AttributeError for EDerivedCollection

[fabtrie]

I am trying to load an AUTOSAR arxml file. Therefore, I created the Ecore model from the publicly available XSD. When I try to load the resource, I am getting the following error:

File "c:\workspace\pyecore_issue\test.py", line 22, in main
    rset.get_resource(URI("test.arxml"))
  File "C:\python\3.8.1.0.7\python-3.8.1.amd64\lib\site-packages\pyecore\resources\resource.py", line 94, in get_resource
    resource.load(options=options)
  File "C:\python\3.8.1.0.7\python-3.8.1.amd64\lib\site-packages\pyecore\resources\xmi.py", line 63, in load
    self._decode_eobject(child, modelroot)
  File "C:\python\3.8.1.0.7\python-3.8.1.amd64\lib\site-packages\pyecore\resources\xmi.py", line 173, in _decode_eobject
    self._decode_eobject(child, eobject)
  File "C:\tpython\3.8.1.0.7\python-3.8.1.amd64\lib\site-packages\pyecore\resources\xmi.py", line 167, in _decode_eobject
    parent_eobj.__getattribute__(feat_container.name).append(eobject)
  File "C:\python\3.8.1.0.7\python-3.8.1.amd64\lib\_collections_abc.py", line 962, in append
    self.insert(len(self), value)
  File "C:\python\3.8.1.0.7\python-3.8.1.amd64\lib\site-packages\pyecore\valuecontainer.py", line 407, in __len__
    raise AttributeError('Operation not permited '
AttributeError: Operation not permited for "AR-PACKAGE" feature

It looks like it has something to do with the class EDerivedCollection which does not allow to get length information.

I am using pyecore version 0.12.2.

I attached a minimal example to reproduce the issue:
pyecore_issue.zip

[aranega]

Hi @fabtrie,

I'm so sorry, I totally missed your issue. Thanks for the case study and the ticket! Following the message, it looks like indeed, a derived feature is responsible for this behavior. As there is currently no implementation of the AUTOSAR metamodel in Python, PyEcore relies only in dynamic metamodel to load the XMI. I will try to load your project with Eclipse, without any plugin beside EMF to see if the same behavior occurs, if yes, it means that a static AUTOSAR metamodel implementation in Python will be required, if no, it means something is wrong on the assumptions I made while decoding the metamodel. I will try to test all of that asap!

Thanks again for your help in improving PyEcore and sorry again for the late answer.

EDIT> I found the issue there, it comes from the fact that PyEcore doesn't deal right now with ExtendedMetadata. It seems that for derived features, when there is a group annotation let in the ExtendedMetadata, EMF considers the feature pointed by group as a feature to which delegate the addition/removal of the object. I will try to think about a trick or something to overcome this until I propose a proper implementation for ExtendedMetadata. The implementation I have right now introduce an overhead for features accesses on instances.

[fabtrie]

OK. Thank you for the analysis. Actually, I used Artop before, which is an abstraction for AUTOSAR on top of EMF. This is working fine. But I want to get rid of the java dependency. If you can find a workaround, It would be highly appreciated. I need it mainly for extracting information and navigating through references.