I appreciate the hard work that went into the things that did make it into Valhalla eventually, but:
> The model was powerful, but also mentally heavy
No it isn't! it is this interpretation that kills off the null-safety debate entirely. Saying you have a variable that cannot be null is not a mentally taxing distinction, especially since everything is labelled thoroughly.
> The team, faithful to the lesson “simplify the model for the user, even at the cost of the performance ceiling,” ultimately dismantled this dualism.
but it would have simplified it for the user.
The whole attitude and process around this and the other topics gives me very little faith that Java can be steered in a sensible direction here. The type system of a programming language is supposed to give convenient guarantees to the developer on a CPU that can only do numbers. There is no reason to reduce the optional(!) safety guarantees you can offer with the excuse of "too mentally taxing".
Hell, they even get there half way by recognising:
> the language model and the JVM model don’t have to overlap one hundred percent
> The whole attitude and process around this and the other topics gives me very little faith that Java can be steered in a sensible direction here.
I agree. The stewardship of Java seems rather lacking - particularly when compared to that of .net, where MS etc. mostly seemed to make the correct decisions from the start.
Does Java even have any value or mindshare at Oracle nowadays? The company seems to be a datacentre/compute business at this point, with appendiges for its legacy activities and a vast overhang of debt.
I sometimes wonder if the only parts of Oracle that are still profitable are the Legal and Lawnmower divisions.
In what way? If anything Java's main developers (employed by Oracle for the most part, working on the completely open source and free OpenJDK) are extremely knowledgeable and are responsible a big jump in how fast the platform evolves. They have added proper algebraic data types to the language, delivered virtual threads and garbage collectors that decouple pause times from heap size. Like if anything, Java is at the best place it has ever been.
> They have added proper algebraic data types to the language
No they haven't. E.g. they added a class that superficially looks like Option but subtly breaks the rules that Option is meant to follow, ensuring that no-one can ever manage to migrate existing codebases away from using `null`.
Sealed classes/interfaces and records are proper sum and product types.
The stdlib's Option type predates this language update by a long shot, so it doesn't use sealed classes, but it is now possible to have the usual FP "Maybe" type in Java:
```
sealed class Maybe<T> permits Some, None {
record Some<T>(T obj) {}
record None() {}
}
```
(You will probably have to write Maybe.Some and I might have messed up the generic syntax as I wrote it on my phone, but that's mostly how it looks)
How .net got so many things right where java did not is a mystery to me, but appreciated (it has its own flaws, of course). Java, in my understanding, is still of core relevance to Oracle, and tied into a lot of contracts that require very little effort from them to maintain. But you are correct in observing that they want to be a datacentre/compute business more and more these days; they may have in fact overcomitted to this due to the AI craze, since shareholders are already complaining.
> But careful: == looks at internal state, which isn’t always what the object represents, so for “is this the same data” comparisons keep using equals.
So == for value classes will basically be like memcmp(). That is a bit unfortunate, as it breaks encapsulation, exposing implementation details. Client code can use this to do case distinctions based on how a given value is internally represented. In a way, it’s worse than identity comparison, because identity comparison at least doesn’t expose internal state.
Value types are a concept very far away from the "magic black box organism" school of OOP thinking. It's not a novel way of doing classic OOP (does anyone still do that?), it's a way for a language born in OOP ideology get one step further into the post-OOP world.
If your bags of data have internal state, there's something wrong with your bags of data. I assume that the Java guys thought far enough to either exclude padding from comparisons or force padding bytes to be zero.
It should work even for strings: They will surely continue to heap-allocated, and memcmp-ing pointers (inside the new "structs") is exactly an identity comparison.
There’s nothing wrong with having non-normalized representations, that’s why there is equals().
For example, you might have a value class for representing (limited-precision) fractions using two longs internally, for the numerator and denominator. For efficiency trade-off reasons, you don’t want to always shorten the fraction. But now client code can distinguish 2/3 from 4/6 using ==.
Scenarios of that sort are conceivable where this actually leaks sensitive information. In any case, it creates dependencies on implementation details where you don’t want to have them.
