Eclipse-Praktikum WS 12/13
Child pages
  • Textual Modeling with Xtext
Skip to end of metadata
Go to start of metadata

This tutorial presents the Xtext framework, a toolsuite for the generation of plain text based model editors. Such textual editors provide syntax highlighting, content assist (ctrl-space), an outline, and much more out-of-the-box. You will start by creating a textual syntax for Turing Machines.

Contents

Creating a Grammar

An Xtext grammar is always related to a specific EMF meta model. The grammar defines a concrete syntax in which instances of the meta model (the abstract syntax) can be serialized and stored. Xtext supports two ways of linking a grammar with a meta model: either creating a grammar for an existing meta model, or creating a grammar first and generating a meta model out of it. Here we will use the former approach, reusing the meta model for Turing Machines that you already defined earlier.

  1. Install the Xtext Antlr Runtime Feature from http://download.itemis.de/updates/ (click Help → Install New Software... and enter the URL into the topmost text field; once Eclipse has downloaded the list of software available under that URL, select the feature from the list)
  2. File → New → Project... → Xtext → Xtext Project From Existing Meta Models → Next → Add... → turing.genmodel → OK
  3. Select your top-level model element as "Entry rule", e.g. TuringMachine → Next → enter the following values:
    • Project name: de.cau.cs.rtprak.login.turing.text (like in previous tutorials, replace "login" by your login name)
    • Location: your repository path
    • Name: de.cau.cs.rtprak.login.turing.text.Turing
    • Extensions: tuxt (the file extension under which your text files will be stored)
    • Uncheck Create SDK feature project (not required)
  4. Finish → an editor is opened with a predefined grammar (Turing.xtext). The fourth line should contain an import statement followed by the namespace URI of your meta model, which is probably marked with an error. The problem here is that the meta model is not loaded in the Eclipse instance you are using for development. Fix this problem by replacing the URI by the workspace path of your Turing Machines ecore model, e.g. "platform:/resource/de.cau.cs.rtprak.login.turingmodel/model/turing.ecore"
  5. The Turing.xtext file is located in the package de.cau.cs.rtprak.login.turing.text together with a Modeling Workflow Engine file, which can be used to configure code generation. Right-click GenerateTuring.mwe2 and select Run As → MWE2 Workflow. If you get an error message like "*ATTENTION* It is recommended to use the ANTLR 3 parser generator" in the console, type n and install Antlr from the update site.
  6. Now a great amount of Java code should have been generated. Add the new plugins to your Eclipse run configuration and start it.
  7. In the new Eclipse instance: File → New → File and name it test.tuxt (when asked to add the Xtext nature, hit Yes).

  8. Use ctrl + space for getting content assist to create a valid model following the predefined grammar. The syntax will probably look similar to this:

    TuringMachine {
    states {
        State blub {
        },
        State blah {
            outgoing {
                Transition {
                    target blub
                }
            }
        }
    }
}

The predefined grammar is not very nice. That's why now your creative part comes into play. Change the Xtext grammar file to a grammar of your liking, which should support a simple and readable syntax for Turing Machines. All required documentation for this task is found at http://www.eclipse.org/Xtext/.

Some hints:

  • The predefined grammar contains one grammar rule for each meta model class. This general structure can be kept in most cases.
  • Ecore enumerations (EEnum) must be defined with an enum grammar rule (see documentation). This is not correct in the predefined grammar and is marked with a TODO note.
  • At some points you will need to reference other elements, which need to have some kind of identifier string in order to do this. If those elements do not have a string attribute yet, you need to modify the meta model and add such an attribute.
  • There are some predefined terminal rules that can be reused and are already imported in the grammar file. Click on org.eclipse.xtext.common.Terminals at the top of the document and hit F3 to see those terminal definitions.

Test the new grammar by re-generating the code (after the first time, the GenerateTuring workflow should be available in the run configurations menu) and starting a test instance of Eclipse. Use your newly designed textual syntax for writing a Turing Machine that copies the input word infinitely often.

