org.openrewrite.staticanalysis.FinalClass is wrong for nested sub classes

[Bananeweizen]

org.openrewrite.staticanalysis.FinalClass works on wrong assumptions. A non-final class with no public constructor can have nested static sub classes, and in that case it cannot be made final.

What is the smallest, simplest way to reproduce the problem?

public class Reproducer {

    private Reproducer(final String name) {
        //...
    }

    public static final class Sub extends Reproducer {

        public Sub() {
            super("subclass");
        }
    }
}

What did you expect to see?

no change

What did you see instead?

final added to the top level class

Are you interested in contributing a fix to OpenRewrite?

Maybe. Have to look for how to get the information that there are nested sub classes.

[timtebeek]

Thanks for reporting this variant! To detect sub classes you can either expand this loop

rewrite-static-analysis/src/main/java/org/openrewrite/staticanalysis/FinalClassVisitor.java

Lines 72 to 83 in 2163f0f

	for (Statement s : cd.getBody().getStatements()) {
	if (s instanceof J.MethodDeclaration && ((J.MethodDeclaration) s).isConstructor()) {
	J.MethodDeclaration constructor = (J.MethodDeclaration) s;
	constructorCount++;
	if (!constructor.hasModifier(J.Modifier.Type.Private)) {
	allPrivate = false;
	}
	}
	if (constructorCount > 0 && !allPrivate) {
	return cd;
	}
	}

Or add a new JavaIsoVisitor<>() { ... } that overrides visitClassDeclaration, and visit on the outer classDeclaration.getBody() and set an AtomicBoolean if any matching nested class declarations are found.