When designing a value class, you are now in the dilemma of either always having to normalize the representation, costing performance, or having your class be a funnel for leaking implementation details.
Well. I'd be upset if custom operator==() for plain-old-data structs was removed from C++, but Java never had it to begin with, so for Java, it just means that you have to fall back to using traditional classes (or compare using something other than ==) if you need such "fancy" features.
A lot of the comments on here are a bit unfair on what is great work being done and even more awesome work (JEPs) in the pipeline for the future.
If Java was a child, imagine it being brought up by loving parents for the first few years (Sun) then it was thrown in a garage with some other children and neglected by its evil guardian (Oracle)
Neglected and unloved till JDK 8, its basically been playing catch up.
So when people say "oh so its now got structs or value types of X", yes it has but that's because it has been stunted in its development due to big bureaucratic and hostile corporate processes, but its free now and is getting love through the OpenJDK family.
I will continue to enjoy writing once and deploying anywhere!
You could probably a whole tech thriller on the evolution on Value Types in Java.
I’ve been reading the mailing lists and watched all videos on the topic and it is truly inspiring how much they managed to consolidate the design to something that always looked like java.
But while also going far deeper in granularity and understanding what it even means to be a value type and what optimizations can be done where
I don’t see a difference between pass by copy and pass by value.
The mutability difference is that part of a struct can be modified in place, which value classes can’t: the value of a complete value-class variable (or array slot) can only be modified (reassigned) as a whole. This is presumably because object references to value-class objects can be created, and those objects should be immutable so their identity doesn’t matter.
I think that's mostly a semantic difference - Java avoided the problem of strange lifetimes, captures, tearing by fixing the semantics as immutable value objects, while C# has to deal with these issues.
But under the hood it can (and will) do a modification in place.
Functionally they don't - java is just catching up with (by now) ancient practice.
The false dichotomy of
> A struct in C# has identity and mutation, so the semantics of copying on assignment or passing have to be precisely defined, which gives a heavier model for the programmer and less freedom for the runtime.
Doesn't really match with what they're describing. While yes, it will not have identity in a java class ref sense, it of course will still have identity in being a unique structure in memory at a certain address. This is just splitting hairs about Java nomenclature.
> it of course will still have identity in being a unique structure in memory
No, it will not. The design allows multiple objects to share one structure in memory across multiple records, or not have such a structure at all (see Scalarization in the article).
I don't wanna badmouth Java people, but how they push the idea that this thing is some sort of genuine breakthrough that took multiple PhDs years of cutting-edge research to implement, when in fact they basically copied what .NET did from basically year 1, is not a good look.
Again, not trying to turn this into a .NET vs Java thing, I'd have been much happier if they reached some new and interesting conclusions.
Well, it is - because they had to make it with almost perfect backwards compatibility for one of the most popular languages with trillions of lines of code produced over decades.
Sure, adding it to a new language is not hard. Adding it to Java which has primitives, generics and boxing, finding ways that seamlessly cover the differences between objects and primitives, while trying to plan for the future is hard.
As a general note, if you come to the conclusion that one of the best designer teams on Earth "basically copied what .NET did from year 1 is not a good look", then maybe your mental model needs adjusting on how these stuff works? Java has a public mailing list, you can browse through the related discussions. Implementation is the least of these things. But I can assure you they most definitely know what they are doing.
Correct me if I'm wrong, but Rust editions are a source code-level feature. So given you have the source code of newer and older rust code, you can compile them together.
That's materially distinct from Java's model of basically dynamic loading already compiled class files. Though class files do have "editions", and there are extra code to deal with different versions. But still, it should be possible to e.g. send a new value class to an old library's class that has never heard of them, and that should just work.
This is exactly what made it so difficult. It is much easier to have a feature like this from year 1 than to add it to a language that has grown and evolved for 18 years already.
I agree with this sentiment. The work they put in deserves a lot of respect, and took a lot of effort, no doubt. It's just the framing they push to the public that could use some work.
