-
Notifications
You must be signed in to change notification settings - Fork 4
GuideTutorialMarkedGraph
TODO: Provide .model, .model.edit, .model.editor.
This tutorial aims at demonstrating the power of the GEMOC studio to define an executable semantics and provide graphical animation for a DSML. It relies of the marked graph language.
TODO: Show an animated GIF or a video.
|
Note
|
A screencast of this tutorial can be found here (give URL when it will be available)… |
Please, read the introduction of tutorials to understand the GEMOC approach to define executable DSML or to have a better understanding of the abbreviations used in this tutorial.
The domain model is implemented with several EMF projects that you need to import into your workspace. A graphical editor defined with Sirius is also available to visualize and edit Marked Graph models.
Marked Graph is a kind of Petri net in which every place has exactly one incoming arc and exactly one outgoing arc. As a consequence, it a concurrent language (several transitions may be fired) and has no conflict.
The Domain Model, also called Abstract Syntax or Metamodel, defines :
-
A marked graph as a set of places and transitions.
-
Each Place has exactly one input transition and one output transition and a token count.
-
A transition has several input places and several output places.
The graphical concrete syntax draws places as circles and transitions as squares. Inputs and outputs of places and transitions are designated by arrows. The following picture shows the graphical representation of the Marked Graph model of wikipedia.
Here the graphical syntax of the wikipedia example using the view defined with Sirius.
In this section, we first create an xDSML project for MarkedGraph and initialize it with the provided Abstract Syntax (AS) and Concrete Syntax (CS).
Select an xDSML project (New > Project > GEMOC Project / new xDSML Project).
The first dialog of the wizard asks for the name of the project. Define it as org.gemoc.sample.markedgraph.xdsml.
Click on Next and define the name of the language (markedgraph).
A file project.xml has been created in the org.gemoc.sample.markedgraph.xdsml project.
When opened, it provides the xDSML view which summarizes all the important resources used in an xDSML project (which are part of and managed by other projects). This view is a kind of control center to have quick access to the main resources of the project.
In the "Domain Model" section, click on the "Browse" button to select the project defining the AS: org.gemoc.sample.markedgraph.model.
Then, select the "Genmodel URI".
Finally, select the "Root container model element" thanks to the "Select" buttons.
The Graphical Editor defines a graphical concrete syntax which is user-friendly to view and edit a model.
In the "Concrete syntax definition / Graphical editor" of the project.xdsml editor, click on "Browse" to select the "org.gemoc.markedgraph.design" project.
A transition can be fired if there is at least one token in every of each input place. When fired, one token is removed from each of its input places and one token is added to each of its output places. Several transitions can be fired as the same time.
Defining the execution semantics consists in implementing the previous behavior. In the GEMOC approach, it is split in different concerns:
-
The definition of Execution Data (ED) like the runtime count of tokens in a place and Execution Functions (EF) like fire a transition. ED and EF constitute the DSA.
-
The definition of the model of concurrency as a set of events and constraints on these events. It is the MoCC concern that is defined in a DSE project (using ECL, Event Constraint Language) possibly completed with MoCCML projects to define libraries of constraints.
-
The mapping between the DSA and the MoCC.
In the current version of the GEMOC studio, the MoCC and the mapping are tightly coupled and described in ECL (Event Constraint Language).
During execution of a MarkedGraph, the number of tokens of a place has to be recorded and changed according to the fired transitions. Thus, we have to manage an execution data (ED) called runtimeTokenCount and an execution function (EF) on Transition called fire(). Furthermore, the runtimeTokenCount of each place must be initialized at the start of the execution. It is the purpose of the EF called initialize() on the MarkedGraph element.
The DSA of Marked Graph is composed of :
-
one ED called runtimeTokenCount defined on Place . It represents the number of token in a place when the model is executed.
-
one EF called initialize() defined on MarkedGraph. It initializes the runtime token count of each place with the initial token count.
-
one ED called fire() on Transition. It to remove one token from each of its input places and add one token to all its output places.
At the moment, we need to complete the AS (markedgraph.ecore) with the ED and EF. In the next release of the GEMOC Studio this steps will disappear and extended the AS will be done automatically thanks to Melange
Add the 'runtimeTokenCount' ED on Place, 'fire()' on Transition and 'initialize()' on MarkedGraph.
