Community
Participate
Working Groups
Build Identifier: I20110613-1736 The default qualified names provider does not correctly split qualified names into segments (the wrong '::' delimiter is used) when linking objects. Reproducible: Always Steps to Reproduce: 1. Create Xtext grammar with rule 'grammar=[xtext::Grammar|QualifiedName]' 2. Try linking to exported Domainmodel grammar instance 3. Failure to link with message "The feature 'language' of 'com.x.y.z.CatalogImpl@5cdf858e{platform:/resource/com.x.y.z.tests/src/test/Sample.check#//@catalog}' contains an unresolved proxy 'org.eclipse.xtext.impl.GrammarImpl@4a378244{platform:/resource/com.x.y.z.tests/src/test/Sample.check#xtextLink_::0.1::4::/6}'" Workaround: overwrite *org.eclipse.xtext.xtext.XtextLinkingService.getLinkedObjects(EObject, EReference, INode)*
See XtextQualifiedNameProvider. Simplest workaround if you want to link to an Xtext grammar: Override DefaultLinkingService#getLinkedObjects and query the scope with a hand-made qualified name when the grammar should be resolved. Downside: Imports don't work.
Since this is the first complaint on the wrong exported name, I assume that other languages, that refer to a grammar, use classpath uris to load it (e.g. code templates or Xtext itself). I'll fix the exported name and won't export a grammar instance twice. Any objections?
Pushed to master.
The problem Johan described is still reproducible with Xtext 2.2.1.
(In reply to comment #4) > The problem Johan described is still reproducible with Xtext 2.2.1. I cannot reproduce what you describe. I would assume that the index in your workspace contains obsolete entries for grammars? Did you try a clean build?
I've tried running a clean build but the problem persists, the grammar won't be found. So we still need the workaround (which I was attempting to remove) with a hand-made qualified name when querying the scope in getLinkedObjects.
My test grammar: grammar org.some.Sample with org.eclipse.xtext.common.Terminals generate sample "sample" import "http://www.eclipse.org/2008/Xtext" Model: name=QualifiedName '{' imports+=Import* grammarRef+=GrammarRef* '}' ; GrammarRef: 'ref' name=ID ('in' ^grammar=[Grammar|QualifiedName])? ; Import: 'import' importedNamespace=QualifiedNameWithWildcard ; QualifiedNameWithWildcard: QualifiedName '.*'? ; QualifiedName: ID ('.' ID)*; ==== The test models: org.abc { import org.xyz.* ref Thing in Sample ref Thing2 in org.xyz.Sample } === grammar org.xyz.Sample generate model "model" Model: {Model} "STRING"; === Could you please provide more details about your concrete setup?
Actually, qualified names _are_ now being correctly split into segments. So given the bug description, this bug has been fixed. So we have a local problem in our implementation that is merely related to this entry, my bad. Case closed.
Closing all bugs that were set to RESOLVED before Neon.0