I get that this makes objects behave like primitive types. Maybe thats reason enough. But is it necessary for the performance boost and de-fluffing the objects? Seems like an orthogonal objective
> There’s a catch worth knowing about here, though: flattened data has to be readable and writable atomically (otherwise it risks “tearing” under concurrent access).
Isn't this a race condition and "undefined bahvior"..? Having to limit yourself to atomic sizes seems like a huge limitation, to accomodate what is most likely buggy code. Is all the effort only gunna help lil toy ColorRGB examples?
> The points array is a million pointers. Each pointer leads to a separate Point object lying somewhere on the heap.
Does this happen in actuality? One would assume the allocator tries to put stuff sequentially on the heap? Its not a guarantee as with these Value Types, but I'd think you could get similar-ish perf with prefetching in cache. I dunno whats happening under the hood.. But when writing Clojure apps the JVM always reserves absurd amounts of heapspace on my machine (to my annoyance). Id assume it can find some place to do contiguous allocations..
Which i guess gets me to my last question... where are the benchmarks broski? It all sounds great, but does it actually yield the insane speedups promised?
Great article, well written. But a benchmark would have been a nice "punchline"
> is it necessary for the performance boost and de-fluffing the objects?
Yes. The one part of the JVM GC that can't run concurrently is heap compaction; objects that can be moved by copying and then deleting would be a huge help for that. And it would be awkward to say the object has an identity but can't be wait/notify'd, at which point you need somewhere for the monitor to go.
> Does this happen in actuality? One would assume the allocator tries to put stuff sequentially on the heap?
Yes. Of course it tries, but semantically the pointers are just pointers and the prefetcher can guess but the system still has to chase them.
> Is all the effort only gunna help lil toy ColorRGB examples
Arguably flattening mostly makes sense for these only.
And yeah, you are right that allocations happen on something called a thread local allocation buffer, which is basically just a pointer bump in cost and objects allocated one after the other should be physically close in memory for the most part (though an object's creation may require a bunch of other object's creation that would sit in-between). But these have headers, so not as dense as they could be (though due to GCs being generational, they may end up actually closer in the next gen? The in-between temporary objects wouldn't survive for the most part)
> I get that this makes objects behave like primitive types. Maybe thats reason enough. But is it necessary for the performance boost and de-fluffing the objects? Seems like an orthogonal objective
It feels like an orthogonal objective and honestly arbitrary distinction, yes.
> Isn't this a race condition and "undefined bahvior"..? Having to limit yourself to atomic sizes seems like a huge limitation, to accomodate what is most likely buggy code.
I think they meant it like the appearance of atomic behavior from a java multithreading view.
> Does this happen in actuality?
Yes, it does happen. Having guarantees on this front leads to better performance.
> But when writing Clojure apps the JVM always reserves absurd amounts of heapspace on my machine (to my annoyance)
It would perhaps make Java look a little bad to say that, in 1995, the prevailing attitude in certain circles was "If it's too slow, just wait for faster hardware - Moore's Law forever baby!"
(Of course, Sun was selling, at the time, relatively fast hardware - the slower the software, the faster the required hardware)
Yes, this also stood out to me. I usually think of CPUs and memory having parity in the early 80s, but I never bothered to check for sure. I do remember some early computer architects writing about memory being faster than the CPU!
I appreciate the hard work that went into the things that did make it into Valhalla eventually, but:
> The model was powerful, but also mentally heavy
No it isn't! it is this interpretation that kills off the null-safety debate entirely. Saying you have a variable that cannot be null is not a mentally taxing distinction, especially since everything is labelled thoroughly.
> The team, faithful to the lesson “simplify the model for the user, even at the cost of the performance ceiling,” ultimately dismantled this dualism.
but it would have simplified it for the user.
The whole attitude and process around this and the other topics gives me very little faith that Java can be steered in a sensible direction here. The type system of a programming language is supposed to give convenient guarantees to the developer on a CPU that can only do numbers. There is no reason to reduce the optional(!) safety guarantees you can offer with the excuse of "too mentally taxing".
Hell, they even get there half way by recognising:
> the language model and the JVM model don’t have to overlap one hundred percent
> The whole attitude and process around this and the other topics gives me very little faith that Java can be steered in a sensible direction here.
