Face à la complexité des applications logicielles multitâches, une approche prometteuse est l'automatisation des développements. En pratique, cela se concrétise par des générateurs capables de produire les implantations propres aux plates-formes logicielles d’exécution multitâche. Ces générateurs sont figés autour d'heuristiques d'implantations propres à chaque plate-forme visée. En vue d'obtenir des solutions plus flexibles, cette étude a pour objectif d'externaliser les formalismes propres aux plates-formes d'exécution dans des modèles explicites. Ces modèles sont alors utilisés en entrée des ingénieries ce qui permet de capitaliser et de réutiliser les générateurs. Pour y parvenir, cette étude définit un motif générique pour la modélisation des plates-formes logicielles d'exécution, une extension au langage UML pour la modélisation des plates-formes multitâches et une infrastructure de transformation intégrant ces modèles de plates-formes. / To minimize the inherent complexity of multitasking programs, a promising approach is to automate developments. In practice, automation is achieved by generators. Those generators produce applications which execute on software multitasking platforms (for example multitasking operating systems). Such generators are in fact specific to selected platforms. They are made of implementation rules which are specific to each platform. In order to cope with adaptable and flexible solutions, this study aims to explicitly describe executing platforms as parameters of generators. For that, it defines, firstly, a generic pattern dedicated to modelling software execution platforms, secondly, an extension to the UML language for modelling multitasking software execution platforms (the Software Resource Modeling profile) and, thirdly, a transformation framework based on explicit platform models.

No description available.

<p>o aid the software and database developer in the development process, specialised software is needed, known as CASE-tools. To form a productive work environment, multiple CASE-tools should be able to cooperate with each other, as it is likely that a single tool cannot give full support for the whole development process. An aid in the integration of tools is that information is stored in a central repository, available for any tool that needs it. A problem which arises is that not all tools are capable of directly accesing the contents of the repository. Thus export procedures are needed that transforms the repository contents into a format that an external tool can read and understand. If these transformation procedures are specified directly in the implementation language, modification can be difficult or even impossible.</p><p>This work proposes a general transformation model for storing rules in a repository. These rules operate on the information in a source model and transform it to a textual representation ready for export to external tools. An example ruleset is given that transforms an example relational model into SQL code.</p>

It is widely accepted within the software engineering community that the support for integration is necessary for requirement models. Several methodologies, such as the role-based software development, that have appeared in the literature are relying on some kind of integration. However, current integration techniques and their tools support are insufficient. In this research, we discuss our solution to the problem. More precisely, we present a general integration approach for scenario-based models, particularly for UML Sequence Diagrams, based on the colimit construction known from category theory. In our approach, Sequence Diagrams are represented by SD-graphs, a special kind of typed graphs. The merge algorithm for SD-graphs is an extension of existing merge operations on sets and graphs. On the one hand, the merge algorithm ensures traceability and guarantees key theoretical properties (e.g., “everything is represented and nothing extra is acquired” during the merge). On the other hand, our formalization of Sequence Diagrams as SD-graphs retains the graphical nature of Sequence Diagrams, yet is amenable to algebraic manipulations. Another important property of our process is that our approach is applicable to other kinds of models as long as they can be represented by typed graphs. A prototype Sequence Diagram integration tool following the approach has been implemented. The tool is not only a fully functional integration tool, but also served as a test bed for our theory and provided feedback for our theoretical framework. To support the discovery and specification of model relationships, we also present a list of high-level merge patterns in this dissertation. We believe our theory and tool are beneficial to both academia and industry, as the initial evaluation has shown that the ideas presented in this dissertation represent promising steps towards the more rigorous management of requirement models. We also present an approach connecting model transformation with source transformation and allowing an existing source transformation language (TXL) to be used for model transformation. Our approach leverages grammar generators to ease the task of creating model transformations and inherits many of the strengths of the underlying transformation language (e.g., efficiency and maturity). / Thesis (Ph.D, Computing) -- Queen's University, 2009-08-28 13:03:08.607

