Quality-Driven Detection and Resolution of Metamodel Smells

Abstract

In model-driven engineering, analogously to any software development practice, metamodel design must be accurate and performed by considering relevant quality factors, including maintainability, reusability, and understandability. The quality of metamodels might be compromised by the introduction of smells that can be the result of inappropriate design decisions. Detecting and resolving metamodel smells are a complex task. The existing approaches deal with this problem by supporting the identification and resolution of smells without providing the means to explicitly trace them with the quality attributes that can be potentially affected. In this paper, we present an approach to defining extensible catalogues of metamodel smells. Each smell can be linked to the corresponding quality attributes. Such links are exploited to automatically select only those smells that have to be necessarily resolved for enhancing the quality factors that are of interest for the modeler. The implementation of the approach is based on the Edelta language, and it has been validated on a corpus of metamodels retrieved from a publicly available repository. INDEX TERMS Domain-specific languages, model-driven engineering, software quality engineering.

Edelta 2.0

“…3This phase consists of an analysis process on the candidate metamodels in order to understand what types of refactorings should be applied. Automatic bad smell matching [5] or anti-patterns [30] can be applied in order to identify candidate elements to be refactored. (4) In this phase, the modifications are eventually applied in a dedicated Edelta program.…”

“…Because of all the linguistic mechanisms just mentioned, the programmer can easily write statements and expressions that are compact and more readable than in Java ( 5 ). For example, by leveraging the Edelta Java API described in Section 2.1, we can create an EClass with a required EAttribute like this in Edelta: We use the addNewXXX methods from our API.…”

“…The referred code must be specified as a dependency of the project. This works both for your own reusable operations and the ones that we provide together with Edelta, described in [5].…”

“…For example, in [5] we presented our Edelta reusable library of metamodel bad smell finders and the corresponding refactorings. The bad smell finders have been modified so that they use showWarning when they spot a bad smell.…”

“…In [31] the authors analyzed source code version control system logs of popular open source software systems to detect refactoring applications and see how software metrics are affected by these refactorings. We applied Edelta to a similar process in [5] where the refactoring is applied in order to understand if correlated metamodeling quality attributes are improved.…”