I agree. The stewardship of Java seems rather lacking - particularly when compared to that of .net, where MS etc. mostly seemed to make the correct decisions from the start.
Does Java even have any value or mindshare at Oracle nowadays? The company seems to be a datacentre/compute business at this point, with appendiges for its legacy activities and a vast overhang of debt.
I sometimes wonder if the only parts of Oracle that are still profitable are the Legal and Lawnmower divisions.
> The stewardship of Java seems rather lacking
In what way? If anything Java's main developers (employed by Oracle for the most part, working on the completely open source and free OpenJDK) are extremely knowledgeable and are responsible a big jump in how fast the platform evolves. They have added proper algebraic data types to the language, delivered virtual threads and garbage collectors that decouple pause times from heap size. Like if anything, Java is at the best place it has ever been.
> They have added proper algebraic data types to the language
No they haven't. E.g. they added a class that superficially looks like Option but subtly breaks the rules that Option is meant to follow, ensuring that no-one can ever manage to migrate existing codebases away from using `null`.
Sealed classes/interfaces and records are proper sum and product types.
The stdlib's Option type predates this language update by a long shot, so it doesn't use sealed classes, but it is now possible to have the usual FP "Maybe" type in Java:
``` sealed class Maybe<T> permits Some, None { record Some<T>(T obj) {} record None() {} } ```
(You will probably have to write Maybe.Some and I might have messed up the generic syntax as I wrote it on my phone, but that's mostly how it looks)
optional is not how algebraic data types are implemented in Java. Basically it's the combination of sealed types and records.
How .net got so many things right where java did not is a mystery to me, but appreciated (it has its own flaws, of course). Java, in my understanding, is still of core relevance to Oracle, and tied into a lot of contracts that require very little effort from them to maintain. But you are correct in observing that they want to be a datacentre/compute business more and more these days; they may have in fact overcomitted to this due to the AI craze, since shareholders are already complaining.
> particularly when compared to that of .net, where MS etc. mostly seemed to make the correct decisions from the start.
Wut? I did worked on .net projects and all it achieved was making me like java a lot more then previously.
This is mainly for performance and memory layouts, it would not have improved safety guarantees of java.
It would have implicitly brought some null-safety to java with primitive-like classes that can not be null.
> But careful: == looks at internal state, which isn’t always what the object represents, so for “is this the same data” comparisons keep using equals.
So == for value classes will basically be like memcmp(). That is a bit unfortunate, as it breaks encapsulation, exposing implementation details. Client code can use this to do case distinctions based on how a given value is internally represented. In a way, it’s worse than identity comparison, because identity comparison at least doesn’t expose internal state.
Value types are a concept very far away from the "magic black box organism" school of OOP thinking. It's not a novel way of doing classic OOP (does anyone still do that?), it's a way for a language born in OOP ideology get one step further into the post-OOP world.
If your bags of data have internal state, there's something wrong with your bags of data. I assume that the Java guys thought far enough to either exclude padding from comparisons or force padding bytes to be zero.
It should work even for strings: They will surely continue to heap-allocated, and memcmp-ing pointers (inside the new "structs") is exactly an identity comparison.
There’s nothing wrong with having non-normalized representations, that’s why there is equals().
For example, you might have a value class for representing (limited-precision) fractions using two longs internally, for the numerator and denominator. For efficiency trade-off reasons, you don’t want to always shorten the fraction. But now client code can distinguish 2/3 from 4/6 using ==.
Scenarios of that sort are conceivable where this actually leaks sensitive information. In any case, it creates dependencies on implementation details where you don’t want to have them.
When designing a value class, you are now in the dilemma of either always having to normalize the representation, costing performance, or having your class be a funnel for leaking implementation details.
Well. I'd be upset if custom operator==() for plain-old-data structs was removed from C++, but Java never had it to begin with, so for Java, it just means that you have to fall back to using traditional classes (or compare using something other than ==) if you need such "fancy" features.
A lot of the comments on here are a bit unfair on what is great work being done and even more awesome work (JEPs) in the pipeline for the future.