Adaptive distributed systems have the potential to revolutionize how humans and computers interact. They can enable software to adapt to dynamic human demands, as users change their focuses, goals, locations and devices. However, producing these systems is currently very challenging: developers must solve daunting user interface issues while mired in distributed systems problems. We present a new class of toolkit, intended to ease the development of adaptive distributed systems. Unlike existing alternatives, we provide a high-level programming model in which developers can easily specify runtime adaptations. Meanwhile, our toolkit automatically generates a fast and tunable implementation. Partial failures in the distributed system are reflected back into the high-level programming model. As a result, developers can remain insulated in their high-level model while building highly-dynamic, high-performance and failure-resistant applications. Our Fiia.Net toolkit relies on model transformation at runtime to bridge between the programmer's high-level model and the actual implementation of the distributed system. Our novel model transformation is the first that can practically maintain this transformation, and enables our toolkit to easily support user-driven adaptations, dynamic optimization and self-healing. / Thesis (Ph.D, Computing) -- Queen's University, 2011-02-25 14:08:13.69

Dans le monde des affaires d'aujourd'hui, les besoins des clients changent plus rapidement que jamais et le nombre de concurrents augmente chaque seconde. De plus, la capacité à gérer les changements et l'imprévisibilité sont devenus un facteur crucial pour que les entreprises gagnent en valeur et restent compétitives [Oracle 2013]. Il en résulte que les entreprises sont aujourd'hui confrontées non seulement à la gestion de processus métier structurés, mais également à des processus de plus en plus non structurés. Dans un processus métier structuré commun, tout ce qui concerne le processus peut être prédéterminé au moment de la conception, tel que les activités, la séquence d'exécution des activités, etc. Cependant, dans une structure non structurée, les activités ne peuvent pas être définies précisément au préalable, ainsi que la séquence à exécuter. Pour rester compétitives, répondre aux demandes du marché en constante évolution et améliorer l'efficacité opérationnelle de leurs processus d'affaires, les organisations ont besoin d'une approche de processus novatrice qui peut les aider à gérer les changements, la dynamique et l'imprévisibilité. Dans ce contexte, le concept de Case Management est proposé. Différent du Business Process Management (BPM) qui standardise et automatise les processus métier structurés, CM surmonte les limites de l'approche BPM et fournit une infrastructure pour gérer les changements, la dynamique et l'imprévisibilité dans les processus métier non structurés. Le CM procède en grande partie en fonction de l'évolution des circonstances, et les décisions sont prises à la volée. BPM nécessite un haut niveau de prévisibilité; considérant que le CM a un niveau de prévisibilité moindre mais un niveau d'adaptabilité et de flexibilité plus élevé. Avec l'approche CM, les entreprises sont en mesure de gérer leurs processus métier non structurés de manière plus souple et plus flexible.Cependant, pour ce nouveau domaine, il manque des méthodes de soutien et des outils logiciels. Les principales préoccupations sont: (1) la modélisation de cas (la construction de modéles de cas); (2) la découverte de modèles (l'établissement de modèles de cas à partir de données brutes); (3) l'analyse de modèles (l'analyse de modèles à la fois statiques et dynamiques, par exemple, la dérivation de propriétés avant la mise en place du cas); (4) amélioration du modèle (réduction des coûts, optimisation de la performance opérationnelle, etc.); et (5) la promulgation du modèle (l'exécution d'un scénario avec des case workers dans la boucle). Après un examen approfondi de la littérature, nous avons constaté que seulement quelques efforts ont été faits dans (1) et (5), et aucune contribution notable n'a été faite dans d'autres aspects.Ceci présente notre approche CM qui fournit aux assistants un support complet dans tout le cycle de vie du CM: de l'établissement de modèles de cas à partir de données brutes jusqu'à l'optimisation des modèles de cas. Process Tree est notre choix pour formaliser le modèle découvert, et CMMN (Case Management Model and Notation, une spécification de modèlisation de cas est choisie comme formalisme pour la présentation et la construction de modèles de cas.) En outre, nous adoptons le langage HiLLS. L'analyse de modèle dynamique est permise par le formalisme DEVS, l'analyse de modèle statique est fournie par des méthodes formelles. Quand à la mise en oeuvre de modèle, elle est permise par la mise en ?uvre d'une spécification orientée objet du cas. Nous proposons principalement deux modules: un module concernant la découverte du modèle de cas à partir des historiques d'événements, et un autre module concernant l'amélioration et l'optimisation du modèle de cas. / In today's business world, customer requirements change more rapidly than ever before, and new competitors are increasing every second. Moreover, the ability of managing changes and unpredictability has become a crucial factor for enterprises to make more value and stay competitive [Oracle 2013]. This results in a fact that nowadays enterprises are challenged with not only managing structured business processes, but also more and more unstructured ones. In a common structured business process, everything regarding the process can be predetermined at design time, such as activities, the execution sequence of activities, and so on. However, in an unstructured one the activities cannot be defined precisely beforehand, as well as the sequence to execute. To stay competitive, meet the ever-changing market demands and improve their business process operational efficiency, organizations need a novel process approach that can help them manage changes, dynamics and unpredictability. Under this context, the concept of Case Management is proposed. Different from Business Process Management (BPM) which standardizes and automates structured business processes, CM overcomes the BPM approach limitations and provides an infrastructure for managing changes, dynamics and unpredictability in unstructured business processes. CM proceeds largely depending on evolving circumstances, and decisions are made on the fly. BPM requires a high level of predictability; whereas CM has a lower level of predictability but a higher level of adaptability and flexibility. With CM approach, enterprises are able to manage their unstructured business processes in a more adaptive and flexible manner.However, for this new area it lacks supporting methods and software tools. Major concerns are: (1) case modeling (the construction of case models); (2) model discovery (the establishment of case models from raw data); (3) model analysis (the analysis of models in both static and dynamic manners, e.g., the derivation of properties before the case is enacted); (4) model improvement (the reduction of cost, the optimization of operational performance, etc.); and (5) model enactment (the execution of a case scenario with case workers in the loop). After a thorough literature review we found that only a few efforts have been done in (1) and (5), and no noticeable contribution has been done in other aspects.This these presents our CM approach that provides case workers full supports in the whole lifecycle of CM: from establishing case models from raw data to optimizing case models. Process Tree is our choice to formalize the discovered model, and CMMN (Case Management Model and Notation, a case modeling specification is selected as the formalism for presenting and constructing case models. In addition, we adopt the HiLLS (High Level Language for Systems Specification) formalism to conciliate usability, simulation ability and formal analysis capabilities together. Dynamic model analysis is enabled by DEVS formalism, static model analysis is provided by formal methods, and model enactment is given by the implementation of an object-oriented specification of the case. We propose mainly two modules in this these: one module concerning the discovery of the case model from historical event logs, and another module concerning the improvement and the optimization of the case model.

