XbaseInterpreter.java:_doEvaluate(XCastedExpression - `throw new EvaluationException(new ClassCastException(typeName))` - not enough information to show to the user what cannot be casted to what

class A {
    public static void main(String[] arg) {
        try {
            Object x = Integer.valueOf(0);
            System.out.println("A" + (String)x);
        } catch (ClassCastException e) {
            System.out.println(e.getMessage());
        }
    }
}

produces with Java 21

$ java A.java
class java.lang.Integer cannot be cast to class java.lang.String (java.lang.Integer and java.lang.String are in module java.base of loader 'bootstrap')

That is: ClassCastException.getMessage() is expected to spell both source and target classes of casting.

When XbaseInterpreter.java:protected Object _doEvaluate(XCastedExpression castedExpression, IEvaluationContext, CancelIndicator) creates a ClassCastException, it puts as detailed message the typeName - the target type.

To print a casting error to the user, containing both source and target classes of casting, one has in a derived class to:

override Object _doEvaluate(XCastedExpression castedExpression, IEvaluationContext context, CancelIndicator indicator) {
    try {
        return super._doEvaluate(castedExpression, context, indicator)
    } catch (RuntimeException e) {
        if (e.cause instanceof ClassCastException) {

and then constructs a message, containing both source and target types of the casting. Obtaining the target type is easy, it is a matter of calling castedExpression.getType().getType().getQualifiedName(). In fact this is the value of the parameter passed to the constructor of ClassCastException in XbaseInterpreter.java:_doEvaluate(XCastedExpression, IEvaluationContext, CancelIndicator).

However there is no way to know, when the EvaluationException(new ClassCastException()) is catched, what was the source type of the expression: In a as B the class of a.

In openHAB to get the source type is re-executed val Object result = internalEvaluate(castedExpression.getTarget(), context, indicator);. The problem is that calling once val Object result = internalEvaluate(castedExpression.getTarget(), context, indicator); in the super class, and calling again val Object result = internalEvaluate(castedExpression.getTarget(), context, indicator); in the derived class might not necessary have the same result: states might have changed in the meantime, so that re-evaluating the same expression, for the sake of logging an error, might produce no more errors. This in fact does happen, as described at https://github.com/openhab/openhab-core/issues/5323 .

That said in

protected Object _doEvaluate(XCastedExpression castedExpression, IEvaluationContext context, CancelIndic
ator indicator) {
…
                try {
                        expectedType.cast(result);
                } catch (ClassCastException e) {
                        throw new EvaluationException(new ClassCastException(typeName));
                }

• spell in the ClassCastException constructor parameter as sentence the target and source types of the casting, as java does in the example of the beginning, or
• instead of passing the target class name (which can be implied when overriding the method by calling castedExpression.getType().getType().getQualifiedName() , pass the name of the source class.

In both cases the aim is to present to the user a message telling what cannot be casted to what, and currently this is not possible.

Start with XbaseInterpreter.java at _doEvaluate(XCastedExpression, IEvaluationContext, CancelIndicator) and compare its ClassCastException handling with Java's message. Read openHAB's ScriptInterpreter.xtend usage and openHAB issue 5323 to understand the dangers of re-evaluating the target expression. Done means the resulting evaluation error identifies both source and target types without relying on a second evaluation.