Table of Contents
List of Figures
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.
The following animation shows the expecting results of this tutorial: according to the fired transitions, tokens move from place to place.
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 Marked Graph models.
Download the provided archive containing the projects and unzip it. Then, with the GEMOC Studio, select import… / General / Existing Projects into Workspace and import all the projects from the archive file.
Here is the archive containing all the projects that will be created during this tutorial.
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 :
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 (using a graphical syntax defined using 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).
First, check that the current perspective is "xDSML" (top right of Eclipse frame). The perspective can be changed thanks to the menu "Window > Open Perspective > Others".
Create a "New xDSML project" (New > Project > GEMOC Project / new GEMOC Concurrent 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 properties of the language (markedgraph). The 'Package' property is already defined (based on the project name). The othe properties include the name of melange file (used to link together the different parts on an xDSML), the name of the language and the location of the Ecore file. The two other properties 'DSA projects' and 'ECL path' will be created latter in this tutorial.
Use 'markedgraph' for the Melange file name, 'MarkedGraphL' for the Language name and select the markedgraph.ecore file by Browsing the 'org.gemoc.sample.markedgraph.model' project (found in 'model' directory).
The Melange file summarizes all the important resources used in an xDSML project (which are part of and managed by other projects). Completion (CTRL-SPACE) can be used to launch wizard but it will not be used in this tutorial. The Melange file is found in the 'org.gemoc.sample.markedgraph.xdsml' project in the 'src' directory.
A transition can be fired if there is at least one token in every of each input place. When a transition is 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:
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 :
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 step replaced by the use of Melange to automatically extend the AS with ED and EF.
The ED and EF are already defined on the provided metamodel. Thus, there is no need to add the 'runtimeTokenCount' ED on Place, 'fire()' on Transition and 'initialize()' on MarkedGraph.
Figure 4.1. Abstract Syntax of MarkedGraph extended with Execution Data (ED) and Execution Functions (EF)
In the Project Explorer, right click on any file of the xDSML project and choose GEMOC Languge > Create DSA Project for language to start the wizard.
The wizard first add for the language: select 'MarkedGraphL' and click OK. On the next screen, the project name is already initialized. Click 'Finish' and the 'org.gemoc.sample.markedgraph.k3dsa' is created and the Melange file is completed with references to aspects implemented the ED and EF.
Open the markedgraphAspects.xtend file. It is located in the org.gemoc.sample.markedgraph.k3dsa project in src/markedgraph.aspects. The file has been initialized with a template that can be discarded and replaced with the following text.
package markedgraph.aspects import fr.inria.diverse.k3.al.annotationprocessor.Aspect import markedgraph.MarkedGraph import markedgraph.Place import markedgraph.Transition import static extension markedgraph.aspects.PlaceAspect.* @Aspect(className=MarkedGraph) class MarkedGraphAspect { def public void initialize() { println("Graph " + _self.name + " initialized.") _self.places.forEach [ pl | pl.initialize ] } } @Aspect(className=Place) class PlaceAspect { public int runtimeTokenCount // @Helper def public void initialize() { println("Place " + _self.name + ": initialized.") _self.runtimeTokenCount = _self.tokenCount } } @Aspect(className=Transition) class TransitionAspect { def public void fire() { println("Transition " + _self.name + ": fired.") _self.inputs.forEach[ runtimeTokenCount = runtimeTokenCount - 1 ] _self.outputs.forEach[ runtimeTokenCount = runtimeTokenCount + 1 ] } }
There is an error in the Melange file because no aspect has been defined on the NamedElement metaclass. Just comment out the corresponding line.
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 define when they may occur.
Like for the DSA Project, right click on any file of the xDSML project and choose GEMOC Languge > Create DSE Project for language to start the wizard.
Then :
On the last screen, complete the required fields:
Once the wizard has been completed, the org.gemoc.sample.markedgraph.dse project is created with a template for the MarkedGraphL.ecl file (found in the ecl directory).