o aid the software and database developer in the development process, specialised software is needed, known as CASE-tools. To form a productive work environment, multiple CASE-tools should be able to cooperate with each other, as it is likely that a single tool cannot give full support for the whole development process. An aid in the integration of tools is that information is stored in a central repository, available for any tool that needs it. A problem which arises is that not all tools are capable of directly accesing the contents of the repository. Thus export procedures are needed that transforms the repository contents into a format that an external tool can read and understand. If these transformation procedures are specified directly in the implementation language, modification can be difficult or even impossible. This work proposes a general transformation model for storing rules in a repository. These rules operate on the information in a source model and transform it to a textual representation ready for export to external tools. An example ruleset is given that transforms an example relational model into SQL code.

No description available.

Model Driven Engineering (MDE) has gained a considerable attention in the software engineering domain in the past decade. MDE proposes shifting the focus of the engineers from concrete artifacts (e.g., code) to more abstract structures (i.e., models). Such a change allows using the human intelligence more efficiently in engineering software products. Model Transformation (MT) is one of the key operations in MDE and plays a critical role in its successful application. The current MT approaches, however, usually miss either one or both of the two essential features: 1) declarativity in the sense that the MT definitions should be expressed at a sufficiently high level of abstraction, and 2) formality in the sense that the approaches should be based on precise underlying semantics. These two features are both critical in effectively managing the complexity of a network of interrelated models in an MDE process. This thesis tackles these shortcomings by promoting a declarative MT approach that is built on mathematical foundations. The approach is called Query Structured Transformation (QueST) as it proposes a structured orchestration of diagrammatic queries in the MT definitions. The aim of the thesis is to make the QueST approach –that is based on formal foundations– accessible to the MDE community. This thesis first motivates the necessity of having declarative formal approaches by studying the variety of model synchronization scenarios in the networks of interrelated models. Then, it defines a diagrammatic query framework (DQF) that formulates the syntax and the semantics of the QueST collection-level diagrammatic operations. By a detailed comparison of the QueST approach and three rule-based MT approaches (ETL, ATL, and QVT-R), the thesis shows the way QueST contributes to the development of the following aspects of MT definitions: declarativity, modularity, incrementality, and logical analysis of MT definitions. / Thesis / Doctor of Philosophy (PhD)

