Ensure enumeration values are rendered correctly

[szschaler]

This works around a bug in EMF, where enumeration attributes that hold the default value are reported as unset (even though by default enumeration attributes are set to not be unsettable).

[kolovos]

This is also the case for non-enumeration attributes. For example, in https://eclipse.dev/epsilon/playground/?9e36d1d1 the name attribute of the first box is marked as unset as the provided and the default values ("B1") match. If we want to make some/all attributes with default values appear on the diagram, I think we should do this using @diagram annotations as we already do for other customisations e.g.

@namespace(uri="bl")
@diagram(showDefaultValues="true") // Enable for all attributes
package bl;

class Box {
	attr String name = "B1";
	@diagram(showDefaultValue="true") // Enable for this specific attribute
	attr Color color;
}

[szschaler]

Huh. It gets even weirder: in that example, if I change the EOL program to

var box = Box.all.first();

box.isSet("name").println();
box.isUnsettable("name").println();

box.isSet("color").println();
box.isUnsettable("color").println();
box.color.println();

operation Any isSet(attr : String) {
    var eAttribute = self.eClass().getEStructuralFeature(attr);
    return self.eIsSet(eAttribute);
}

operation Any isUnsettable(attr : String) {
    var eAttribute = self.eClass().getEStructuralFeature(attr);
    return eAttribute.isUnsettable();
}

then I get

false
false
false
false
Red

In other words, EMF has correctly marked each of these attributes as not unsettable, and still maintains that they haven't been set -- which feels like a contract violation. My proposed fix to the renderer would catch this case, too, though, in a way that I feel would make the visualisation more intuitive.

Given that this is just in the context of the education platform / the playground, I don't think there's a need for special markup in the meta-model, but that's of course a decision for you to make.

The change I'm proposing here is the same I have implemented in the Xtext tool's visualisation of Xtext models.

[kolovos]

Doesn't making this change defeat the purpose of checking for e.eIsSet(attr)? For example, if I change the model as follows, the default value of the (unset) name attribute of the first box appears on the model diagram.

<?nsuri bl?>
<_>
    <box color="red"/>
    <box name="B2" color="green"/>
</_>

[szschaler]

But isn't that what one would expect to see? After all, that's the value of the object's attribute at that point, regardless of how it got set. However, if the attribute is unsettable and eIsSet() returns false, then the value is null or undefined. In that case, it feels OK to leave out the attribute altogether. Alternatively, one may wish to generate a line stating that something like attr = undefined.

[agarciadom]

Looks good to me, thank you!

I'll rebase this on top of the latest tip and then merge it.