While the new features look great, I still can't help but feel like there's a lot of redundancy compared to things that are already commonplace in Scala. Fast forwarding 2 years, I think Kotlin will still be catching up with where Scala is today, and Scala may have migrated to Dotty / 3.0 by then.
Of course, there's still plenty of room for both languages, and I'm sure Kotlin will introduce interesting innovations of its own.
I don't think the intention for Kotlin is to have feature parity with Scala.
I feel that part of Scala's problem in every project I've been involved with is that it has too many features which means if you're not incredibly careful about specifying exactly what features should be used you end up with an unmaintainable mess.
> I feel that part of Scala's problem in every project I've been involved with is that it has too many features which means if you're not incredibly careful about specifying exactly what features should be used you end up with an unmaintainable mess.
Whenever I've asked people to go into the specific "too many features" in Scala it always turns out they were talking about library features, not language features. The language is powerful enough that people can write very complex libraries in it. If Kotlin ever becomes popular it will have exactly the same problem.
It's quite possible that in many senses Kotlin will overtake Scala. Maybe not in terms of advanced language features, but in terms of adoption, tooling.
Kotlin has things Scala doesn't, like efficient optionality in the type system, and better Java interop. These are things that the Scala guys probably aren't interested in adding, and likewise, the Kotlin guys don't want to add some things that Scala has, like implicits.
So thinking of it as a one dimensional line isn't really correct. Kotlin and Scala overlap in many ways, but the ways in which they don't (and will not) define them.
Scala does want efficient optionality in the type system - just not at the cost of a consistent representation of optionality. I'm sure an optimization for the common case would be very welcome.
I constantly see Kotlin advocates claiming better Java interop but I don't see how; Scala already has 100% Java interop as far as I can see.
Luckily JVM languages mostly interop with each other very well and the question is just convenience and efficiency.
For instance, Kotlin is using the same collections library as Java. So there are no adapters or conversions needed. Scala has its own collection library instead.
Scala has some constructs that can make it impossible or very difficult to call code from Java. Kotlin doesn't.
Those constructs are opt-in. People who are writing library code can choose to not use them. But in practice, people writing libraries in Scala that serve the Java community export a Java-specific API.
> For instance, Kotlin is using the same collections library as Java. So there are no adapters or conversions needed. Scala has its own collection library instead.
You can write Scala using the standard Java collections (or Guava immutable collections) if you like (especially if you're using a ScalaZ-heavy style where everything happens via typeclasses - just write typeclass instances for the collections you're using, and then it's easy to mix and match). For my money immutability is too important for it to be worth using the Java collections - even in Java I'll use the Guava ImmutableList etc.
> Scala has some constructs that can make it impossible or very difficult to call code from Java. Kotlin doesn't.
Such as? I mean, implicit parameters won't be implicit (but they couldn't be in Kotlin either), and higher-kinded types compile to raw methods for which you have to cast (but you'd have to do the same to express the same thing in Kotlin).
For instance, Kotlin is using the same collections library
as Java. So there are no adapters or conversions needed.
Scala has its own collection library instead.
Years ago Scala tried the same thing Kotlin currently tries to sell as the hot new stuff. That approach was abandoned altogether, because. it. just. doesn't. work.
It's a death by thousand cuts. You can keep sprinkling compiler magic on top of it, but it will be an ongoing game of whack-a-mole. See all the typesystem unsoundness and how they had to abandon (non-)null annotations for existing code because it just didn't scale/work.
The whole concept of having mutable and immutable collections within the same API hierarchy that Kotlin tries to sell is also considered to be one of the remaining design issues of Scala collections.
Those were terrible times. So happy that Scala doesn't do that anymore!
Actually Scala 2.12 will have more Java 8 Interop than Kotlin.
Scala will hit soon while I don't see Kotlin Java7/Java8 soon not even a Milestone release yet.
For reference, M4 was released 3 days ago. JDK8 was released a little over 3 years ago, and JDK9 is about a year away which will introduce some interesting challenges.
We've yet to see how fast JetBrains will iterate on Kotlin and support for JDK8, while we know exactly how fast Scala iterates.