If Java was a child, imagine it being brought up by loving parents for the first few years (Sun) then it was thrown in a garage with some other children and neglected by its evil guardian (Oracle)
Neglected and unloved till JDK 8, its basically been playing catch up.
So when people say "oh so its now got structs or value types of X", yes it has but that's because it has been stunted in its development due to big bureaucratic and hostile corporate processes, but its free now and is getting love through the OpenJDK family.
I will continue to enjoy writing once and deploying anywhere!
You could probably a whole tech thriller on the evolution on Value Types in Java.
I’ve been reading the mailing lists and watched all videos on the topic and it is truly inspiring how much they managed to consolidate the design to something that always looked like java.
But while also going far deeper in granularity and understanding what it even means to be a value type and what optimizations can be done where
I know its a faux pas in the Java world to acknowledge the existence of .NET, but how does this differ from .NET structs?
Value types, generic specialization, boxing - a quick skim makes it looks like they picked the same choices.
The article has a section about that.
For me, a struct in C/C# can be modified and is passed by copy while a value class can not be modified and is passed by value.
I do not think you can do stack allocation in Java.
I don’t see a difference between pass by copy and pass by value.
The mutability difference is that part of a struct can be modified in place, which value classes can’t: the value of a complete value-class variable (or array slot) can only be modified (reassigned) as a whole. This is presumably because object references to value-class objects can be created, and those objects should be immutable so their identity doesn’t matter.
I think that's mostly a semantic difference - Java avoided the problem of strange lifetimes, captures, tearing by fixing the semantics as immutable value objects, while C# has to deal with these issues.
But under the hood it can (and will) do a modification in place.
I think pass by copy is a consequence of being modifiable.
The other solution is to stack allocate and pass a pointer but as i said, unlike in C#, i do not think it's possible to do that in Java.
In Go, you can stack allocate but when you send a pointer (that escapes), the compiler will heap allocate the object.
My point is that pass by copy and pass by value do the same thing, they copy the value representation.
Functionally they don't - java is just catching up with (by now) ancient practice.
The false dichotomy of
> A struct in C# has identity and mutation, so the semantics of copying on assignment or passing have to be precisely defined, which gives a heavier model for the programmer and less freedom for the runtime.
Doesn't really match with what they're describing. While yes, it will not have identity in a java class ref sense, it of course will still have identity in being a unique structure in memory at a certain address. This is just splitting hairs about Java nomenclature.
> it of course will still have identity in being a unique structure in memory
No, it will not. The design allows multiple objects to share one structure in memory across multiple records, or not have such a structure at all (see Scalarization in the article).
I don't wanna badmouth Java people, but how they push the idea that this thing is some sort of genuine breakthrough that took multiple PhDs years of cutting-edge research to implement, when in fact they basically copied what .NET did from basically year 1, is not a good look.
Again, not trying to turn this into a .NET vs Java thing, I'd have been much happier if they reached some new and interesting conclusions.
> genuine breakthrough
Well, it is - because they had to make it with almost perfect backwards compatibility for one of the most popular languages with trillions of lines of code produced over decades.
Sure, adding it to a new language is not hard. Adding it to Java which has primitives, generics and boxing, finding ways that seamlessly cover the differences between objects and primitives, while trying to plan for the future is hard.
As a general note, if you come to the conclusion that one of the best designer teams on Earth "basically copied what .NET did from year 1 is not a good look", then maybe your mental model needs adjusting on how these stuff works? Java has a public mailing list, you can browse through the related discussions. Implementation is the least of these things. But I can assure you they most definitely know what they are doing.
idk maybe java should adopt something similar to rust's "edition"?
Correct me if I'm wrong, but Rust editions are a source code-level feature. So given you have the source code of newer and older rust code, you can compile them together.
That's materially distinct from Java's model of basically dynamic loading already compiled class files. Though class files do have "editions", and there are extra code to deal with different versions. But still, it should be possible to e.g. send a new value class to an old library's class that has never heard of them, and that should just work.
