Discussion

Project Panama: Interconnecting JVM and native code

grg0: I love it how Java is "innovating" by catching up to things that other programming languages have had for three decades.

latchkey: "We are improving and enriching the connections between the Java virtual machine and well-defined but “foreign” (non-Java) APIs, including many interfaces commonly used by C programmers."Where does it say "innovating"?

grg0: They are doing something new in the language -> innovating.JNI was always the wrong way to do FFI. FFI should require no changes or wrappers in the native code; anything short of that is unnecessary and inefficient. Yet, somehow, in Java land, this is still the norm in 2026?I'd really want to love Java, but man, it has a long laundry list of warts and a near-zero pace of innovation.

marginalia_nu: Well, yeah. It's the explicit design philosphy of the language to wait and see what works as other languages do the experimenting.

re-thc: > Yet, somehow, in Java land, this is still the norm in 2026?FFM (what this article refers to) was released some releases ago. So what is the issue? If you mean what 3rd party libraries use - is that a concern to you? That's like saying there exists legacy code.> it has a long laundry list of wartsIt's such a surprise because you haven't even mentioned 1.> and a near-zero pace of innovationGarbage collection? ZGC?

millipede: What other language does it better?

andy800: Paet of Panama is the Vector API, currently in "incubation". Kotlin Notebooks are a great dataframe alternative to pandas or polars in Python (and dplyr in R), and work fine for relatively small data sets, but are indeed slower when dealing with calculations on large data. Vectors should reduce that gap significantly.

latchkey: This isn't new or innovating. This is "improving and enriching".You're unfairly trying to hold making improvements against them.

corroclaro: Check out the Clojure tech.ml.dataset libraries - they are very very fast