Yeah the schedule change was sad. And Scala 2.12 takes really long, but a Scala 2.11 is really solid, it just hasn't the same interop than it should have in a Java 8 world.
Scala Some is a box (i.e. it takes up 8 bytes on top of whatever's in it). It would be nice to represent Some(x) as just the value x (at the Java bytecode level) and None as null. The problem with that is that if you have an Option[Option[A]] you can't tell the difference between None and Some(None) (which might be semantically important), and whatever you do to solve this your Option is no longer generic and behaves in different ways depending on what's inside it, which is a disaster for being able to understand and reason about code.
Kotlin treats nullability as an ad-hoc special case. This allows them to save those 8 bytes occasionally, but it means you simply can't handle optionality generically (in Scala you can write methods that will work generically with Option and also with other cases, e.g. "sequence" does the same thing when you call it on a List[Option[Int]] or a List[Future[Int]] or a List[Writer[String, Int]] or... and it all makes sense, because Option is just another plain old type in the language). IMO that's the wrong tradeoff for almost all use cases, but evidently the Kotlin folk disagree.
• 4 bytes to point to it, unless you need a big heap and OOP compression isn't usable, in which case it's 8 just on the pointer.
• Either 4 or 8 bytes of mark word, depending again on 32 vs 64 bit mode.
• Either 4 or 8 bytes of class pointer, ditto.
• Then another 4 or 8 bytes for the pointer inside the box.
• GC algorithms impose some additional overhead too.
That's hoping there's no alignment padding going on.
If the JVM supported real value types, then a lot of this overhead would boil away, but not always all of it.
When you use Option[] as a local variable or as a return type of a method, then HotSpot can sometimes optimise it out using escape analysis. Unfortunately the escape analysis in the C2 compiler is quite conservative and can often fail. The Graal compiler has a different design for its escape analysis and can remove the overhead a lot more often. Graal does show better speedups on Scala code than on Java code:
The ability to nest optionalities is not something I've wanted to do so far, whereas the ability to freely represent optionality without having to worry about cost is quite freeing. I think Kotlin has the right tradeoff here, especially as in the rare cases where you do want the ability to represent Optional<Optional<T>> you can just use the Optional class from the Java 8 standard library to do so.
I very rarely want to explicitly use an Optional<Optional<T>>. But every day I write code that works with a generic Option[A] and I want to not have to worry about what happens if someone calls it with A=Option[B]. I also frequently use Option as a parameter to some other datastructure like Kleisli or WriterT, which only works because it's an ordinary datatype that behaves like any other.
You say "you can just use the Optional class from the Java 8 standard library", but as far as I can see porting code that was written to use nullables would be a pretty big refactor, and there's no way at all to write generic code that works with both, so you'd end up having to maintain two parallel copies of your common functions.
Efforts to improve the performance of Option are of course to be applauded, but for a lot of use cases we're talking about the difference between (say) 50x faster than Ruby and 40x faster than Ruby. Scala is more than fast enough for a huge range of use cases; I don't think I've ever seen someone port code off Scala because the runtime performance wasn't good enough.
You're basically talking about using an untagged union as an optional type instead of a tagged union. Last I checked, Dotty will have union types [1], so we may well see that become a more common way to express optionality. Or not. Either way, that's a pretty small distinction.
I haven't programmed Kotlin before, so I'm not sure what their answer to implicits is. For ad hoc polymorphism, none of the alternatives is particularly pretty. Perhaps the current answer is to resort to adapter classes, which would be the way to do it in Java.
This is awesome news. yield-return semantics was something I missed dearly.
Note that we’re still in the process of estimating
the effort needed to implement this feature,
and we don’t know whether it would be reasonable
to support it in the 1.1 timeframe or it would be
postponed to a later release.
:) I can already tell you: yield is probably doable, async-await will have to be postponed.
As the post says, yield and async/await will be based on exactly the same mechanism. Either the entire mechanism will be implemented, or it will be postponed.
All JVM languages "are" the same speed, because they compile to the same bytecode.
