-R, respectively.

A Feature-Based Classification of Model Repair Approaches

“…These may be built around standard languages/architectures like XML, MDA or EMF, or other specific to the technique. This technical space defines the concrete model syntax that the technique is able to process, like XML [49], XMI [35,43,45], UML [15,21,29,40,48,60,67,70], or a technique-specific language [57,68]. These concrete artifacts are translated by the technique into their representation in the underlying formalism (Section 3.1.1).…”

“…If the user is allowed to define or customize constraints (Section 3.2.1), this feature defines the language in which he is able to do so. Typically this amounts to some version of MDA's OCL [15,16,33,35,45,61], that is also prescribed in EMF, or it can be designed specifically for the technique [71]. In techniques with support for inter-model constraints (Section 3.2.2), standard languages include MDA's QVT [53] standard [43,44].…”

“…The most common means to define intra-model consistency is through the definition of logical constraints [10,13,[15][16][17]21,23,35,36,39,45,49,51,57,58,60,62,[67][68][69]71]. These may also be used to define inter-model consistency, assuming they are able to refer to elements from different models [20,26,68].…”

“…In the most typical scenario of violation selection, inconsistency levels I take the shape I 1 × I 2 , a pair whose first element states whether the selected violation was removed, and the second element provides information regarding the remainder environment constraints, allowing the user to be aware of possible side-effects. Composite reports may also arise when the techniques distinguish different classes of constraints, for instance intra-and inter-model constraints [45]. In these composite reports there is more than a single sensible partial order over each shape of I.…”

Writing bidirectional model transformations as intentional updates

“…For instance, the backward transformation inferred by EMoflon [1] (TGGs) would ignore current source people and generate a new set of people afresh from a new set of tokyoites; Echo [10] (QVT-R) would return a nondeterministic choice between all consistent source sets at a minimal distance to the current source set under some distance measure; while SyncATL or GRoundTram would try to propagate the modifications made to the view set of tokyoites as inferred replace/insert/delete operations on the current source set of people 1 .…”