org.eclipse.xtext.xbase.jvmmodel
Interface IJvmModelInferrer

All Known Implementing Classes:

AbstractModelInferrer, IJvmModelInferrer.NullImpl, PureXbaseJvmModelInferrer, XtendJvmModelInferrer

public interface IJvmModelInferrer

Model inference is done after a resource has been loaded. It is used to map you language's constructs to the elements defined in TypesPackage. Those elements closely reflect the Java type model.

For example if you have a language representing state machines with its states, transitions and actions, one way to map them to the types model would be to create a single class per state machine, an Enum for the different states, methods for the various actions and one method which contains the state machine logic. You would do that mapping in your language's implementation of this interface.

The inferred Jvm type model is used to compute how scoping is done and how the language compiles to Java. That is having a proper implementation of this interface is often the only thing you need to do to make your language executable on the JVM.

Implementers should use the JvmTypesBuilder API to create a model of Jvm elements in a convenient way.

Implementors have to traverse the given source object in infer(..) and decide which types to create according to the state of the model.

Author:

Jan Koehnlein - Initial contribution and API, Sven Efftinge

See Also:

IJvmModelAssociations, IJvmModelAssociator

Nested Class Summary

static class

IJvmModelInferrer.NullImpl A null-implementation.

Method Summary

void	infer(org.eclipse.emf.ecore.EObject e, IJvmDeclaredTypeAcceptor acceptor, boolean preIndexingPhase) This method is called at two different times in a resource's life-cycle, reflected by whether {preIndexingPhase} is set to true or false.

Method Detail

infer

void infer(org.eclipse.emf.ecore.EObject e,
           @NonNull
           IJvmDeclaredTypeAcceptor acceptor,
           boolean preIndexingPhase)

This method is called at two different times in a resource's life-cycle, reflected by whether {preIndexingPhase} is set to true or false. When set to true everything is still in a pre-indexing phase, that means linking hasn't been done yet. In this phase you just need to create the Jvm-elements which should be indexed (i.e. found globally). For regular Xbase expressions only the JvmTypes with the correct qualified name are needed at this point.

You must only infer Jvm elements which directly result from elements contained in the current resource!

When this method is called with preIndexingPhase set to false, you need to do the full inference including setting all links. But still you have to it in a particular order. First you need to create the complete tree structure and make sure you have passed the created JvmTypes to the acceptor. Only pass top level JvmTypes to the acceptor. Only if the tree structure is created and its root types have been passed to the acceptor, you are free to resolve any cross references.

Parameters:

e - the root element from the parse result

acceptor - an acceptor awaiting derived root JvmDeclaredTypes

preIndexingPhase - whether the call is done in before indexing. During this phase clients may not rely on any global indexing information and only need to to provide the information needed by the language's IDefaultResourceDescriptionStrategy. IF not implemented differently this is just the JvmDeclaredTypes with their qualified name, but no members and no other data.