L'Ingénierie Dirigée par les Modèles (IDM) est un domaine de recherche en pleine émergence qui considère les modèles comme des éléments de base. Chaque modèle est conforme à un autre modèle, appelé son méta-modèle, qui définit sa syntaxe abstraite et ses concepts. Dans un processus IDM, différents types de modèles sont manipulés par des transformations de modèles. Une transformation génère un modèle dans un langage cible à partir d'un modèle dans un langage source. Pour concevoir une transformation, les développeurs doivent avoir une bonne connaissance des méta-modèles concernés ainsi que des langages de transformation, ce qui rend cette tâche difficile. Dans cette thèse, nous proposons d'assister l'écriture des transformations et plus généralement de comprendre comment une transformation opère. Nous adhérons à l'approche de transformation de modèles par l'exemple qui propose de créer une transformation de modèles à partir d'exemples de transformation. Cela permet d'utiliser la syntaxe concrète définie pour les méta-modèles, et cela évite donc de requérir que les développeurs aient une bonne maîtrise des méta-modèles utilisés. Dans ce contexte, nous proposons deux contributions. La première consiste à définir une méthode pour générer des règles de transformation opérationnelles à partir d'exemples. Nous nous basons sur une approche qui utilise l'Analyse Relationnelle de Concepts (ARC) comme technique d'apprentissage pour obtenir des patrons de transformation à partir d'un appariement de type 1-1 entre les modèles. Nous développons une technique pour extraire des règles de transformation opérationnelles à partir de ces patrons. Ensuite, nous utilisons le langage et le moteur de règles JESS pour exécuter ces règles. Nous étudions aussi comment mieux apprendre des règles de transformations à partir d'exemples, en utilisant séparément chaque exemple ou en réunissant tous les exemples. La deuxième contribution consiste à récupérer les traces de transformation à partir d'exemples de transformation. Ces traces peuvent être utilisées par exemple pour localiser des erreurs durant l'exécution des programmes de transformation ou vérifier la couverture de tous les modèles d'entrée par une transformation. Dans notre contexte, nous supposons que ces traces vont servir pour un futur apprentissage des règles de transformation. Nous traitons tout d'abord le problème de récupération des traces avec des exemples provenant d'un programme de transformation. Nous proposons une approche basée sur une méta-heuristique multi-objectifs pour générer des traces sous forme d'appariement de type n-m entre des éléments de modèles. La fonction objectif s'appuie sur une similarité lexicale et structurelle entre ces éléments. Une extension de cette méthode est proposée pour traiter le problème plus général de l'appariement entre modèles. / Model Driven Engineering (MDE) considers models as first class artifacts. Each model conforms to another model, called its metamodel which defines its abstract syntax and its semantics.Various kinds of models are handled successively in an MDE development cycle. They are manipulated using, among others, programs called model transformations. A transformation takes as input a model in a source language and produces a model in a target language. The developers of a transformation must have a strong knowledge about the source and target metamodels which are involved and about the model transformation language. This makes the writing of the model transformation difficult.In this thesis, we address the problem of assisting the writing of a model transformation and more generally of understanding how a transformation operates.We adhere to the Model Transformation By example (MTBE) approach, which proposes to create a model transformation using examples of transformation. MTBE allows us to use the concrete syntaxes defined for the metamodels. Hence, the developers do not need in-depth knowledge about the metamodels. In this context, our thesis proposes two contributions.As a first contribution, we define a method to generate operational transformation rules from transformation examples. We extend a previous approach which uses Relational Concept Analysis as a learning technique for obtaining transformation patterns from 1-1 mapping between models. We develop a technique for extracting relevant transformation rules from these transformation patterns and we use JESS language and engine to make the rules executable. We also study how we better learn transformation rules from examples, using transformation examples separately or by gathering all the examples.The second contribution consists in recovering transformation traces from transformation examples. This trace recovery is useful for several purposes as locating bugs during the execution of transformation programs, or checking the coverage of all input models by a transformation. In our context, we expect also that this trace will provide data for a future model transformation learning technique. We first address the trace recovery problem with examples coming from a transformation program. We propose an approach, based on a multi-objective meta-heuristic, to generate the textit{many-to-many} mapping between model constructs which correspond to a trace. The fitness functions rely on the lexical and structure similarity between the constructs. We also refine the approach to apply it to the more general problem of model matching.