Formatting

Xtext supports automatic formatting, which is available in the text editor with right-click → Format or ctrl+shift+F. However, the formatter must be configured in order to generate good results. Write a formatter configuration that fits well to your syntax by editing the generated file TuringFormatter in the formatting subpackage. Learn how this is done by reading the Xtext reference documentation → Runtime Concepts → Formatting.

Validation

The generated code includes some automatic validation of models with respect to syntactic issues. If the token sequence in a text file does not conform to the grammar, error markers are shown at appropriate points in the text. However, this should be augmented by semantic validation by checking high-level properties of the model. Implement such a semantic validation by editing the generated file TuringJavaValidator in the validation subpackage. Learn how this is done by reading the Xtext reference documentation → Runtime Concepts → Validation → Custom Validation. You should implement at least the following checks:

  • Is there exactly one initial state?
  • Are all states reachable through transitions starting from the initial state?

Parsing

The generated code includes a parser for text files in your syntax. This parser is simply used with the same interface as for any other EMF models: resource sets (see the Xtext reference documentation → Integration with EMF and Other EMF Editors). Your final task for this tutorial is to reuse the Turing controller you implemented in the EMF tutorial for simulating Turing Machine models, this time applying it to .tuxt text files. However, instead of using a fixed path to your model file, we will now use a menu contribution in order to set the file path dynamically. This contribution will be put into the popup menu of the Navigator / Project Explorer view by configuring a visibility expression that tests the currently selected (right-clicked) elements. The contribution shall only be visible if the selected files have the extension tuxt, which is assigned to our mighty textual syntax.

  1. Go to the plugin where you created and registered TuringHeadController, the Turing Machine simulator. Open plugin.xml → Dependencies and add org.eclipse.ui if not on the list yet. Go to Extensions and add org.eclipse.ui.menus.
  2. Add a menuContribution element to the new extension and set "popup:org.eclipse.ui.popup.any?after=additions" as locationURI (without quotation marks).
  3. Add a command element to the menuContribution with the following attributes:
    • commandId: de.cau.cs.rtprak.login.setSimFile
    • label: Set Simulation File
  4. Add a visibleWhen element to the command, and add an iterate element to the visibleWhen with the following attributes:
    • operator: and
    • ifEmpty: false
  5. Add an adapt element to the iterate with the following attribute:
    • type: org.eclipse.core.resources.IResource
  6. Add a test element to the adapt with the following attributes:
    • property: org.eclipse.core.resources.extension
    • value: tuxt
  7. Add a new extension org.eclipse.ui.commands and add a command element to it with the following attributes:
    • id: de.cau.cs.rtprak.login.setSimFile
    • name: Set Simulation File
    • Click on defaultHandler to open a dialog for creation of a new handler class. Name the new class SetFileHandler and put it into some package of that plugin. Remove the suggested interface and set org.eclipse.core.commands.AbstractHandler as superclass instead.

  8. Use the following method stub for SetFileHandler (this requires a plugin dependency to org.eclipse.core.resources):

    /**
 * {@inheritDoc}
 */
@Override
public Object execute(ExecutionEvent event) throws ExecutionException {
    ISelection selection = HandlerUtil.getCurrentSelection(event);
    if (selection instanceof IStructuredSelection) {
        Object element = ((IStructuredSelection) selection).getFirstElement();
        if (element instanceof IFile) {
            IFile file = (IFile) element;
            // TODO update the static reference to the simulation file
        }
    }
    return null;
}
  9. Add a public static field named modelFile to TuringHeadController and directly set that field in the TODO part of the execute method shown above.
  10. Use the resource set code shown in the EMF tutorial for loading model files (without the stand-alone part) in order to load the modelFile in the initialize method.
  11. Now you should be able to simulate models written in your textual syntax: start Eclipse, right-click a textual Turing Machine file (*.tuxt), select Set Simulation File, and run simulation in your Tape view using the correct controller.


  • No labels