Currently, I have an existing grammar for my DSL, how can I make use of it
in xtext?
Do I have to translate the grammar to xtext grammar? or some way else to use
it directly

I'd like to use my language editor inside a
dialog for a field (content assist etc).

Essentially, I would give the field a EMF model
and the field should edit the with the xtext
content assist and error handling.

Is this possible?

Michael
--
Michael Scharf
Wind River Systems GmbH
http://www.WindRiver.com
http://MichaelScharf.blogspot.com/

omfg, you are right. Input was "-1" and '-' was not in the definition of INT.
I've overwrite it with

terminal INT returns ecore::EInt: '-'?('0'..'9')+;

and everything was fine. Argh, sorry. The INT in XText is just NAT ;-)

I'm glad to hear you found the issue :)

That was a bug in Xtext in the 0.7.2-release.

You are my best friend when you'll deploy the current version of XText to the
Maven2 repository [http://openarchitectureware.org/m2/].

regards,
  Moritz

--- Original-Nachricht ---
Absender: Moritz Eysholdt
Datum: 16.09.2009 22:18

Hi Michael,

you are right, was not the correct line, but a very similar. The source is long and the corresponding exception print out is much longer.

But, back to topic.

If I try this [1], I got an exception in line 11.

Exception: org.eclipse.xtext.parsetree.reconstr.XtextSerializationException: Serialization failed

But why?

Generally speaking, this happens when your model contains values that are not written during serialization or when your model does *not* contain values that are required for serialization. Your grammar specifies which values these are. But since your are serializing the model right after you've parsed it, I guess that you haven't added any model elements in the meantime, right? One could do things like that in the Linker, for example. In case there is multiple inheritance in your Ecore model, this is likely to be the reason. That was a bug in Xtext in the 0.7.2-release. Updating to a newer release can help.

Model checked and OK.

OK.

Parsing OK.

To be sure parsing is *really* OK, please check whether resource.getWarnings() and resource.getErrors() are empty after calling resource.load().

Am I right, that Resource.load() is for parsing and Resource.save() for serialization?

Yes.

If my answers don't tackle your problem, please post your grammar and the textual model you've been parsing.

cheers,
  Moritz

[1] - java code
---------------
line 1: Injector injector = new DslStandaloneSetup().createInjectorAndDoEMFRegistration();
line 2: XtextResourceSet xtextResourceSet = injector.getInstance(XtextResourceSet.class);
line 3: xtextResourceSet.addLoadOption(XtextResource.OPTION_RESOLVE_ALL, Boolean.TRUE);
line 4: Map<?, ?> options = xtextResourceSet.getLoadOptions();
line 5: URI dslResource = org.eclipse.emf.common.util.URI.createURI("platform:/resource/dummy.dsl");

line 6: Resource resource = xtextResourceSet.createResource(dslResource);
line 7: InputStream in = new ByteArrayInputStream(dslCode.getBytes());
line 8: resource.load(in, xtextResourceSet.getLoadOptions());
line 9: ret = (Model) resource.getContents().get(0);
line 10: ByteArrayOutputStream bos = new ByteArrayOutputStream();
line 11: resource.save(bos, options);
==== end ====

--- Original-Nachricht ---
Absender: Moritz Eysholdt
Datum: 14.09.2009 16:15

Hi Michael,

what you are listing below is actually an error message and *not* your serialized model. However, this error message contains fragments of your serialized model.

Are you sure the message resulted from serializing the model from parsing "$varname = 5;"?

The message contains the serialized part of your model:
"] = $current ; $message = "square root found" ; } }"
And this doesn't look to me like it has originated from "$varname = 5;".

regards,
  Moritz

Hi Sebastian,

ok, I've got it running, but not in a way I want.

Parsing and serialization are divided into two classes in my implementation. Thats the reason why XtextResourceSet.getResource() reports: "resource was not registered".


If I merge the methods, I got a different serialization as I hoped for.
I put into the parser a line like this:

 $varname = 5;

after parsing and serializing I got something like that:
======
scripts[0] = Assign {
variable = Variable(name='de.btopia.ghandara.charme.impl.VariableNameImpl@55eb1db2 (name: $result)') {
69: "] = $current ; $message = "square root found" ; } }...":
-> Variable_IndexAssignment_1_1: Variable.index is not set.
68: "= $current ; $message = "square root found" ; } } }...":
-> Variable_PropertyAssignment_2_1: Variable.property is not set.
}
value = Expression {
headTerm = IntegerConst(...
====
OK, I found no easy-to-read code snipplet. Sorry. I can give you much more complex examples.

