Designing, Animating, and Verifying Partial UML Models

layout: true

<div style="text-align: center;position: absolute; bottom: 15px; width: 85%; font-size: 24px">
.hiddenlink[[MODELS 2020](http://modelsconference.org/)] — 
Virtual Event — 
October 16-23,
2020</div>

---
class: center, middle

# .hiddenlink[[Designing, Animating, and Verifying Partial UML Models](https://conf.researchr.org/details/models-2020/models-2020-technical-track/35/Designing-Animating-and-Verifying-Partial-UML-Models)]

### Frédéric Jouault<sup>1</sup>, Valentin Besnard<sup>1</sup>, Théo Le Calvar<sup>2</sup>,
### Ciprian Teodorov<sup>3</sup>, Matthias Brun<sup>1</sup>, and Jerome Delatour<sup>1</sup>

<sup>1</sup>ERIS, ESEO-TECH,
Angers, France,
*firstname*.*lastname*@eseo.fr

<sup>2</sup>LERIA, Université d’Angers,
Angers, France,
*firstname*.*lastname*@univ-angers.fr

<sup>3</sup>Lab-STICC UMR CNRS 6285, ENSTA Bretagne,
Brest, France,

*firstname*.*lastname*@ensta-bretagne.fr

---
class: bigger

# Context and Problem

* Early model validation is useful to fix design mistakes before they become costly
--
count: false
* This requires models to be executable in some way
	* In particular, a significant subset of UML has a standardized precise execution semantics (with [fUML](https://www.omg.org/spec/FUML), [PSCS](https://www.omg.org/spec/PSCS), and [PSSM](https://www.omg.org/spec/PSSM))
--
count: false
* However, most existing approaches typically only work with relatively consistent and complete models
	* Reaching this level of consistency and completeness is generally hard
--
count: false
* How could we help modelers with their inconsistent and incomplete models?

---
class: bigger

# Approach Overview

* Objectives
	* Making models **executable as early as possible**
	* Instead of only relying on typically inconsistent mental simulations
--
count: false
* Means: making partial models executable by **relaxing modeling language compliance**
--
count: false
* Rationale
	* Makes it possible to **incrementally improve testable models**
		* Until they become complete and consistent (standard compliant, verifiable)
	* While providing designers with **tool support** during the whole process
--
count: false
* Coming next: application to UML

---

# Relaxation Point: Instances Before Classifiers

---

# Making it Easier by Borrowing from Communication Diagrams

<div style="position: absolute; right: 10%; top: 20%">
.bigger[
(and showing [clickable](javascript:alert("You may want to click on diagram elements, not on this link style example."%29) elements)
]
</div>

---

# More Relaxation Points

* Receiving undeclared messages
* Time triggers may always match

---

# Relaxation Point: Message Triggers May Find Messages

---

# Relaxing Actions and Expressions

* They may be non-executable
* All guards may be considered true

---

# Relaxing Inter-object Communication

* Messages may transit through the *ether* instead of via event pools

???

Shortcuts:
* s: switch to TCSVG sequence diagram
* p: switch to PlantUML sequence diagram
* t: switch to PlantUML timing diagram

---

# More Features: Methods & Available State Machine Actions

---

# Advanced Usage Example: Verification

---

# Tool Support

* Approach implemented as a web application
	* https://github.com/fjouault/AnimUML
		* Includes a link to a public server to try without installing
--
count: false
* Usable to create interactive documents
	* Including interactive slide decks like this very presentation
--
count: false
* Some features
	* Share model by link (with or without current execution trace)
	* Export diagrams (e.g., as PlantUML code, SVG)
	* Generate code
		* To tUML, which can be imported into Eclipse UML (notably for use in Papyrus)
		* To other languages (work in progress)
	* Import from Eclipse UML
	* Analyse models directly in-browser
	* Connect to more powerful remote analysis engines (http://www.obpcdl.org/)
	* Connect to remote execution engines (http://www.obpcdl.org/bare-metal-uml/)
	* Node.js compatible (e.g., to create command line tools)

---
class: bigger

# Conclusion

* AnimUML: a tool-supported approach for early model execution
	* Model animation, testing, verification
--
count: false
* Animating models is actually useful to discover inconsistencies, or missing elements
--
count: false
* Some possible perspectives
	* Semi-automatic inference of a model's missing parts
--
count: false
	* Application to teaching
		* Teachers can create interactive AnimUML tutorials
		* Students can play with models (modify, execute, repeat)
			* To gain insight into notation and semantics
		* (Semi-)automatic grading with analysis tool?
--
count: false
	* Application to SysML v2?

---

# Questions?

</iframe>