[Modelinterpreter] MODELS 2015 konferencia

[Dévai Gergely]

MODELS 2015

09-27

Executable Modeling Workshop

Francis Bordeleau: Why and Where do we need Model Execution?
- "pragmatic, not dogmatic": business cases and showing that modeling brings values is important
- going from PowerPoint to Papyrus-based DSML for network modeling
- software modeling: "might be enough to generate skeletons"
- software design modeling: ObjectTime and xtUML/BridgePoint mentioned
- functional modeling: Matlab, OpenModelica
- requirements:
  - language with executable semantics
  - action language
  - execution (generation or interpretation)
  - tooling (tracing, debugging, both "on host" and "on target")
- reference:
http://modeling-languages.com/executable-models-vs-code-generation-vs-model-interpretation-2/
- tooling is named a key requirement
- the talk was quite general, no technical details

Arnaud Cuccuru: Executable Modeling with fUML and Alf in Papyrus: Tooling and Experiments
- controlling the execution because of debugging: call out to control code whenever a token is passed in the activity
- using Alf fragments in a graphical model would be easier with more modular grammar
- work in progress FMI integration

Gergely Dévai: UML Model Execution via Code Generation
- Questions/comments:
    - Is the translation time calculated in the runtime of the 'code generation' solution?
    - What is the reason for convergence of the runtime in the two JIT cases?
    - How is incremental translation solved?
    - Instead of placing Java breakpoints we could communicate to the runtime module what are we interested in and at those points it could send information to the GUI.
    - Instance-level breakpoints would be better than class-level breakpoints.
    - Arnaud has no idea how to have separate Papyrus diagrams for two instances of the same class. He likes the debug view showing the instances, would like to have that in Papyrus.

Hamoud Aljamaan: UmpleRun: a Dynamic Analysis Tool for Textually Modeled State Machines using Umple
- trace specification defined independently of the model
- UmpleRun:
  - based on Java code generation and there is a validator module that provides validation verdicts
  - execution specification describes instance creation, signals sent and also the required outcome (eg. state to be in, or attribute value)
- UmpleRun is in progress:   
  - for example association support is not yet in place
  - no animation or debugging support is in focus

Colin Atkinson: On the Real-time Execution of Deep Models
- many meta modelling layers used to control execution
- www.melanee.org

Domenik Pavletic: Testing Extensible Language Debuggers
- mbeddr
- they write the mapping from the language extensions to the base language and the debugging support for the extension separately
- result: it gets out of synch, so they created a language extension to test debug support... :)
- no incremental support yet

Discussion session:
- Francis: let's go to business process modeling and system modeling first; software modeling in not realistic in two years
- Everyone: tooling issue is the most important
- tools state of the art: BridgePoint, Rhapsody, RSA-RTE, Papyrus/Moka, MagicDrow/Cameo simulation toolkit, Enterprise Architect/LieberLieber AMUSE
- No real commertial implementation of Alf known
- OMG has no means to check conformance to its starndards
- Semantics of profiles need to be made precise
 
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://plc.inf.elte.hu/pipermail/modelinterpreter/attachments/20150928/568d1751/attachment.html>