Pyecore

++++++

**Bugfixes**

- Fix issue with serialization of multiple empty strings.

- Fix relative HttpURI computation.

- Add inheritance mix for static/dynamic metaclass.

- Fix issue with moving object from resource root to object back and forth.


**Miscellaneous**

- Add better pythonic implementation of "behavior".

- Add more detailed messages for XMI deserialization.

- Pass to full f-strings.

- Add cache for resources.

++++++

**Features**

- Add listeners on resources. A listener can be added on a resource to be notified of all the modifications performed on objects of its hierarchy.

- Add new ``container`` attribute on ``EReference`` that indicates if a reference is a container (it owns an opposite which is a containment).

**Bugfixes**

- Fix issue with ``xmi:type`` and ``xsi:type`` for mixed XMI. In some XMI files, the field related to type can be mixed and sometimes prefixed by ``xmi`` or ``xsi``. The flip-flop between the two mode was only performed once in the previous algorithm, now it is adaptive. Thanks `aacebedo <https://github.com/aacebedo>`_ for the pull request.

- Fix missing ``EShortObject``. Thanks `aacebedo <https://github.com/aacebedo>`_ for the pull request.

- Fix MRO problem when updating inheritance. The previous algorithm was not taking into account dynamic additions of inheritance relationship. Thanks `mn3mos <https://github.com/mn3mos>`_ for the issue.

- Fix issue on metamodel with ``abstract`` decorator.

- Fix issue with ``EEnumLiteral`` serialization for JSON resource. Thanks `jinhu <https://github.com/jinhu>`_ and Marc Hamilton for the issues and the pull request.

- Fix issue with clearing non unique collections. A set operation was performed on the collection to safely iterate on it, implying that some elements were forget during a full clear. Thanks `ewoudwerkman <https://github.com/ewoudwerkman>`_ for the issue.

**Miscellaneous**

- Improve slightly runtime typechecking performances.

- Improve collection speed.

- Improve eType computation for speed.

- Disabling resolve cache for resources when a resource had been properly load. Thanks `ewoudwerkman <https://github.com/ewoudwerkman>`_ for the suggestion.

++++++

**Bugfixes**

- Fix issue with MRO calculation. This happend when the Ecore inheritence tree is not compatible with Python C3 linearization algorithm. Thanks `mn3mos <https://github.com/mn3mos>`_ for the issue.

- Fix non-serializable ``EDataType`` ``instanceClassName``. The property was not serialized by the XMI resource. Thanks `4ekin <https://github.com/4ekin>`_ for the pull requests.

++++++

**Bugfixes**

- Fix issue with new ``ordered_set`` version. This new version removed some
parts functions about iterables. Thanks `fgro93 <https://github.com/fgro93>`_ for the issue.

++++++

**Bugfixes**

- Fix issue with default value computation using ``defaultValueLiteral``. PyEcore
was providing it's own "default value" mechanism that had an higher priority than
the one usually used by EMF. This resulted in the default value not being computed
correctly. Thanks `annighoefer <https://github.com/annighoefer>`_ for the PR
and pointing out the issue.

++++++

**Bugfixes**

- Fix issue with xmi ecore supertype references. When an ecore model has two elements A and B and that B inheriths from
A which is another element in a second ecore model, then, as their name
are equivalent, the resolution could, sometimes have problems in
resolving the elements, mistaking A from the first ecore with A from the
second one. In the case the wrong ecore was referenced, it implied a
circular reference for inheritency that caused the ecore file loading to
fail.

- Fix issue with weird URI (``plateform:/``). Weird uri that does not contains ``://`` provoques problems in the
automatic object resolving. This fix avoid that the relative uri
computation adds the relative path to the external uri it tries to
resolve.

- Fix issue with computation of ``many`` feature. The "algo" used to compute the ``many`` feature was wrong and introduced non-logical behavior when a reference or an attribute was having its lower and upper bound set to a same value higher than ``0`` (e.g.: ``2..2``).Thanks
`jorensips <https://github.com/jorensips>`_ for the issue.