5

Was introducing default methods to java inevitable? As far as I know multiple class inheritance was not introduced to avoid difficulties with the method signature clash in base classes.

So we avoided it with classes and on the other hand we introduce this problem with default methods as we have to override default method if it exists with the same signature in two or more interfaces that we implement.

Is the cost of updating old systems to newer versions of interfaces without using default methods greater than cost of using language with garbage in its syntax?

7
  • 1
  • 1
    @gnat Wrong. This has an answer that is not opinion. – JimmyJames Nov 12 '20 at 19:53
  • "Is the cost of updating old systems to newer versions of interfaces without using default methods greater than cost of using language with garbage in its syntax?" <-- per my reading this is as opnon based as it gets @JimmyJames (you've got to see Clinton Begin's (of myBatis) essay on Java 5 features to see how much opinion is in stuff like that) – gnat Nov 12 '20 at 19:59
  • 1
    @gnat I really don't understand the obsession with killing questions like this. It's a perfectly reasonable question to ask given how well Java is/was known for disallowing multiple inheritance. And the answer is illuminating with regard to how to use the new feature effectively. – JimmyJames Nov 12 '20 at 20:14
  • 2
    @JimmyJames just because gnat is decent enough to comment when he votes to close is no reason to pick on him. Debate only on the merits here. If you think practice needs to change take it to meta. – candied_orange Nov 12 '20 at 20:28
12

The reason that default methods were introduced was that without them, it's impossible to add new methods to an interface and maintain backwards compatibility. As explained here:

Default methods enable you to add new functionality to the interfaces of your libraries and ensure binary compatibility with code written for older versions of those interfaces.

They were introduced in Java 8 along with the stream APIs. Looking at the default methods that were added to Collection, it starts to become clear. The addition of the streaming methods such as stream, parallelStream, and spliterator would break every existing Collection implementation outside of the JDK.

This is clearly the primary driver of the addition. I think the worries around them have proven to be unwarranted. There is no ambiguity in the spec around how clashes are handled. They also have some nice features aside from the above. For example, there's no longer a need to implement a remove method in an Iterator implementation because it has a default implementation.

So to the point of your question which I initially failed to answer:

Is the cost of updating old systems to newer versions of interfaces without using default methods greater than cost of using language with garbage in its syntax?

So whether default methods are 'garbage' is highly subjective but the first part is answerable. The Collection interface is one of the most implemented interfaces in Java code. There is a truly immense amount of code that would need to be updated if a non-backwards-compatible changes were made to it. The cost seems immeasurable and I think the result would be that a lot of Java code would never be able to run on 1.8 or better given it's history of backwards-compatibility.

We could ask whether new features such as this could be added in another way, such as by using wrappers. Perhaps but such an approach tends to be unwieldy and create a lot of extra code. In the years since this was introduced, I've seen no evidence that it has caused any significant issues.

2
  • Thank you for this comment. It made me aware how immense amount of work would be needed to adjust everything to new methods in interfaces. Maybe my initial question should be "Is there some other reason for disallowing multiple class inheritance than method signature clashes in base classes". If not then I think it is inconsistency to disallow this in classes and allow in interfaces and I do not see a reason to keep it disallowed for class case. – bridgemnc Nov 13 '20 at 18:42
  • 2
    @bridgemnc It's valid to ask why, if we think it's OK in interfaces, why not in classes? It's important though to be clear that what you can do with default methods in interfaces is quite limited. Firstly, you cannot directly access object state in a default method. You can reference other methods within the interface and these are always ultimately controlled by the (single) implementing class. I think this limited scope is a smart design and why this has not created challenges. You can't really impose those constraints on classes without some fundamental language changes. – JimmyJames Nov 13 '20 at 19:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.