Validation framework behavior issues

[abelhegedus]

In the current version, there are a number of issues in the Validation framework that needs to be clarified and probably fixed:

• The initialization uses the selection to create matchers (through the adapters), therefore if an EObject is selected, only the problems existing in the EMF model part reachable from the object is created.
• It seems that it is possible to initialize validators multiple times for the same selection, resulting in multiple markers.
• Instead of adding new adapters to an already initialized editor, they are overwritten. This also means that markers from the earlier initializations are not removed when closing the editor.
• When the marker is double clicked, it is not the registered editor where the location is shown but a new editor is opened for only the given resource... This happens when there are multiple models loaded into an editor and the model containing the location is not part of the first model that the editor was opened with.

For now, I have assigned this to myself and will see what I can fix before forwarding it to someone else if necessary.

[bergmanngabor]

"in the EMF model part reachable from the object " more precisely, in the containment hierarchy below the selected object

should the matcher scope rather be a Resource or the entire ResourceSet? Ábel and I vote for ResourceSet.

[abelhegedus]

Thanks for the clarification, Gábor.

Also, to be more precise, I vote for the ability to initialize the validation for the resource set, even if the Resource/EObject options remain in the framework.

[abelhegedus]

In the current version, the code includes the support for initializing validators for a selected subtree, but this part is commented out as there are multiple issues with such a possibility:

• The selection is used as the scope of the matchers, so there may (and probably will) be false positive and false negative violations due since only part of the model is traversed by the engine.
• There is no clear semantics for dealing with multiple initializations on different selecitons. E.g. what should happen if the initialization is called for a parent element of an earlier selection? What should happen with existing violations?

However, the option remains open for such a possibility e.g.:

• The selection might be used as a post-matching restriction on the results, but how should that work?
• The selection may be used in a future extension where parameterized validators can be defined and invoked (e.g. transitively reachable elements with a complex path query).

[abelhegedus]

The double-clicking behavior is a result of the internal workings of the EMF marker location handling. I have not been able to find a workaround for that yet.

[abelhegedus]

Moved to Eclipse.org BugZilla https://bugs.eclipse.org/bugs/show_bug.cgi?id=398739