Why are class static methods inherited but not int

Question

I understand that in Java static methods are inherited just like instance methods, with the difference that when they are redeclared, the parent implementations are hidden rather than overridden. Fine, this makes sense. However, the Java tutorial notes that

Static methods in interfaces are never inherited.

Why? What's the difference between regular and interface static methods?

Let me clarify what I mean when I say static methods can be inherited:

class Animal {
    public static void identify() {
        System.out.println("This is an animal");
    }
}
class Cat extends Animal {}

public static void main(String[] args) {
    Animal.identify();
    Cat.identify(); // This compiles, even though it is not redefined in Cat.
}

However,

interface Animal {
    public static void identify() {
        System.out.println("This is an animal");
    }
}
class Cat implements Animal {}

public static void main(String[] args) {
    Animal.identify();
    Cat.identify(); // This does not compile, because interface static methods do not inherit. (Why?)
}

Answer 1

Before Java 8, you couldn't define static methods in an interface. This is heavily discussed in this question. I'm going to refer to this answer (by user @JamesA.Rosen) as to why the Java designers probably didn't want static methods in an interface initially:

There are a few issues at play here. The first is the issue of declaring a static method without defining it. This is the difference between

public interface Foo {
  public static int bar();
}

and

public interface Foo {
  public static int bar() {
    ...
  }
}

Java doesn't allow either, but it could allow the second. The first is impossible for the reasons that Espo mentions: you don't know which implementing class is the correct definition.

Java could allow the latter, as long as it treated Interfaces as first-class Objects. Ruby's Modules, which are approximately equivalent to Java's Interfaces, allow exactly that:

module Foo
  def self.bar
    ...
  end
end

However, since the release of Java 8, you can actually add default and static methods inside an interface.

I'm going to be quoting this source a lot here. This is the initial problem:

Java's interface language feature lets you declare interfaces with abstract methods and provide implementations of those methods in the classes that implement the interfaces. You are required to implement each method, which is burdensome when there are many methods to implement. Also, after publishing the interface you cannot add new abstract methods to it without breaking source and binary compatibility.

This was the solution Java 8 provided default:

Java 8 addresses these problems by evolving the interface to support default and static methods. A default method is an instance method defined in an interface whose method header begins with the default keyword; it also provides a code body. Every class that implements the interface inherits the interface's default methods and can override them

And for static:

A static method is a method that's associated with the class in which it's defined, rather than with any object created from that class. Every instance of the class shares the static methods of the class. Java 8 also lets static methods be defined in interfaces where they can assist default methods.

When you implement an interface that contains a static method, the static method is still part of the interface and not part of the implementing class. For this reason, you cannot prefix the method with the class name. Instead, you must prefix the method with the interface name

Example:

interface X
{
   static void foo()
   {
      System.out.println("foo");
   }
}

class Y implements X
{
}

public class Z 
{
   public static void main(String[] args)
   {
      X.foo();
      // Y.foo(); // won't compile
   }
}

Expression Y.foo() will not compile because foo() is a static member of interface X and not a static member of class Y.

Answer 2

Static methods in interfaces could create a diamond of death if they were being inherited. So, calling a static method from the appropriate interface is good enough compared to the risk of calling it from a concrete class that may implement multiple interfaces that contain static methods of the same name.

Why are static methods any different?

Static methods are just functions unrelated to the objects. Instead of placing them in utility abstract classes (like calling Collections.sort() ) we move those functions (static methods) to their appropriate interfaces. They could be bound to the inherited objects like the default methods do, but that is not their job. Static methods provide functionality which is unrelated to the instances of the class.

Example:

interface Floatable {

    default void float() {
        // implementation
    }

    static boolean checkIfItCanFloat(Object fl) {
         // some physics here
    } 
}

class Duck implements Floatable { }

So, the point is that a Duck may float but the function that checks if an Object really floats is not something that a Duck can do. It is an irrelevant functionallity that we could pass to our Floatable interface instead of having it sit inside some utility class.

Answer 3

Let's begin with some background ...

Java doesn't support multiple inheritance (the ability to extend more than one class). This is because multiple inheritance is prone to the deadly diamond of death (also known as the diamond problem) which the designers of Java chose to preempt.

If B and C override a method inherited from A, which method does D inherit?

A class can implement multiple interfaces because interface methods are contracted for overriding; if a class C implements two interfaces A and B that declare the same method, then the same method in C will be invoked by clients of either interface (A or B). The introduction of default methods for interfaces in Java 8 was made possible by forcing the implementer to override the default in case of ambiguity. This was an acceptable compromise since default methods are intended to be defensive (to be used if no other implementation is explicitly provided by an implementer). However, since the compiler can’t force you to override a static method (static methods inherently can't be overridden), the introduction of static methods for interfaces in Java came with one restriction: the static methods of an interface are not inherited.

Answer 4

Here's my guess.

Since Cat can only extend one class if Cat extends Animal then Cat.identify has only one meaning. Cat can implement multiple interfaces each of which can have a static implementation. Therefore, the compiler would not know which one to choose?

However, as pointed out by the author,

Java already has this problem, with default methods. If two interfaces declare default void identify(), which one is used? It's a compile error, and you have to implement an overriding method (which could just be Animal.super.identify()). So Java already resolves this problem for default methods – why not for static methods?

If I was to guess again, I'd say that with default the implementation is part of Cat's vtable. With static it cannot be. The main function must bind to something. At compile time Cat.identify could be replaced with Animal.identify by the compiler but the code wouldn't match reality if Cat was recompiled but not the class that contains main.