OTOH some languages make it more practical / cheaper to write higher-performance code, e.g. in Scala you can use @specialized to automatically generate unboxed special cases for generic code that you'd have to write (and keep track of) by hand in Java. But if you ask "in which JVM language is it most practical/cheapest to write feature X with performance Y?" then the biggest component of the answer is going to be how easy programming in general in that language is, not anything specific to performance.
The question "how fast will a naive implementation in language X perform?" does have an answer, but it's not a question you should ever be asking - if a naive implementation in language X is faster than a naive implementation in language Y, but takes 3x as long to write as an optimised implementation in language Y that's faster than the naive implementation in language X, then language Y is the better choice.
They compile to the same bytecode, but some languages may have features that compile to a lot more bytecode than others. A dynamic language such as Clojure, for example, is definitely slower than Java, although they are both JVM-based, because it has to do a lot more.
Kotlin relies quite heavily on bytecode level method inlining, for instance.
The advantage is that it makes code written in a functional style effectively free (well, almost, unless you could gain performance from step fusion). The disadvantage is that the JVM wasn't designed for that and it messes up stack traces.
I'd be satisfied with it not being noticeably slower than Java and according to some it isn't. Do you speak from experience or just assume that it must be slower?
Yes, the plan is for Kotlin to replace Java at JetBrains. I don't see how the performance requirement follows from that. The performance of JetBrains products is determined far more by the implementation algorithms than by the compiler used to build their source code. Kotlin's performance is on par with Java, and that is all that JetBrains needs for its internal use.
I'm very excited for 1.0.x, type aliases and a library for concurrent programming are great news. Not to mention support for Java 7-9 bytecode generation.
40 comments
[ 2.4 ms ] story [ 112 ms ] threadOf course, there's still plenty of room for both languages, and I'm sure Kotlin will introduce interesting innovations of its own.
I feel that part of Scala's problem in every project I've been involved with is that it has too many features which means if you're not incredibly careful about specifying exactly what features should be used you end up with an unmaintainable mess.
Whenever I've asked people to go into the specific "too many features" in Scala it always turns out they were talking about library features, not language features. The language is powerful enough that people can write very complex libraries in it. If Kotlin ever becomes popular it will have exactly the same problem.
The funny thing is that even as people say that, every new release tends to look a step closer to Scala.
So thinking of it as a one dimensional line isn't really correct. Kotlin and Scala overlap in many ways, but the ways in which they don't (and will not) define them.
I constantly see Kotlin advocates claiming better Java interop but I don't see how; Scala already has 100% Java interop as far as I can see.
For instance, Kotlin is using the same collections library as Java. So there are no adapters or conversions needed. Scala has its own collection library instead.
Scala has some constructs that can make it impossible or very difficult to call code from Java. Kotlin doesn't.
You can write Scala using the standard Java collections (or Guava immutable collections) if you like (especially if you're using a ScalaZ-heavy style where everything happens via typeclasses - just write typeclass instances for the collections you're using, and then it's easy to mix and match). For my money immutability is too important for it to be worth using the Java collections - even in Java I'll use the Guava ImmutableList etc.
> Scala has some constructs that can make it impossible or very difficult to call code from Java. Kotlin doesn't.
Such as? I mean, implicit parameters won't be implicit (but they couldn't be in Kotlin either), and higher-kinded types compile to raw methods for which you have to cast (but you'd have to do the same to express the same thing in Kotlin).
It's a death by thousand cuts. You can keep sprinkling compiler magic on top of it, but it will be an ongoing game of whack-a-mole. See all the typesystem unsoundness and how they had to abandon (non-)null annotations for existing code because it just didn't scale/work.
The whole concept of having mutable and immutable collections within the same API hierarchy that Kotlin tries to sell is also considered to be one of the remaining design issues of Scala collections.
Those were terrible times. So happy that Scala doesn't do that anymore!
I think generally when people say this they mean things like:
1. You can freely mix and match Java and Kotlin files in the same package
2. Being a more lightweight layer on top of Java makes interop with all Java environments (specifically Android) more practical
The original roadmap: http://www.scala-lang.org/news/2.12-roadmap/ The current roadmap: http://www.scala-lang.org/news/2016-schedule/
For reference, M4 was released 3 days ago. JDK8 was released a little over 3 years ago, and JDK9 is about a year away which will introduce some interesting challenges.
