How to highlight an Xtext cross-reference differently for targets of different types? - eclipse-plugin

Question

I have an Xtext grammar which reads (in part):
grammar mm.ecxt.MMLanguage hidden(WS, COMMENT)
import "http://www.eclipse.org/emf/2002/Ecore" as ecore
...
Statement:
ConstantStatement |
VariableStatement |
LabeledStatement |
...
LabeledStatement:
EssentialHypothesisStatement |
...
ConstantStatement:
DOLLAR_C (constants+=ConstDecl)+ DOLLAR_DOT;
VariableStatement:
DOLLAR_V (variables+=VarDecl)+ DOLLAR_DOT;
EssentialHypothesisStatement:
name=LABEL DOLLAR_E (symbols+=[Decl|MATHSYMBOL])+ DOLLAR_DOT;
Decl: ConstDecl | VarDecl;
ConstDecl returns ConstDecl: name=MATHSYMBOL;
VarDecl returns VarDecl: name=MATHSYMBOL;
MATHSYMBOL: PARENOPEN | PARENCLOSE | QUESTIONMARK | COMPRESSED | TLABEL | WORD;
...
(The full grammar is MMLanguage.xtext from current commit 328a5e7 of https://github.com/marnix/metamath-eclipse-xtext/.)
My question: How do I highlight the symbols in an EssentialHypothesisStatement, by using a different color for constants and variables? So if the MATHSYMBOL refers to a ConstDecl, then it should be highlighted one way, and some other way for a VarDecl.
I've tried to create an ISemanticHighlightingCalculator in all kinds of ways, but I can't seem to detect what the actual reference type is, neither through the node model nor through the Ecore model. On the one hand, the grammar-related methods only tell me that the reference goes to a Decl. On the other hand, the Ecore model's EReferences tell me whether the target is a ConstDecl or a VarDecl, but there I can't find the location of the source MATHSYMBOL.
Note that I prefer to use the node model (as opposed to the Ecore model) since I also want to highlight comments, and for performance reasons I cannot afford multiple passes over the document.
What is a good/canonical/efficient/simple way to achieve this?

Answer 1

from EObject perspective have a look at org.eclipse.xtext.nodemodel.util.NodeModelUtils.findNodesForFeature(EObject, EStructuralFeature)
from the node model perspective you can use EObjectAtOffsetHelper
sample grammar
Model:
defs+=Def*
uses+=Use*
;
Def:
ADef | BDef;
ADef:
"adef" name=ID
;
BDef:
"bdef" name=ID
;
Use:
"use" def=[Def]
;
And here the Impl
public class MyDslSemanticHighlightingCalculator implements ISemanticHighlightingCalculator {
#Inject
private MyDslGrammarAccess ga;
#Inject
private EObjectAtOffsetHelper helper;
#Override
public void provideHighlightingFor(XtextResource resource, IHighlightedPositionAcceptor accoptor, CancelIndicator cancelIndicator) {
if (resource == null)
return;
IParseResult parseResult = resource.getParseResult();
if (parseResult == null || parseResult.getRootNode() == null)
return;
BidiTreeIterable<INode> tree = parseResult.getRootNode().getAsTreeIterable();
for (INode node : tree) {
if (cancelIndicator.isCanceled()) {
return;
}
if (node.getGrammarElement() instanceof CrossReference) {
if (ga.getUseAccess().getDefDefCrossReference_1_0() == node.getGrammarElement()) {
EObject target = helper.resolveElementAt(resource, node.getOffset());
if (target instanceof ADef) {
accoptor.addPosition(node.getOffset(), node.getLength(), HighlightingStyles.COMMENT_ID);
} else if (target instanceof BDef) {
accoptor.addPosition(node.getOffset(), node.getLength(), HighlightingStyles.STRING_ID);
}
}
}
}
}
}