Replace the content of the ECL file with the following code (explanations on this code are given bellow):
The file MarkedGraphL.ecl.
import 'platform:/resource/org.gemoc.sample.markedgraph.model/model/markedgraph.ecore' ECLimport "platform:/plugin/fr.inria.aoste.timesquare.ccslkernel.model/ccsllibrary/kernel.ccslLib" ECLimport "platform:/plugin/fr.inria.aoste.timesquare.ccslkernel.model/ccsllibrary/CCSL.ccslLib" package markedgraph context MarkedGraph def: initIt: Event = self.initialize() context Transition def: fireIt: Event = self.fire() context Place inv tokenCountIsNull: (self.tokenCount = 0) implies (Relation Precedes(self.input.fireIt, self.output.fireIt)) inv tokenCountIsNotNullDelayRequired: (self.tokenCount > 0) implies let delay: Integer = self.tokenCount in let outputDelayed: Event = Expression DelayFor(self.output.fireIt, self.output.fireIt, delay) in Relation Precedes(self.input.fireIt, outputDelayed) context MarkedGraph inv initFirst: let firstInit : Event = Expression OneTickAndNoMore(self.initIt) in let allFire : Event = Expression Union(self.transitions.fireIt) in let firstOfAllFire : Event = Expression OneTickAndNoMore(allFire) in Relation Precedes(firstInit, firstOfAllFire) inv onlyOneInit: let firstInit2 : Event = Expression OneTickAndNoMore(self.initIt) in Relation Coincides(self.initIt, firstInit2) endpackage
This step has three main purposes:
First, it specifies DSE in the context of metaclasses of the AS. For Marked Graph xDSML, we identify 2 DSE:
Then, it links them to EF from DSA --- when a DSE will occur 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 MoCC libraries. Here constraints are expressed in CCSL and only relies on relations and expression from the CCSL core library.
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. For example, there will be one fireIt MSE for each Transition element of MarkedGraph model. For the wikipedia example, there will a fireIt event for transitions t1, t2, t3 and t4. In the same way, constraints apply to the MSE.
The executable MarkedGraph Language is now defined. We can use the GEMOC Modeling Workbench to execute MarkedGraph models. In the next section, we will see how to define a graphical animation a MarkedGraph model.
First, we will create a run configuration. Select "run / run Configurations".
Double click on "Eclipse Application" and change the name "New Configuration" into "MarkedGraph Modeling Workbench".
Then click "Run" to start the new runtime Eclipse which indeed corresponds to the Modeling Workbench for MarkedGraph.
Import the modeling project org.gemoc.sample.markedgraph.sample in the Modeling Workbench (Import / General / Existing project into Workspace). It contains the wikipedia.markedgraph file which corresponds to the Wikipedia example.
To open the graphical visualization of the wikipedia MarkedGraph example, double-click on wikipedia.aird resource, then open the tree view until "wikipedia MarkeGraph diagram" appears. Open it.
The properties view (bottom right) allow to see the properties on an element. We can see that a place object has the "Runtime Token Count" properties (the added ED).
We can now run the "gemoc" configuration.
It selects the different views related to the execution on an xDSML model.
As we have not yet defined any animation for the MarkedGraph Language, there will be no animation and the graphical visualization will not change.
The Logical Steps view, presents the possible steps, and for each step, the MSE that will occur if it is selected.
Double-click on a Logical Step to choose it. The corresponding MSEs occur and trigger the associated execution functions (EF) which make the state of the model evolve. Here, we will see traces in the Console are the EF we have defined print messages. The first MSE was initIt, which init the Graph and all its places (by initializing the runtime token count with the token count). The is then only one possible step: fireIt applied on t1 (__MSE_ta_fireIt).
Then, we can select a new logical step and so one. In the following screenshots, we have first selected the t1 transition, and then the t2 transition.
To stop the execution, simply click on the red button in the Logical Steps Decider view (or in the Gemoc Engines view).
During the execution of a Marked Graph model, we want to visualize :
As the animation view is close to the graphical concrete syntax, we extend the existing diagram description.
