DSL中Xtext插件下独立EMF模型创建及应用技术问询

Great question! Let’s walk through exactly how to create a standalone EMF model from your registered Ecore model—one that’s ready for code generation and graphical representation—after validating your Xtext DSL files.

Xtext generates an Ecore model alongside your DSL grammar, but this model is tightly coupled to Xtext’s runtime. To create a standalone version:

• Locate your Xtext project’s .ecore file (usually in src/main/resources). If it was auto-generated from your grammar, right-click it and select Export > EMF Model > Ecore Model to save a clean, standalone copy.
• Create a new EMF project (or reuse an existing non-Xtext project) and import this standalone .ecore file. Right-click the imported Ecore, choose Generate EMF Model Code to generate the standalone model classes, factory, and package—these will have no dependencies on Xtext.

Once your Xtext DSL file passes validation, you’ll need to map the Xtext-specific model to your standalone EMF model. Here’s how to do it in code:

• First, fetch the validated Xtext model from the resource:

  // Get the validated Xtext model from the resource
  Resource xtextResource = ...; // Your validated Xtext resource
  EObject xtextModel = xtextResource.getContents().get(0);
  

• Initialize your standalone EMF package to ensure it’s registered with EMF:

  // Trigger package initialization (generated with your EMF code)
  YourStandalonePackage.eINSTANCE.eClass();
  

• Map the Xtext model to your standalone model. You can use manual mapping (for full control) or EcoreUtil.copy() (for simple, identical structures):

  // Example: Manual mapping for an Order model
  XtextOrder xtextOrder = (XtextOrder) xtextModel;
  StandaloneOrder standaloneOrder = YourStandaloneFactory.eINSTANCE.createStandaloneOrder();
  
  // Copy core properties
  standaloneOrder.setOrderId(xtextOrder.getOrderId());
  standaloneOrder.setCustomerName(xtextOrder.getCustomerName());
  
  // Map child elements like OrderItems
  for (XtextItem xtextItem : xtextOrder.getItems()) {
      StandaloneItem standaloneItem = YourStandaloneFactory.eINSTANCE.createStandaloneItem();
      standaloneItem.setProductName(xtextItem.getProductName());
      standaloneItem.setQuantity(xtextItem.getQuantity());
      standaloneOrder.getItems().add(standaloneItem);
  }
  

• Save the standalone model to an XMI file for later use:

  ResourceSet standaloneResourceSet = new ResourceSetImpl();
  standaloneResourceSet.getResourceFactoryRegistry()
      .getExtensionToFactoryMap()
      .put("xmi", new XMIResourceFactoryImpl());
  
  URI outputUri = URI.createFileURI("/path/to/your/standalone-model.xmi");
  Resource standaloneResource = standaloneResourceSet.createResource(outputUri);
  standaloneResource.getContents().add(standaloneOrder);
  standaloneResource.save(Collections.emptyMap());
  

To use this model for code generation:

• Use EMF’s GenModel: Right-click your standalone .ecore file, create a .genmodel file, then configure code generation templates (Java, C++, etc.) and generate code directly.
• Or use Acceleo: Write template files that read the standalone XMI model and generate your target code. Ensure your Acceleo project depends on your standalone EMF model project (not the Xtext project) to keep it decoupled.

For visualizing or editing the standalone model, you have two main options:

• EMF Tree Editor: Right-click your .genmodel file, select Generate Editor Code to create a basic tree-based editor for your model.
• Sirius or GMF:
  • For Sirius: Create a Sirius Viewpoint Specification Project, define diagrams, nodes, and edges based on your standalone Ecore model. You can then open the standalone XMI model in a Sirius editor for graphical editing.
  • For GMF: Build a graphical editor by creating a GMF Tooling project tied to your standalone Ecore model, then generate the editor code for full-featured graphical support.

• Ensure your standalone Ecore model matches the structure of your Xtext model (or handle structural differences in the conversion step).
• Always initialize your standalone EMF package at application startup (e.g., in your plugin’s Activator.start() method) to avoid EMF not recognizing model types.
• If your Xtext model has cross-resource references, make sure to resolve and map these correctly in the standalone resource set.

内容的提问来源于stack exchange，提问作者Johan