But first, the Injector needs an inject with the registered resources. The use cases (parse and serialize) can be made independent from each other. The parsed model has all information about the original DSL code. There is no need to store resource URIs _OR_ the authority for registration resources is unique in the whole system.

But in my case:
How can I reach my simple DSL code line "$varname = 5;" from serialization?

Ciao, Micha

--- Original-Nachricht ---
Absender: Sebastian Zarnekow
Datum: 11.09.2009 18:17

Hi Michael,

your resource is emtpy and has not been loaded.

Please try something like:

Resource res = resourceSet.getResource(URI.createURI(..existing uri..));
res.save(..);

resourceSet.createResource will create a new resource but won't load it.

res = resourceSet.createResource(..uri..);
res.load(new StringInputStream("valid model"));
res.save(..)

should do the trick as well.

Hope that helps,
Sebastian

--
Need professional support for Eclipse Modeling?
Go visit: http://xtext.itemis.com

afaik you can restrict the reference to attributes of the same class by providing your own ScopeProvider (see [1]). Here is some pseudo code, hope it helps:

public IScope scope_Attribute_name(Attribute context, EReference reference){

	// get all elements of the class
	EList<Element> elements = ((Class)context.eContainer).getElt();

	// get attributes
	Collection<Attribute> attributes = EcoreUtil.getObjectsByType(elements, YourDSLPackage.eINSTANCE.getAttribute());

	Iterable<IScopedElement> scopedElements = Scopes.scopedElementsFor(attributes, new Function<EObject, String>() {
		public String apply(EObject from) {
			return (((Attribute)from).getName());
		}
	});
		
	return new SimpleScope(IScope.NULLSCOPE, scopedElements);
}

Regards,
ILyas

------
[1] http://www.eclipse.org/Xtext/documentation/0_7_2/xtext.html#scoping

Hi all ,
this is my grammar :

Root: (root+=Model)* ;
Class: "class" name=ID (elt+=Element)*  ;
Element : Attribute | Ref ;
Attribute : "var" name=ID ;
Ref : "ref" name=[Attribute|ID] ;

the attribute "name" in the element "Ref" references all the elements "Attribute" in the model How I can restrict this reference to reference only the "Attributres " which exists in then same "Class"
for instance in this model :

class class1
    var var1     var var2
class class 2
   var var3
  var var4
  ref  var1

the reference to var1 from class2 must not work because var1 must be in class1.

Root: (root+=Model)* ;
Class: "class" name=ID (elt+=Element)*  ;
Element : Attribute | Ref ;
Attribute : "var" name=ID ;
Ref : "ref" name=[Attribute|ID] ;

the reference to var1 from class2  must not work  because var1 must be in class1.

Sven

Also, the generation part is not in the tutorial (as it used to be for the previous xtext), is it?

cheers
    Lorenzo

--
Need professional support for Eclipse Modeling?
Go visit: http://xtext.itemis.com

cheers
	Lorenzo

--
Lorenzo Bettini, PhD in Computer Science, DI, Univ. Torino
ICQ# lbetto, 16080134     (GNU/Linux User # 158233)
HOME: http://www.lorenzobettini.it MUSIC: http://www.purplesucker.com
http://www.myspace.com/supertrouperabba
BLOGS: http://tronprog.blogspot.com  http://longlivemusic.blogspot.com
http://www.gnu.org/software/src-highlite
http://www.gnu.org/software/gengetopt
http://www.gnu.org/software/gengen http://doublecpp.sourceforge.net

Michael,

We're embedding Xtend expressions in our grammar to do that. The major drawback is that the expressions themselves are not captured by grammar rules, but are instead defined within a String literal. So to the user of the language, it looks a bit like this:

define template 'ClassDescription' for ecore::EClass
if '!this.isAbstract && this.eAttributes.size() > 0'
heading '"EClass " + this.name'
body
...
end body
end template

But using Xtend has to nice advantages:

1) We can reuse the Xtend content assist facility. So at the right places, we just delegate from our Xtext language content assist code to Xtend content assist. Works like a charm.

Regards,

--knut