First, we will define in the Language Workbench a graphical representation based on the graphical Marked Graph syntax thanks to the wizard called "Create Animator Project for language", in whichwe will add a layer to describe the graphical animation of a Marked Graph model.
Then we will be to able to visualize the states of mobel being executed in the MarkedGraph Modeling Workbench.
Keep the MarkedGraph Modeling Workbenchrunning and come back to the Language Workbench (first Gemoc Studio).
Rigth-click on any file of the org.gemoc.sample.markedgraph.xdsml project and select GEMOC Language > Create Animator Project for language.
On the first screen of the wizad, select "Extends an existing diagram description".
Select the viewpoint to extend "MarkedGraph diagram".
Finally, we can fill in the Project Name in which the newly creating viewpoint file will take place (org.gemoc.sample.markedgraph.animation), the name of the viewpoint file (markedgraph-animation.odesign), the viewpoint name (MarkedGraphAnimationViewpoint) and the diagram name (MarkedGraphAnimation).
The last screen allows to choose the name of the layer to be created. Debug is a good name.
Press Finish to create the project. The markedgraph-animation.odesign file is opened.
Open the markedgraph-animation.odesign file and unfold its content.
To be able to complete the definition of the odesign file, we first need to load the existing markedgraph.odesign as a resource :
After the debug presentation has been defined, we can complete it to add an animation layer.
On "Style Customization", set the "Predicate Expression" to:
[self.eGet('inputs')->forAll(p | p.eGet('runtimeTokenCount').toString().toInteger() > 0) /]
Perform the same action as above to set the background color of places to yellow.
You have to select the first element named "ecliple white" (the second one is used to diplay the name).
In the project, org.gemoc.sample.markedgraph.animation project, in "src/org.gemoc.sample.markedgraph.animation.services", copy the MarkedgraphDebugServices.java file to MarkedgraphAnimationServices.java, and edit the file to
The executable MarkedGraph Language is now defined with a graphical visualization. We can come back to the MarkedGraph Modeling Workbench to run it.
First of all, we need to restart (File > restart) the MarkedGraph Modeling Workbench to take into account the modifications of the MarkedGraph xDSML.
Right click on wikipedia.aird and select Viewpoints Selection. If Viewpoints Selection is not in the menu, double-click on the file wikipedia.aird.
Select Viewpoint Selection.
Select MarkedGraphAnimationViewpoint.
You can now open the wikipedia graphical representation (wikipedia.air / Representation per category / markedgraph / MarkedGraph diagram / wikipedia MarkedGraph diagram). On the visualization, all places should be with a yellow background.
Open the "run wikipedia example" configuration" (Debug > Debug Configuration…). Complete the "Animator" field with "/org.gemoc.sample.markedgraph.sample/wikipedia.aird" (use Browse to select it).
First, we will create a run configuration. Select "run / run Configurations"
To be able to debug a model, we need to complete the org.gemoc.sample.markedgraph.animation project using the Language Workbench.
The "Debug" layer which has been generated has to be completed and corrected. A future version of the GEMOC Studio will tackle down the burdon.
On the "Mapping Based Decoration Enabled breakpoint" and "Mapping Based Decoration Disabled breakpoint" of the "Debug" layer, under "Decorations" element, select the "Place" and "Transition" element. Change "Position" to "North East".
It indicates that a breakpoint can be enabled or disabled on Place and Transition elements of a MarkedGraph model. It will be graphically shown as icon displayed in the corresponding graphical element, at the nort east.
In both mappings, change the directory of the icon from "org.gemoc.executionframework.extensions.sirius" to "org.gemoc.gemoc_heterogeneous_modeling_workbench.ui". Do the same change for the icons of "Operation Action Debug" and "Operation Action Toggle breakpoint".
@Override public String getModelIdentifier() { return org.gemoc.execution.concurrent.ccsljavaengine.ui.Activator.PLUGIN_ID+".debugModel"; }
To improve readability of the ECL file and also to favor capitalization and reuse of MoCC elements, it is possible (and encouraged) to define libraries. Those libraries looks like the predefined ones that we have already used like kernel.lib and CCSL.lib.
