Shameless plug: CheerpJ is our solution to run any JVM language in the browser, including Clojure. Reflections, Multithreading and Swing / AWT apps are all supported.
I used it to make a program that logs all activity happening on the Pioneer CDJs. The best reverse engineering of the Pioneer protocols is a Java project, but I wanted to write the rest of my application in Go.
GraalVM plus a GitHub action spits out native binaries that I can exec and interact with over stdio from Go.
If/when the WASM backend supports UDP networking and threads I'd love to run it as WASM instead of a binary.
Clojure code ends up using a lot of reflection if you're doing generic Java interop. Most code destined for the GraalVM will add `(set! warn-on-reflection true)` and get repl warnings and you can set type hints accordingly.
My understanding is GraalVM's JIT keeps some optimizations for the enterprise edition, and is otherwise comparable to Hotspot, with some niche exceptions. I'm not sure "many" would benefit without paying.
I vaguely remember using it about 10 years ago for work, can't remember what for, or anything about that situation, but the one takeaway that I do remember is that it was neat and innovative, but ultimately not good enough to overthrow whatever we were using instead.
Just like most FOSS licenses, that in big corps always got through legal.
In the projects where that isn't required, usually we have licence validation tooling on the CI/CD pipeline, that breaks the build if the legal wishes aren't fulfilled.
Not really, but one thing that bothers me is how unreproducible GraalVM is. AFAIK every distro that has binaries for it just repacks the binaries released from Oracle, and the last time I searched I couldn't find instructions on how to build from scratch (I was the maintainer of GraalVM in nixpkgs, not anymore because I just got fed-up with it).
yes sir, i kinda love seeing all the different ways folks mash stuff together to get dev setups working, so much trial and error that never gets seen - you ever feel like the real blocker is always some weird little detail nobody warns you about?
> The output WASM of this simple program is 5.6MB binary, which can be pruned a bit via wasm-opt tool, just make sure that it doesn't break anything for you. Luckily when compressed (gzip, brotli, etc) the binary becomes just ~2.5MB in size.
That’s much better than I expected! Very impressive work here. It actually looks viable for certain applications.
For quite some time (maybe even still today?), ClojureScript was compiled to JS using the Google Closure Compiler. I felt sympathy for anyone who had to discuss that out loud.
Shameless plug: CheerpJ is our solution to run any JVM language in the browser, including Clojure. Reflections, Multithreading and Swing / AWT apps are all supported.
https://labs.leaningtech.com/blog/cheerpj-4.0
And yes, it can run Minecraft :-)
https://browsercraft.cheerpj.com/
The more the merrier.
GraalVM is neat.
I used it to make a program that logs all activity happening on the Pioneer CDJs. The best reverse engineering of the Pioneer protocols is a Java project, but I wanted to write the rest of my application in Go.
GraalVM plus a GitHub action spits out native binaries that I can exec and interact with over stdio from Go.
If/when the WASM backend supports UDP networking and threads I'd love to run it as WASM instead of a binary.
- https://github.com/nzoschke/vizlink
- https://github.com/nzoschke/vizlink/blob/main/.github/workfl...
> Starting from v25 GraalVM added support for WASM
GraalVM is so amazing technically, but gets so little love by HN.
Just to clarify -- GraalVM had support for RUNNING WASM for quite a long time.
This post references the ability to compile programs via native images to WASM as an output format.
When I tried it it was great but also not easy to use. Things become hard quickly, e.g. If your jvm code uses something like reflection.
Clojure code ends up using a lot of reflection if you're doing generic Java interop. Most code destined for the GraalVM will add `(set! warn-on-reflection true)` and get repl warnings and you can set type hints accordingly.
Native images are just one feature though. Many people would benefit from the Graal JIT even if they don't care about anything else.
My understanding is GraalVM's JIT keeps some optimizations for the enterprise edition, and is otherwise comparable to Hotspot, with some niche exceptions. I'm not sure "many" would benefit without paying.
It's been entirely free since mid-2023. Even if you care rather about native images, you probably want to use the proprietary distribution anyway.
I vaguely remember using it about 10 years ago for work, can't remember what for, or anything about that situation, but the one takeaway that I do remember is that it was neat and innovative, but ultimately not good enough to overthrow whatever we were using instead.
First release happened in 2019
My memory is not great, as implied by that very comment.
GraalVM has evolved from another Sun Research Labs project, MaximeVM, but that was never released as a product.
Does it have to do with the somewhat complicated licensing?
Current licenses for Oracle JDK and GraalVM are essentially the same terms. It's pretty straightforward.
https://www.oracle.com/downloads/licenses/graal-free-license...
So for shops that are already open to using the JDK they're obviously already used to the legalese/implications for companies built on this software.
Everyone else in the world probably does not see this as "straight forward".
So Step 0, be a lawyer.
Just like most FOSS licenses, that in big corps always got through legal.
In the projects where that isn't required, usually we have licence validation tooling on the CI/CD pipeline, that breaks the build if the legal wishes aren't fulfilled.
Oracle has specifically reworked its license to make terms clear after the bad press around the initial release of Oracle JDK 17:
https://blogs.oracle.com/java/post/free-java-license
Any company using Java should be willing to read and understand Oracle's terms, whether they use third party OpenJDK distributions or Oracle's builds.
If you're leaving significant performance gains on the table because you can't read, that's on you.
And Oracle's army of lawyers.
Not really, but one thing that bothers me is how unreproducible GraalVM is. AFAIK every distro that has binaries for it just repacks the binaries released from Oracle, and the last time I searched I couldn't find instructions on how to build from scratch (I was the maintainer of GraalVM in nixpkgs, not anymore because I just got fed-up with it).
That is how people lose out on great technology, while worshiping the ways of 1980's server room computing.
Nitpick maybe: It should be Wasm not WASM.
https://youtu.be/nnDo0i6NbsI?si=XSqgSVoV6ISBWg2n&t=185
https://webassembly.org/
yes sir, i kinda love seeing all the different ways folks mash stuff together to get dev setups working, so much trial and error that never gets seen - you ever feel like the real blocker is always some weird little detail nobody warns you about?
The analysis of the benchmark is wrong. Native is faster than JVM for 2 out of 4 operations. The 2-3x vs 5-12x are hence not correct.
> The output WASM of this simple program is 5.6MB binary, which can be pruned a bit via wasm-opt tool, just make sure that it doesn't break anything for you. Luckily when compressed (gzip, brotli, etc) the binary becomes just ~2.5MB in size.
That’s much better than I expected! Very impressive work here. It actually looks viable for certain applications.
It's acceptable. Meanwhile 5.6MB of Wasm is about the size of the entire test suite of Scala.js. :-p
GraalVM is excellent technology, but when it comes to targeting Wasm, I believe the core language compilers will always have an edge.
Nice, I will revisit closure to try this out
Clojure. ;)
For quite some time (maybe even still today?), ClojureScript was compiled to JS using the Google Closure Compiler. I felt sympathy for anyone who had to discuss that out loud.
The trick is to call the latter GCC. Much less confusing. ;-)
And now I've got coffee in my sinus. Thanks for that :D
Would have been good to create a pamphlet to inform on the differences. A Clojure/Closure Brochure, if you will
Imagine the pain of talking about Clozure Common Lisp.
You have no idea how many times I typed clojure when I meant to type closure throughout my career. Bizarrely backwards.