Some Eclipse Foundation services are deprecated, or will be soon. Please ensure you've read this important communication.

Bug 322875

Summary:

Missing validation in xtext grammar when importing existing metamodels

Product:

[Modeling] TMF

Reporter:

Sebastian Benz <sebastian.benz>

Component:

Xtext

Assignee:

Project Inbox <tmf.xtext-inbox>

Status:

CLOSED FIXED

QA Contact:

Severity:

enhancement

Priority:

CC:

clay, sebastian.zarnekow, sven.efftinge

Version:

1.0.0

Flags:

sebastian.zarnekow: indigo+

Target Milestone:

Hardware:

All

OS:

All

Whiteboard:

Attachments:

Description	Flags
Sample project that demonstrates the problem.	none
proposed patch pls review	none

Description Sebastian Benz

2010-08-17 03:57:21 EDT

Created attachment 176761 [details]
Sample project that demonstrates the problem. 

I created a new Xtext language for an existing metamodel (Sample.ecore). In my model I reference elements in the Ecore metamodel (e.g. EClass):

import "platform:/resource/model/model/Sample.ecore" 

import "http://www.eclipse.org/emf/2002/Ecore" as ecore

Model returns Model:
   reference=[ecore::EClass] //<- validation error
;

The problem is that I get a validation error for the "reference" feature("Cannot find referenced feature reference in sealed EClass model...:"). I took me a while to figure out that the problem was mixing imports by platform resource URI and NSURI. When I import my existing metamodel via its NSURI ("http://sample") everything works fine. I think it would help a lot here to have a corresponding validation. I attached also an example project that demonstrates the problem.

Comment 1 Sebastian Zarnekow

2011-01-10 10:28:09 EST

Scheduled for M5 as grammar mixin and Ecore import will become more popular
with Xbase.

Comment 2 Michael Clay

2011-01-31 16:40:49 EST

(In reply to comment #1)
> Scheduled for M5 as grammar mixin and Ecore import will become more popular
> with Xbase.

thats because Sample.ecore refers to Ecore.ecore as platform:/plugin-uri 

<eStructuralFeatures xsi:type="ecore:EReference" name="reference" eType="ecore:EClass platform:/plugin/org.eclipse.emf.ecore/model/Ecore.ecore#//EClass"/>

instead of its ns-uri form

    <eStructuralFeatures xsi:type="ecore:EReference" name="reference" eType="ecore:EClass http://www.eclipse.org/emf/2002/Ecore#//EClass"/>

should we add an additional check for this kind of errors  

see 

http://www.eclipse.org/forums/index.php?t=msg&goto=551227&S=b33a349f818bb2e41cfceb0b7048056b

and

http://www.eclipse.org/forums/index.php?t=msg&goto=540574&S=85734ae7f35dff9fae4e185c63d0dc08

Comment 3 Sven Efftinge

2011-02-01 03:02:16 EST

(In reply to comment #2)
> should we add an additional check for this kind of errors  

Yes. How would you do it?

Comment 4 Sebastian Zarnekow

2011-02-01 03:06:19 EST

(In reply to comment #3)
> (In reply to comment #2)
> > should we add an additional check for this kind of errors  
> 
> Yes. How would you do it?

I'd assume something along 

if (eClass.ePackage.nsURI == otherEClass.ePackage.nsURI && eClass.fragment == otherEClass.fragment && eClass.uri != otherEClass.uri) {
  ... 
}

would work.

Comment 5 Sebastian Zarnekow

2011-02-01 03:08:24 EST

(In reply to comment #4)
> (In reply to comment #3)
> > (In reply to comment #2)
> > > should we add an additional check for this kind of errors  
> > 
> > Yes. How would you do it?
> 
> I'd assume something along 
> 
> if (eClass.ePackage.nsURI == otherEClass.ePackage.nsURI && eClass.fragment ==
> otherEClass.fragment && eClass.uri != otherEClass.uri) {
>   ... 
> }
> 
> would work.

As for the other validation issues that cover misleading errors in the Xtext grammar editor, this would require changes in the grammar inference code.

Comment 6 Sven Efftinge

2011-02-01 03:13:50 EST

We could just check whether any EPackages are referenced twice instead of checking the EClass references. The EPackage reference should be flagged with an error not the EClass reference.

Comment 7 Michael Clay

2011-02-07 07:51:32 EST

Created attachment 188432 [details]
proposed patch pls review

Comment 8 Sebastian Zarnekow

2011-02-07 10:13:48 EST

Michael,

I had to refactor the patch to support grammar inheritance.

Pushed fix to master.

Comment 9 Karsten Thoms

2017-09-19 17:59:53 EDT

Closing all bugs that were set to RESOLVED before Neon.0

Comment 10 Karsten Thoms

2017-09-19 18:10:37 EDT

Closing all bugs that were set to RESOLVED before Neon.0