We've yet to see how fast JetBrains will iterate on Kotlin and support for JDK8, while we know exactly how fast Scala iterates.
Using Java 8 has worked with 2.11 for quite some time already.
Honest question - what is it that makes Scala's optionality inefficient compared to Kotlin? I'd never heard of this before.
Kotlin treats nullability as an ad-hoc special case. This allows them to save those 8 bytes occasionally, but it means you simply can't handle optionality generically (in Scala you can write methods that will work generically with Option and also with other cases, e.g. "sequence" does the same thing when you call it on a List[Option[Int]] or a List[Future[Int]] or a List[Writer[String, Int]] or... and it all makes sense, because Option is just another plain old type in the language). IMO that's the wrong tradeoff for almost all use cases, but evidently the Kotlin folk disagree.
A box takes at least:
• 4 bytes to point to it, unless you need a big heap and OOP compression isn't usable, in which case it's 8 just on the pointer.
• Either 4 or 8 bytes of mark word, depending again on 32 vs 64 bit mode.
• Either 4 or 8 bytes of class pointer, ditto.
• Then another 4 or 8 bytes for the pointer inside the box.
• GC algorithms impose some additional overhead too.
That's hoping there's no alignment padding going on.
If the JVM supported real value types, then a lot of this overhead would boil away, but not always all of it.
When you use Option[] as a local variable or as a return type of a method, then HotSpot can sometimes optimise it out using escape analysis. Unfortunately the escape analysis in the C2 compiler is quite conservative and can often fail. The Graal compiler has a different design for its escape analysis and can remove the overhead a lot more often. Graal does show better speedups on Scala code than on Java code:
http://lampwww.epfl.ch/~hmiller/scala2013/resources/pdfs/pap...
(old paper)
The ability to nest optionalities is not something I've wanted to do so far, whereas the ability to freely represent optionality without having to worry about cost is quite freeing. I think Kotlin has the right tradeoff here, especially as in the rare cases where you do want the ability to represent Optional<Optional<T>> you can just use the Optional class from the Java 8 standard library to do so.
You say "you can just use the Optional class from the Java 8 standard library", but as far as I can see porting code that was written to use nullables would be a pretty big refactor, and there's no way at all to write generic code that works with both, so you'd end up having to maintain two parallel copies of your common functions.
Efforts to improve the performance of Option are of course to be applauded, but for a lot of use cases we're talking about the difference between (say) 50x faster than Ruby and 40x faster than Ruby. Scala is more than fast enough for a huge range of use cases; I don't think I've ever seen someone port code off Scala because the runtime performance wasn't good enough.
I haven't programmed Kotlin before, so I'm not sure what their answer to implicits is. For ad hoc polymorphism, none of the alternatives is particularly pretty. Perhaps the current answer is to resort to adapter classes, which would be the way to do it in Java.
[1] https://d-d.me/talks/scalaworld2015/#/6
if yes, then i would assume the primary objective should be performance .. but this doesnt seem to be the case
and my question is, is there any jvm language which is faster than java
OTOH some languages make it more practical / cheaper to write higher-performance code, e.g. in Scala you can use @specialized to automatically generate unboxed special cases for generic code that you'd have to write (and keep track of) by hand in Java. But if you ask "in which JVM language is it most practical/cheapest to write feature X with performance Y?" then the biggest component of the answer is going to be how easy programming in general in that language is, not anything specific to performance.
The question "how fast will a naive implementation in language X perform?" does have an answer, but it's not a question you should ever be asking - if a naive implementation in language X is faster than a naive implementation in language Y, but takes 3x as long to write as an optimised implementation in language Y that's faster than the naive implementation in language X, then language Y is the better choice.
The advantage is that it makes code written in a functional style effectively free (well, almost, unless you could gain performance from step fusion). The disadvantage is that the JVM wasn't designed for that and it messes up stack traces.