Click on K3 project in the project.xdml editor (Behavioral definition / DSA definition). The wizard to create of new Kermeta 3 project is launched with the name of the project initialized (k3dsa is the last name). Click "Finish". The project has been created.
The purpose of the DSE project is to define events (called DSE) on AS elements that will trigger EF calls when they occurs. Furthermore, constraints can be defined on these events to ensure they occur in the right order.
To create the DSE Project, click on the "ECL Project" in the project.xdsml editor ("Behavior definition / DSE definition"). Check that the name is org.gemoc.sample.markedgraph.dse.
Click "finish" and the project is created.
Click on "ECL Project" to edit the "markedgraph.ecl" file and replace its content with the following:
This step has three main purposes:
-
First, it specifies DSE in the context of metaclasses of the AS.
-
Then, it links them to EF form DSA --- when a DSE will occurs the associated EF will be executed.
-
Finally, it defines constraints on the DSE to rule the possible scheduling. Constraints generally rely on relations which are defined in the MoCC.
|
Note
|
Please notice that, as often, DSE are defined at the language level, but at runtime they are instantiated as MSE on each object instance of the metaclasse they are defined on. In the same way, constraints apply to the MSE. |
Replace the content of the ECL file with the following code:
link:MarkedGraph/markedgraph.ecl[role=include]
TODO: Give some rationals on the relation and their instancitions. In particular the delay.
What should be explained?
-
The main principles (we can fire as much transition has the initial token count of a place, thus a delayed is introduce)
-
We want to ensure that one init is done before any fire (and only once)
The executable MarkedGraph Language is now defined. We can use the GEMOC Modeling Workbench to execute MarkedGraph models.
Launch the Modeling Workbench by running a new Eclipse: right click, select "run as / Eclipse Application". The new eclipse is the Modeling Workbench.
Import the modeling project markedgraph.sample in the Modeling Workbench (Import / General / Existing project into Workspace).
Create a Run Configuration: right click on the model and select "Run As… > Run Configurations". Create a new "Gemoc eXecutable Model" configuration. We can call it "gemoc" and fill in the fields: Model to execute: "first.automata", xDSML: "markedgraph".
Change the "Decider" to "Step by step user decider" (the user will decide which is the next step to execute).
We can now run the "gemoc" configuration. The graphical representation is opened, the VCD is created, etc.
TODO: Give some explanations.
Questions on the MoCC from summerSchool :
-
why delayedBy p.initialMarking and not delayedBy 1? ⇒ Because we can make as much read as there is initialMarking token in the place. Any write in that place, means that a new read can be done.
-
is it still right if the initial marking is P2 = 1 (all other places is 0) ? ⇒ No. But in any case, this initialMarking will result in a deadlock.
-
what if there is several places which are initially marked? ⇒ See read/write semantics
Questions on the implementation :
-
A transition has no more that one input or output place. Should be any number !
Questions:
-
use execute for the EF on transition instead of fire and use fire (or trigger) for the DSE instead of fireCall.
Questions on the form :
-
Use of abbreviations like MoCC, DSA, EF, ED, AS, CS, etc.? Yes because using the full name will make the tutorial hard to read. ⇒ Explanations of acronyms must be provided somewhere.
-
Avoid to split files in several parts? For example ED and EF are defined in the same files. It is thus useless to try to have two steps, one to define the ED and the other to define EF.
-
How to name the project defining the abstract syntax? model or as? ⇒ If 'as' or 'metamodel' is used, then the wizard should be changed in order to avoid to have the language named 'as' or 'metamodel'.
TODO :
-
Remove "example" in the name of the project?
-
DONE: Domain model and then MetaModel (Abstract Syntax)
-
Give insight on the debugging, tricks to resolve problms, etC. ⇒ A kind of FAQ must be provided.
-
Look at the Mélanie’s Dashboard so that the steps in the tutorial corresponds to activities in the dashboard.
-
Look at the tools guide written by Didier, Florent et Issa.
Tips and tricks :
-
When the Ecore is changed to add new operations (and the genmodel is used to regenerate the projects), the ECL editor keep signaling errors concerning the unresolved operations. The solution consists in closing the editor and opening again the ECL file.
-
xDSML Editor: It is sometimes required to save the file (for example when the AS has been selected in order to create the k3dsa project).
-
ECL Editor: If Precedes is not found, quit Eclipse and launch it again. It should solve the problem!