Hacker News Reader

Comment by jwn | Hacker News Reader

Looking at http://openjdk.java.net/projects/valhalla/, I don't see a whole lot of value compared to JDK9. What's your basis for wanting 10 so badly?

View on HN

well as others already pointed out, value types. this changes the way how the JVM uses Memory in certain ways and might be a huge improvement for a lot of stuff.

specialized generics. this is a huge one, too, less boxing is always a win.

JEP-286 less typing.

Maybe Project Panama and maybe a even better AOT.

Compared to what JDK 9 brings, this is huge. JDK 9 brings a Module system which was already possible (and a lot of stuff around it which didn't exist), a new GC algorithm, Tiff Image I/O, jshell, ALPN, http2 client, reactive-streams. (P.S.: not everything is complete here, but this stuff will probably be used by most)

JDK9 mostly brings stuff that was already missing and provided by other libraries. JDK10 will probably change a lot of more things in the JVM ecosystem.

Good answer, thanks =) My question earlier was an honest question, asked out of curiosity...

If you want less typing, then you definitely want Lombok (https://projectlombok.org/). At JavaOne this year, they also showed a proposed shortcut for data object (@Data in Lombok).

  public class something(String name, int age) {
    ...
  }

I'm embarrassed to say that I've never understood the use case for Lombok. So, I write my code in Java with Lombok extensions. Then, another Java programmer goes to maintain it... and they have to learn Lombok?

That was my reaction the first time I was introduced to it. I looked at it again two years later and it just clicked. What's really cool is that the developer who has to go maintain it has 20% as much code to maintain.

Say you want a POJO with accessors, mutators along with equals and hashcode. Let it generate the methods by defining a class like:

  @Data
  public class something {
    String name;
    int age;
  }

Lombok generates:

  public String getName() { ... }
  public void setName(String name) { ... }
  public int getAge() { ... }
  public void setAge(int age) { ... }
  public boolean equals(Something other) { ... }
  public int hashcode() { ... }
  public String toString() { ... }

The real win is when you do Effective Java chapter 1 immutable objects with @Builder. I worked at a Big Dumb corp that eschewed lombok, and required full test coverage of getter, setters, builders, toString, equals, and hash mehtods, and 100% javadoc coverage for all methods (even private methods). It was insanity, and carpal tunnel inducing. Lombok would have reduced the workload by 90%.

Yes! ... And there are a lot of stages between a fully immutable object @Value and @Data (described above). The @Builder is so much better than most but you can also easily create a POJO with a portion of its fields immutable too (a combination of @NonNull, @Getter, @Setter and @RequiredArgsConstructor).

It might be worth a look at Immutables (http://immutables.github.io/) or FreeBuilder (https://github.com/google/FreeBuilder), both of which also use Java's built-in annotation processors to do code generation, but involve slightly less magic and produce generated classes which are not overwriting the original class.

One of the nice aspects of it is that if you have, say, a setter method which does something slightly unusual (e.g. setLastName does a call to setName(getFirstName() + " " +lastName) - unlikely example perhaps) then it stands out from the crowd of otherwise straightforward getters and setters.

But then you switch to JavaFX and you realize Lombok won't help you with JavaFX properties generation :(

I believe a similar argument could be made against using any 3rd party library. At some point who ever is maintaining your code will have to understand its dependencies at some level.

For lombok specifically, other languages have these types of features built in which of course people need to learn in order to use it.

I agree it isn't a perfect solution, but java is in dire need of boilerplate reduction features like this and lombok fills this need well.

if you want somthing that will send people over the edge and let you cackle with glee...

https://projectlombok.org/features/val.html

I've been doing C# for while... soo nice to have this. Coming back to java was like "why am I repeating myself". They already did the diamond operator on the right hand side, this makes the left just as nice.

now let the flames begin!!!

Oh also found a use for SneakyThrows today... I'm bootstrapping a test server in my unit test, where all of the code and interfaces are designed for safety and no-crash... and this lets me short circuit right out of it when something bad happens, which in a test setup is what I want.

Lombok is a bad idea, IMO.

Getters and Setters are verbose, but usually model classes are contained in a single class anyway, and can be generated automatically with any IDE. I can glance at a typical bean and ignore the code bloat without any problem.

The real issue is that Lombok generated classes don't play well with the dozens of libraries that expect Java bean style methods for auto marshaling and serialization: json generators, commons bean utils, binding libs, etc.

The savings aren't worth the trouble.

What problems do you have? Lombok does bytecode generation, I've never had a problem using it with things such as Jackson. How would an external library even know when the bytecode looks the same? I'm entirely aware I might be missing something obvious, this is a genuine question.

You can argue against Lombok for other reasons but was wondering about this specific one.

Everyone has their own priorities, but ValueTypes are huge.

Also, I don't think this falls under Valhalla (and I'm not sure if it's been committed yet), but the possibility of reasonable type inference in Java 10 would make me rather happy.

I'm pretty excited about that one, it's JEP-286, http://openjdk.java.net/jeps/286.

Valhalla is pretty big. Goetz described it as pulling on a very long string. It's going to touch everything in the JVM, including reifing generics, although as I understand it, erasure of Objects may be here to stay for the language.

Erased generics will surely stay and are a good thing. If they were removed, Scala would be in trouble (or rather, scalac would have to perform erasure itself, which in turn would mean that interop from Java wouldn't be good anymore).

What's broken in the JVM isn't erased generics but instead runtime reflection that is a lie due to erasure. Type erasure though is definitely an example of Java getting something very right for the wrong reasons.

Are you attempting to separate the concepts of specialization vs reified generics? Java does neither right now, and both effects are due to type erasure -- the runtime simply doesn't have the necessary information. I'm not quite sure where you're trying to draw the line.

Furthermore, runtime specialization (as opposed to compile-time) would seem to require reified types. So that further confuses things. But you can definitely have reified types without specialization, which I think is the point that you are trying to convey.

Aside from backwards compatibility, what's good about type erasure for generics?

Doing it a better way is hard

Is it, though? The most basic implementation, where things are still erased at runtime by the JIT, should be fairly simple, although it will not give the expected perf gains, of course.

Either way, CLR and C# did it a long time ago, and in the same time period when Java acquired its type-erased generics.

They care too much about backwards compatibility I think. .NET had done generics this way from the beginning.

It permits languages on the platform with substantially-richer compile-time type systems than the primary language on the platform with good interior stories (this was, IIRC, a substantial problem with Scala.NET.)

Value types combined with generic specialisation results in less memory usage, higher data locality, less gc pressure. Basically a big increase in performance on top of the already very good JIT.

JDK 10 will have TLS Fallback SCSV and native ChaCha20 support, though I wish they were in JDK 9.

Not sure why you have to wait. I use ChaCha in production in JDK 8: https://github.com/bcgit/bc-java/blob/master/core/src/main/j...

Why would native be any better? I would think that as long as the implementation matches the reference it wouldn't matter.

Last weekend, I added Bouncy Castle as the security provider in my dev environment, but my server did not show ChaCha20 in the list of available TLS ciphers. I was not convinced that my server was even using BC. Knowing that if you don't know what you're doing and mess around with crypto too much that you can get burned, I gave up. ECDHE RSA AES GCM for now.

I think I see where you're coming on that. For most of my applications, I terminate TLS connections with nginx. I do use ChaCha for high-speed encryption in other areas, though. The BC libraries have always worked great for me.

Reply on news.ycombinator.com