Investigate JEP 441: Pattern Matching for switch

[tgodzik]

JEP 441 introduced spattern matching for switch statements, which we could possibly use to optimize the generated bytecode. @sjrd mentioned SwitchBootstraps that could be especially useful

Introduced in JDK 21

[He-Pin]

Yes, so it's been discussed on the contributions.scala-lang.org too, I really want the next Scala LTS a Java 21 based.
https://bugs.openjdk.org/browse/JDK-8304394

[SethTisue]

note that even if we don't adopt JDK 21 as a new baseline (11 or 17 seems much likelier), this could still be implemented to be enabled only when the bytecode version is 21+

[He-Pin]

@tgodzik @SethTisue , it's been introduced since JDK 17, and marked as final in JDK 25:

It's marked as final in JDK25; it would be nice to have this, and then, users can avoid something like this:

databricks/sjsonnet#313

[He-Pin]

Even though we define it in a Scala 3 enum, it's still a long chain of instanceOfs:

[sjrd]

Please measure. Don't look at the bytecode. The JVM does all sorts of things to your bytecode to optimize it. It doesn't interpret it as is. Making optimization decisions based on what the bytecode looks like is basically gambling.

[He-Pin]

Yes, but this kind of optimization is landed in zio/ce/akka projects already, by hand.