| 12345678910111213141516171819202122232425262728293031323334353637383940414243444546474849505152535455565758596061626364656667 |
- \chapter{Methodology}
- \label{chapt:Methodology}
- The methodology employed in this thesis outlines the systematic approach taken to translate SCCD into PyDEVS. This section provides an overview of the strategies, tools, and processes utilized to achieve the translation objectives.
- The translation involves converting models represented in one formalism into another while preserving their essential behavioral characteristics. Such translation is critical for interoperability, model verification, and simulation
- interoperability between systems designed using different modeling paradigms.
- We also present the overarching methodology for translating SCCD to PyDEVS, detailing the steps involved, challenges encountered, and the rationale behind the chosen approaches. Additionally, we discuss alternative
- approaches considered during the development process, providing insights into why certain strategies were pursued while others were discarded.
- \section{Translation Tools}
- The translation of SCCD to DEVS involves the use of specialized tools and techniques to facilitate the conversion process. In this section, we discuss the evolution of the translation tools employed in this research, from initial
- parsing methods to the adoption of a visitor pattern for seamless translation.
- \subsection{Parsing line by line}
- Initially, in the endeavor to translate SCCD to DEVS, the SCXML file was parsed meticulously, line by line. However, as we delved deeper into the process, we encountered challenges that made this approach increasingly cumbersome.
- The complexity of the SCXML structure made it difficult to extract the necessary information efficiently, leading to inefficiencies and potential errors in the translation process. Recognizing the need for a more robust and
- structured approach, we sought alternative solutions to streamline our workflow.
- \subsection{Visitor Pattern}
- Upon further investigation, it was discovered that the SCCD compiler already integrates a robust visitor pattern for translating SCXML files into Python and/or Java representations. This discovery marks a significant breakthrough
- in the translation process, as it not only facilitates the traversal of intricate data structures without altering the structure itself but also enables seamless integration of new languages or platforms.
- The compiler generates an Abstract Syntax Tree (AST), which serves as the foundation for the visitor to traverse each node systematically. This thesis elucidates the profound utility of this approach. Firstly, the visitor parses
- the tree from the root to the leaf nodes, thereby enabling a chronological description of the output. This sequential traversal allows for the orderly declaration of imports and their subsequent usage, streamlining the code
- generation process. Furthermore, by compartmentalizing each node, the visitor pattern facilitates the generation of code independent of other nodes, enhancing the modularity and efficiency of the SCCD framework.
- Moreover, the visitor pattern's versatility extends to accommodating various platforms, including Thread, UI Event Loop, and Game Loop, for both Python and Java. This inherent adaptability underscores the feasibility of integrating
- the DEVS formalism seamlessly into the SCCD framework. By developing a visitor tailored for DEVS and integrating it as a platform, extending the SCCD compiler's capabilities becomes a straightforward endeavor.
- \section{Considered Translation Approaches}
- \subsection{First approach: Implementing everything into one AtomicDEVS}
- The first approach is to set the entire SCCD code, which includes the controller, ObjectManager and every class, to one AtomicDEVS model. While this could be possible, it was chosen not to do this for several reasons.
- \subsection{Second approach: Setting every component to AtomicDEVS}
- Another approach is to set every possible component (Controller, ObjectManager and classes) to an AtomicDEVS model. This proved to be impossible with the standard
- atomicDEVS because in SCCD, statecharts can be created at runtime. The classic AtomicDEVS does not allow to create AtomicDEVS objects at runtime and thus mapping statecharts
- to a corresponding AtomicDEVS is impossible.
- \subsection{Chosen approach: TODO}
- The approach that is chosen is to have a Controller component as a CoupledDEVS object and the ObjectManager and class type as an AtomicDEVS object. What is different with the approach explained in \ref{TODO} is that
- the class type is an AtomicDEVS an not every instance of that type.
- \subsubsection{Controller component}
- This component originated from the SCCD runner. While the controller in the SCCD runner is created to control the runtime of the objects, the controller created in PyDEVS serves merely as a structural object. In the
- controller the submodels and their couplings are described. This is perfect because the classes in SCCD need to be coupled to each other and the ObjectManager. To couple them, the SCCD to PyDEVS compiler will generate
- a controller and link every AtomicDEVS to the AtomicDEVS of the ObjectManager.
- \subsubsection{ObjectManager component}
- The ObjectManager has the same use cases as the ObjectManager in SCCD. The only difference is that objects can not explicitely be created in this component but rather the Event to do an action gets now forwarded do the
- specific class component an instance should be created, deleted from. This way the ObjectManager itself still decides if the instance can/may be made or not. Instances can still send events to the ObjectManager with the
- \"cd\" scope.
- When the application is started, the default class
- the object manager now thus not create an instance of the default class and starts its associated Statecharts model. But it sends a \"start\_instance\" Event to the default class allowing
- the default class to start the instance that is already created in the
- From then on, instances can send several events to the object manager to control the set of currently executing objects. The object manager accepts four events. We list them below, including the parameters that have to be sent as part of the event:
- \subsubsection{Class component}
- \subsubsection{Class Instance component}
|