What is the correct way to return a Void
type, when it isn't a primitive? Eg. I currently use null as below.
interface B<E>{ E method(); }
class A implements B<Void>{
public Void method(){
// do something
return null;
}
}
This question is related to
java
generics
return-value
It is possible to create instances of Void
if you change the security manager, so something like this:
static Void getVoid() throws SecurityException, InstantiationException,
IllegalAccessException, InvocationTargetException {
class BadSecurityManager extends SecurityManager {
@Override
public void checkPermission(Permission perm) { }
@Override
public void checkPackageAccess(String pkg) { }
}
System.setSecurityManager(badManager = new BadSecurityManager());
Constructor<?> constructor = Void.class.getDeclaredConstructors()[0];
if(!constructor.isAccessible()) {
constructor.setAccessible(true);
}
return (Void) constructor.newInstance();
}
Obviously this is not all that practical or safe; however, it will return an instance of Void
if you are able to change the security manager.
There is no generic type which will tell the compiler that a method returns nothing.
I believe the convention is to use Object when inheriting as a type parameter
OR
Propagate the type parameter up and then let users of your class instantiate using Object and assigning the object to a variable typed using a type-wildcard ?
:
interface B<E>{ E method(); }
class A<T> implements B<T>{
public T method(){
// do something
return null;
}
}
A<?> a = new A<Object>();
Just for the sake of it, there is of course the possibility to create Void
instance using reflection:
interface B<E>{ E method(); }
class A implements B<Void>{
public Void method(){
// do something
try {
Constructor<Void> voidConstructor = Void.class.getDeclaredConstructor();
voidConstructor.setAccessible(true);
return voidConstructor.newInstance();
} catch (Exception ex) {
// Rethrow, or return null, or whatever.
}
}
}
You probably won't do that in production.
If you just don't need anything as your type, you can use void. This can be used for implementing functions, or actions. You could then do something like this:
interface Action<T> {
public T execute();
}
abstract class VoidAction implements Action<Void> {
public Void execute() {
executeInternal();
return null;
}
abstract void executeInternal();
}
Or you could omit the abstract class, and do the return null in every action that doesn't require a return value yourself.
You could then use those actions like this:
Given a method
private static <T> T executeAction(Action<T> action) {
return action.execute();
}
you can call it like
String result = executeAction(new Action<String>() {
@Override
public String execute() {
//code here
return "Return me!";
}
});
or, for the void action (note that you're not assigning the result to anything)
executeAction(new VoidAction() {
@Override
public void executeInternal() {
//code here
}
});
Java 8 has introduced a new class, Optional<T>
, that can be used in such cases. To use it, you'd modify your code slightly as follows:
interface B<E>{ Optional<E> method(); }
class A implements B<Void>{
public Optional<Void> method(){
// do something
return Optional.empty();
}
}
This allows you to ensure that you always get a non-null return value from your method, even when there isn't anything to return. That's especially powerful when used in conjunction with tools that detect when null
can or can't be returned, e.g. the Eclipse @NonNull
and @Nullable
annotations.
Source: Stackoverflow.com