This is exactly what made it so difficult. It is much easier to have a feature like this from year 1 than to add it to a language that has grown and evolved for 18 years already.
I agree with this sentiment. The work they put in deserves a lot of respect, and took a lot of effort, no doubt. It's just the framing they push to the public that could use some work.
> Will I get a fast, flat `ArrayList<Point>`? Not yet.
Sad. Hope they can do this by the next LTS JDK.
a few questions for the pros
> "The defining trait: no identity"
I get that this makes objects behave like primitive types. Maybe thats reason enough. But is it necessary for the performance boost and de-fluffing the objects? Seems like an orthogonal objective
> There’s a catch worth knowing about here, though: flattened data has to be readable and writable atomically (otherwise it risks “tearing” under concurrent access).
Isn't this a race condition and "undefined bahvior"..? Having to limit yourself to atomic sizes seems like a huge limitation, to accomodate what is most likely buggy code. Is all the effort only gunna help lil toy ColorRGB examples?
> The points array is a million pointers. Each pointer leads to a separate Point object lying somewhere on the heap.
Does this happen in actuality? One would assume the allocator tries to put stuff sequentially on the heap? Its not a guarantee as with these Value Types, but I'd think you could get similar-ish perf with prefetching in cache. I dunno whats happening under the hood.. But when writing Clojure apps the JVM always reserves absurd amounts of heapspace on my machine (to my annoyance). Id assume it can find some place to do contiguous allocations..
Which i guess gets me to my last question... where are the benchmarks broski? It all sounds great, but does it actually yield the insane speedups promised?
Great article, well written. But a benchmark would have been a nice "punchline"
> is it necessary for the performance boost and de-fluffing the objects?
Yes. The one part of the JVM GC that can't run concurrently is heap compaction; objects that can be moved by copying and then deleting would be a huge help for that. And it would be awkward to say the object has an identity but can't be wait/notify'd, at which point you need somewhere for the monitor to go.
> Does this happen in actuality? One would assume the allocator tries to put stuff sequentially on the heap?
Yes. Of course it tries, but semantically the pointers are just pointers and the prefetcher can guess but the system still has to chase them.
> Is all the effort only gunna help lil toy ColorRGB examples
Arguably flattening mostly makes sense for these only.
And yeah, you are right that allocations happen on something called a thread local allocation buffer, which is basically just a pointer bump in cost and objects allocated one after the other should be physically close in memory for the most part (though an object's creation may require a bunch of other object's creation that would sit in-between). But these have headers, so not as dense as they could be (though due to GCs being generational, they may end up actually closer in the next gen? The in-between temporary objects wouldn't survive for the most part)
> I get that this makes objects behave like primitive types. Maybe thats reason enough. But is it necessary for the performance boost and de-fluffing the objects? Seems like an orthogonal objective
It feels like an orthogonal objective and honestly arbitrary distinction, yes.
> Isn't this a race condition and "undefined bahvior"..? Having to limit yourself to atomic sizes seems like a huge limitation, to accomodate what is most likely buggy code.
I think they meant it like the appearance of atomic behavior from a java multithreading view.
> Does this happen in actuality?
Yes, it does happen. Having guarantees on this front leads to better performance.
> But when writing Clojure apps the JVM always reserves absurd amounts of heapspace on my machine (to my annoyance)
Might be a configuration problem?
From the article:
> In 1995, a memory access cost roughly the same as a CPU operation
Uhm... no?!
Here's a CS paper from 1993(!) about prefetching from cache(!!) because the cache was slower than the ALU. https://www.eecs.umich.edu/techreports/cse/93/CSE-TR-152-93....
It would perhaps make Java look a little bad to say that, in 1995, the prevailing attitude in certain circles was "If it's too slow, just wait for faster hardware - Moore's Law forever baby!" (Of course, Sun was selling, at the time, relatively fast hardware - the slower the software, the faster the required hardware)
Yes, this also stood out to me. I usually think of CPUs and memory having parity in the early 80s, but I never bothered to check for sure. I do remember some early computer architects writing about memory being faster than the CPU!
Dupe? https://news.ycombinator.com/item?id=48590056