QVT 1.1 RTF Avatar
  1. OMG Issue

QVT11 — errors and anomalies in QVT_1.0.mdl in the 07-07-08 ZIP

  • Key: QVT11-30
  • Legacy Issue Number: 12518
  • Status: closed  
  • Source: Model Driven Solutions ( Ed Willink)
  • Summary:

    Use of automated tooling to support comparison of the models developed
    initially as part of the Eclipse GMT/UMLX project and being transferred to the
    Eclipse QVT Declarative/QVT Operational Mappings Projects reveals the following
    errors and anomalies in QVT_1.0.mdl in the 07-07-08 ZIP.

    Since the diagrams are printed from QVT_1.0.mdl all the QVT problems also
    occur in 08-04-03.

    Textual errors in 08-04-03 cannot be analyzed automatically. There are
    so many that a thorough proof read is required combined with a statement
    that the diagrams only are normative

  • Reported: QVT 1.0 — Fri, 6 Jun 2008 04:00 GMT
  • Disposition: Resolved — QVT 1.1
  • Disposition Summary:

    The QVT 1.0 diagrams and models were originally from the models in the QVT_1.0.mdl file. The
    diagrams rely on proprietary tooling. Unfortunately some independent evolution occurred and so there
    were many inconsistencies.
    Consistent Ecore/EMOF files from Eclipse were endorsed as the QVT 1.1 non-normative files.
    For QVT 1.2, the primary non-normative files are UML models derived from the QVT 1.1 Ecore files.
    The diagrams are redrawn from the UML using the Open Source Papyrus tool. The diagrams are
    drawn to assist understanding rather than to squeeze as much as possible into the useable area. The
    redrawn diagrams are therefore larger/more numerous. (Unfortunately Papyrus does not support

    {ordered} so {ordered}

    text is added manually.) In all other respects diagrams and UML non-normative
    files should be consistent.
    The multiplicity of unnavigable opposites was not present in the Ecore files and so is set to 0..1 for
    compositions, and 0..* for non-compositions by the Ecore2UML conversion. This causes a few 1
    multiplicities to change to 0..1. The 1 multiplicities were needlessly restrictive prohibiting re-use of
    classes, so the 0..1 seems better. Perhaps we should have an issue to change navigable containers
    too.
    TemplateParameterType is missing from the diagrams since it is not preset in the non-normative files.
    TemplateParameterType has no practical utility implementations are required to work magic for
    templates types which may or may not involve the TemplateParameterType class.
    I find that when I consult the specification, I want to see all the diagrams, which is hard when they are
    spread thoughout the AS section. All diagrams are therefore brought to the start of their AS section.

  • Updated: Fri, 6 Mar 2015 20:58 GMT