< To: ATL
This page contains a list of problems related to ATL usage. It gives hints towards solutions and links to examples.
How do I generate text from models?
There are several possibilies. We only detail some of them here.
You can specify a query over your model in ATL. An example of this is available in the UML2Java transformation.
The atlanticRaster zoo is also using this technique. It first transforms KM3 metamodels into DOT models with the KM32DOT.atl transformation. It then uses DOT2Text.atl to query a String from the DOT model. This string corresponds to the generated DOT code. Then, the build.xml ant scripts makes use of the am3.saveModel task with an ATL extractor pointing to the DOT2Text.atl transformation.
TCS (Textual Concrete Syntax) is a tool enabling the specification of textual syntaxes for metamodels. Once such a syntax has been specified, models can be serialized to text files.
One advantage of TCS is that text can also be parsed into models. It is therefore possible to use the same specification to enable the definition of models with any text editor.
How do I launch transformations programmatically?
The ATL development environment (ATL Development Tools, or ADT) provides several means to launch transformations:
- by specifying a launch configuration,
- by writing an ant script using the AM3 Ant Tasks.
However, in some cases, it is necessary to launch an ATL transformation programmatically (e.g., from Java code).
The AtlLauncher class can be used for this purpose. Complete Milan Milanovic's ATL template bundle is basic solution for integration of ATL engine and for transformation execution in Java applications.
Remark 1: the AtlCompiler class may be used to compile ATL transformations into .asm files executable on the ATL Virtual Machine.
Remark 2: There may be problems when using MDR as model handler, as it shown in KM3Projector class. Call to initMDR() method should be done only once when your class is constructed. This is because if you load your model in this way: getMDRModelHandler().loadModel("model.xmi", getSomeMetamodel(), mod.openStream()); method getSomeMetamodel() will return reference to metamodel which is initialized in initMDR(). If some of your methods later call initMDR() your initial metamodel reference will be lost and someMetamodel variable will get reference to new metamodel. Result of your transformation will be output model file, because input model has lost his metamodel reference.
Libraries and Prequesites
To launch ATL transformations programmaticallly from Java, you will need some libraries. These libraries should be copied from eclipse/plugins folder to folder where libs are stored for Java application, and with they you have complete support for MDR and EMF. They should be also added to the classpath. Files in case are of Eclipse 3.3.x and emf 2.1.0 version. If you have other versions, just copy the same files with those versions.
Here is a description of what is required for all kind of use :
- ATL libraries
- antlr-runtime-3.0.jar, ATL-parser.jar, org.eclipse.gmt.tcs.extractor_1.0.0.jar, org.eclipse.gmt.tcs.injector_1.0.0.jar : those are located on the CVS, here
- emf4atl.jar, vm.jar, engine.jar : those are not exported at this time, so you must get the last ATL version on the CVS and then export these plugins into jar libraries (right-click on the project-> export->JAR file). The real plugin names are org.eclipse.m2m.atl.drivers.emf4atl, org.eclipse.m2m.atl.engine.vm and org.eclipse.m2m.atl.engine.
- Eclipse libraries
- EMF libraries
ATL - MDR use
The following libraries are only required for the use of MDR.
ATL - KM3 use
The following libraries are only required for the use of KM3.
A very simple working example, involving the invocation of an ATL transformation from a plug-in can be seen at JWT to BPMN and vice versa, using regular .ecore metamodels.
Another illustration is given in Dennis Wagelaar's ATL command line tool (here is a direct link to the Java class). This tool was discussed on the ATL mailing list (see the "Related threads" section below).
Here is a list of discussion threads related to this issue that occured on the old ATL mailing list atl_discussion, or on the Eclipse M2M newsgroup:
- Run ATL-Transformation from command line
- Running ATL transformations programmatically [with MDR]!
- Running from command line
- Libraries and Prequesites to launch ATL transformations programmatically from Java: archive (read-only), newsportal (requires password)
- ATL transformations on UML2 models - including profiles - outside Eclipse
How do I use extern parameters in ATL transformations?
Using an XML file
A solution is to save transformation parameters in a extern XML model. In the transformation, you will use an helper to get parameters from your model. The wanted parameters model will be chosen in the transformation launch configuration or in the AM3 Ant Tasks.
The following code is a simple parameters XML file:
<parameters> <param name="source" value="KM3"/> <param name="target" value="ATL"/> </parameters>
In the transformation header, you add the parameters model and its metamodel XML.km3 :
module AMW2ATL; create OUT : ATL from source : MOF, target : MOF, parameters : XML;
From parameters model, the following helper is used to get value attribute select in the param entities according to the given name:
helper def : getParameter(name : String) : String = XML!Element.allInstancesFrom('parameters')->select(e | e.name = 'param' )->select(e | e.getAttrVal('name') = name )->first().getAttrVal('value');
To use the previous helper (due to the getAttrVal call), you need to use XMLHelpers.asm ATL Library and add in the transformation:
The following code is a simple example to show how to use the getParameter helper:
helper def : sourceValue : String = thisModule.getParameter('source');
<am3.loadModel modelHandler="EMF" name="KM3" metamodel="MOF" path="uri:http://www.eclipse.org/gmt/2005/KM3"/> <am3.loadModel modelHandler="EMF" name="XML" metamodel="KM3" path="XML.km3"> <injector name="ebnf"> <param name="name" value="KM3" /> </injector> </am3.loadModel> <am3.loadModel modelHandler="EMF" name="Parameters" metamodel="XML" path="Parameters.xml"> <injector name="xml"/> </am3.loadModel> <am3.atl path="myTransformation.atl"> <inModel name="source" model="left"/> <inModel name="target" model="right"/> <inModel name="MOF" model="%EMF"/> <inModel name="parameters" model="Parameters"/> <inModel name="XML" model="XML"/> <library name="XMLHelpers" path="XMLHelpers.asm"/> <inModel name="ATL" model="ATL"/> <outModel name="OUT" model="ATLmodel" metamodel="ATL"/> </am3.atl> <am3.saveModel model="ATLmodel" path="ATLmodel.ecore"/>
How can I specify a Virtual Machine operation in Java?