Using the GEMOC Studio, we start by defining a MoCCML project. In the xDSML view, we select Behavioral definition / MoCC definition library to create this new project.
We can then define in this project the following markedgraph.ccslLib file. TODO: To be corrected.
The file markedgraph.ccslLib defining a MoCCML library.
Unresolved directive in GuideTutorialMarkedGraph.asciidoc - include::MarkedGraph/markedgraph.ccslLib[Definition of a MoCCML library]
To use this new library, we only have to import it at the beginning of the ECL file and the we can use it as kernel.lib and CCSL.lib libraries.
Here is the new ECL file.
The file MarkedGraphL.ecl using the MoCCML library.
import 'platform:/resource/org.gemoc.sample.markedgraph.model/model/markedgraph.ecore' ECLimport "platform:/plugin/fr.inria.aoste.timesquare.ccslkernel.model/ccsllibrary/kernel.ccslLib" ECLimport "platform:/plugin/fr.inria.aoste.timesquare.ccslkernel.model/ccsllibrary/CCSL.ccslLib" ECLimport "platform:/resource/org.gemoc.sample.markedgraph.mocc/mocc/markedgraph.moccml" package markedgraph context MarkedGraph def: initIt: Event = self.initialize() context Transition def: fireIt: Event = self.fire() context Place inv placeReadAndWrite: let delay: Integer = self.tokenCount in Relation placeReadWrite(self.output.fireIt, self.input.fireIt, delay) context MarkedGraph inv initFirst: let firstInit : Event = Expression OneTickAndNoMore(self.initIt) in let allFire : Event = Expression Union(self.transitions.fireIt) in let firstOfAllFire : Event = Expression OneTickAndNoMore(allFire) in Relation Precedes(firstInit, firstOfAllFire) inv onlyOneInit: let firstInit2 : Event = Expression OneTickAndNoMore(self.initIt) in Relation Coincides(self.initIt, firstInit2) endpackage
Sometimes, it may be easier to define the MoCC using automata. For example, we can see a place as something on which we can read a token (ont output transition reads one token) or we can write a token (one input transition writes one token). A token can only be read if there is at least one in the place. Initially, there is as much token in the place as in the initial token count of the place.
For explanations on how to use the MoCCML editor, please refer to the GEMOC manual.
We first define the declaration of a relation (placeReadWrite). Here after is its graphical representation. This relation is parameterized by the r clock (some one how wants to read a token), the w clock (someone how wants to write a token, and n the initial count of token in the place).
Then we can define this relation. Here after is the resulting automaton. The local variable count represent the number of tokens in this place. s0 is the initial state. The action on the transition from s0 to s initialize count with n (the parameter). The two reflexive transitions on s explain what happens when w or r occurs. The write transition is fired when w occurs and a token is assed in the place (count is incremented). The read transition is fired when r occurs. It removes a token and thus decrements count. A guard prevents the read transition to be fired if there is not at liste one token in the place (count >= one).
Here is the textual representation of this constraint automata.
Textual representation of the constraint automata.
AutomataConstraintLibrary markedgraph{ import "platform:/plugin/fr.inria.aoste.timesquare.ccslkernel.model/ccsllibrary/kernel.ccslLib" as kernel; import "platform:/plugin/fr.inria.aoste.timesquare.ccslkernel.model/ccsllibrary/CCSL.ccslLib" as ccsl; RelationLibrary placeLib { AutomataRelationDefinition placeReadWriteDef[placeReadWrite]{ variables { Integer MAX = 10 Integer count = 0 } init: s0 from s0 to s : init -> (do count = n.value) from s to s : read -> ( when r if( count >= one) do count = (count - one)) from s to s : write -> ( when w do count = (count + one)) State s0 (out : init) State s( in : read, write, init out : write, read ) } RelationDeclaration placeReadWrite(r:clock, w: clock, n: int) } }