Les décisions architecturales ont émergé comme un moyen important pour maintenir la qualité de l’architecture pendant sa évolution. L’une des décisions les plus importantes faite par architecte sont celles à propos des approches de conception, à savoir des patrons ou des styles architecturales. La structure de ce genre de décision donne la possibilité d’être contrôlée automatiquement. Dans la littérature, il existe quelques travaux sur la vérification automatique de la violation des décisions architecturales. Dans cette thèse, nous montrons que ces travaux ne permettent pas de détecter toutes les violations possibles. Pour les compléter, nous proposons une approche qui i) décrit les patrons architecturaux qui contiennent la définition de la décision architecturale, ii) intègre les décisions architecturales au modèle architectural et iii) automatise la vérification de la conformité de la décision architecturale. Notre approche est implémentée en utilisant EMF et ses technologies accompagnées. Nous avons montré la possibilité de formaliser tous les patrons structuraux. A travers de deux expérimentes, nous avons montré que les décisions architecturales sont bien expliquées et toutes les violations sont détectées. Les systèmes logiciels composables sont prouvés capable de supporter l’adaptation aux nouvelles exigences grâce à leur flexibilité. Une méthode typique pour composer ces systèmes est de sélectionner et combiner des patrons qui adressent aux exigences de qualité attendues. Plusieurs propositions ont montré l’intérêt de la composition de patron. En revanche, l’un des défauts de ces propositions est la vaporisation de l’information de composition qui conduit au problème de la traçabilité et la reconstructibilité des patrons. Cette thèse a pour but de réserver le statut première classe aux opérateurs de composition pour stocker l’information de composition. L’approche est implémentée dans un outil et une étude empirique a été aussi conduite pour souligner ses intérêts. / Architectural decisions have emerged as a means to maintain the quality of the architecture during its evolution. One of the most important decisions made by architects are those about the design approach such as the use of patterns or styles in the architecture. The structural nature of this type of decisions give them the potential to be controlled systematically. In the literature, there are some works on the automation of architectural decision violation checking. In this thesis we show that these works do not allow to detect all possible architectural decision violations.To solve this problem we propose an approach which: i) describes architectural patterns that hold the architectural decision definition, ii) integrates architectural decisions into an architectural model and, iii) automates the architectural decision conformance checking. The approach is implemented using Eclipse Modeling Framework and its accompanying technologies. Starting from well-known architectural patterns, we show that we can formalize all those related to the structural aspect. Through two experiments, we show that architectural decisions are well explained and all of their violations are detected. Composable software systems have been proved to support the adaptation to new requirements thanks to their flexibility. A typical method of composable software development is to select and combine a number of patterns that address the expected quality requirements. A lot of work have shown the interest of pattern composition. Nevertheless, one of the shortcomings of these work is the vaporization of composition information which leads to the problem of traceability and reconstructability of patterns. This thesis also proposes to give first- class status to pattern merging operators to facilitate the preservation of composition information. The approach is tool-supported and an empirical study has also been conducted to highlight its interests

The success of a software project is strongly related with architectural design. However, designing the right Software Architecture is a very subjective task and takes a long time, being much influenced by architect's experience and the quality of requirements engineering. This architectural knowledge, usually, is not documented, since it is considered tacit knowledge of architects or other stakeholders, and eventually dissipates. It is also essentially important to ensure the consistency between software architecture and implementation. However, this mapping is usually made manually, based only on the developer's understanding over the software architecture, which requires high discipline. Thus, errors can arise during this phase, compromising the consistency amongst architectural decisions and source code. The objective of this work is to present a tool-based solution that supports young architects by recommending a suitable architectural style, based on the system's requirements, particularly the quality attributes of the system. The tool encompasses both trade-o resolution over quality attributes and recommendation of architectural styles based on quality attributes. Finally, based on the recommended architecture, the tool will generate the system structural source-code, using a component implementation model called COSMOS*, providing traceability between architectural design and implementation. The proposed solution has been evaluated in the context of a specific domain of Learning Management System (LMS), in order to illustrate the tool support in the execution of an architectural design process. / O sucesso de um projeto de software está fortemente relacionado com o projeto arquitetural. No entanto, projetar a arquitetura de software correta é uma tarefa muito subjetiva e leva muito tempo, sendo muito influenciada pela experiência do arquiteto e a qualidade da engenharia de requisitos. Este conhecimento arquitetural, geralmente, não está documentado, uma vez que é considerado o conhecimento tácito dos arquitetos ou dos interessados, e, eventualmente, se dissipa. Também é essencialmente importante assegurar a consistência entre a arquitetura de software e a implementação. No entanto, esse mapeamento é feito manualmente na maioria das vezes, baseado apenas no entendimento do desenvolvedor sobre a arquitetura, exigindo disciplina por parte dele. Assim, erros podem surgir durante esta fase, comprometendo a consistência entre as decisões arquiteturais e o código fonte. Em face destas dificuldades, foi desenvolvido este trabalho, cujo o objetivo é apresentar uma ferramenta que apoie jovens arquitetos com a recomendação de um estilo arquitetural adequado, baseado nos requisitos do sistema, particularmente os atributos de qualidade do sistema. A ferramenta compreende tanto resolução trade-off sobre os atributos de qualidade e recomendação de estilos arquiteturais com base em atributos de qualidade. Por fim, com base na arquitetura recomendada, a ferramenta irá gerar o código estrutural do sistema, utilizando um modelo de implementação de componente chamado COSMOS*, proporcionando rastreabilidade entre projeto arquitetural e a implementação. A solução proposta foi avaliada no contexto de um domínio específico dos Ambientes Virtuais e Aprendizagem (AVA), a m de ilustrar o suporte da ferramenta na execução de um processo de projeto arquitetural.