It is possible to add new Java operations to the ATL Virtual Machine (VM). These operations can then be called like helpers.
This is used in org.eclipse.gmt.atl.oclquery.core.OclEvaluator to redefine the dumpASM method of ASMEmitter. The process to register operations on OCL primitive types (e.g. Boolean, Integer, String) is similar.
How can I tune the XML output of ATL?
The ATL Virtual Machine can write target models to XML files in two different ways:
- XMI writing is delegated to the model handler used to handle the model (e.g., EMF, MDR).
It is usually not possible to change the target format in this case because XMI support in each model handler implements a precise mapping from a metamodel to an XML Schema. However, different model handlers may produce different XMI files (e.g., XMI 2.0 with EMF, XMI 1.2 with MDR).
- Target models can be extracted into XML files using custom XML Schemas.
To do this, the target model is first transformed into a model conforming to a specific XML metamodel. The resulting XML model can then be extracted into an XML document. An example can be found in the Table2SVGPieChart transformation scenario: the target SVG model is transformed into an XML model that can then be extracted (see documentation of scenario for more details and README.txt in zip file for directions). The process of transforming a model to XML then extracting it can be automated using ant tasks. This is also illustrated by the Table2SVGPieChart scenario (see build.xml in the zip file of Table2SVGPieChart for directions).
So to obtain a custom XML file (e.g. WS-BPEL, WSDL etc) you must write an ATL transformation from your target meta-model to an XML meta-model. The model outputted from this transformation can be serialised to custom XML using the AM3 "XML extractor" feature. Details of how to manually perform this function in your eclipse setup are as follows:
- Install AM3 into your eclipse environment
- Restart eclipse
- Change to the new AM3 perspective
- Right click on the model instance to be extracted into XML and select "Extract XML model to XML file"
- You always obtain XMI files (and thus non-customizable XML output) after running a launch configuration.
How can I handle arbitrary XML documents?
Arbitrary XML documents can be used as source or target of ATL transformations. However, this cannot be done directly (see previous section) but rather through injectors and extractors.
The XML injector can be executed from the contextual menu on files with the .xml extension, or from the am3.loadModel ant task. An example of using the XML injector is given in the how to use external parameter section.
The XML extractor can be executed from the contextual menu on files with the .ecore extension, or from the am3.saveModel ant task.
How can I implement my own injector or extractor?
An injection (resp. extraction) is a transformation from (resp. to) another Technical Space (e.g., XML, Grammarware) to (resp. from) the Model Engineering Technical Space. An injector (resp. extractor) is a tool implementing an injection (resp. extraction).
The injector extension point of the AM3 Core plugin enables plugins to contribute injectors. There is no such extension point for extractors yet.
How can I transform UML models?
There are several implementations of UML. The one we consider here is MDT/UML2.
To transform UML2 models with ATL from Eclipse the UML2 plugins must be installed. The UML2 metamodel must be loaded from the EMF registry (i.e., by namespace URI). This can be done using the "EMF Registry..." button of the ATL launch configuration, or using the nsURI attribute of the [] task.
Usage of content assist (i.e., code completion) is especially useful with big metamodels like UML2.
How can I retrieve tagged values from stereotyped UML model elements?
In ATL, a profiled UML model used as source model is viewed like any other UML model. You generally (i.e., when the profile is referenced with a valid relative path) do not have to specify the profile used with the model. So, in the header of the ATL module you just have:
module uml2something; create OUT : targetMetamodel from IN : UML;
If atl seems to ignore your stereotypes, you can add your profile as source model. You obtain the following header:
module uml2something; create OUT : targetMetamodel from IN : UML, INPROFILE : profile;
After that you can use operations like:
- getAppliedStereotype(sterotypeName : String) on a UML element to get the stereotype with the name "stereotypeName" applied on this element. This operation needs the fully qualifed name of the stereotype so it looks like: profileName::sterotypeName.
- getValue(stereotype : Stereotype, propertyName : String) to get the Value of the tagged value with name "propertyName" on the stereotype "stereotype" (the sterotype can be retrieved by getAppliedStereotype).
A helper like:
helper context UML!Element def: hasStereotype(name : String) : Boolean = not self.getAppliedStereotype(name).oclIsUndefined();
can also be useful to check if a stereotype is applied on the UML element before trying to retrieve it or to get a tagged value from it. As it uses getAppliedStereotype, it needs the fully qualified name of the stereotype (e.g., profileName::sterotypeName).