@andrewvc @ph @jordansissel FYI not sure if I should be happy or sad here after all the work that went into Java exec ... but this looks pretty safe to merge to give 6.1 a massive boost imo ... obviously way diminishing people's incentive to try out the Java exec ...

Also, this version beats the Java exec by quite a margin in baseline (17% faster) stdin->stdout or just a single fast filter like the UA (10% faster) one in place :(

Woa, good catch! This change is low risk we could release it in 6.0.x and event 5.6.x if we ever do another release. I am doing some testing, will report soon.

oh sweet <3 🥇
also doing testing.

obviously way diminishing people's incentive to try out the Java exec

Concerning the java execution, we will make it faster eventually, that should not block us from merging it with a feature flag.

@ph "should" or "should not"? :D

should not :)

I did some testing with a really simple pipeline configuration

filter {
    mutate { uppercase => [ "message" ] }
}

I got a 7% improvements with the code changes, but as soon as you add more plugins to the filter stage, I get the 30% speed improvements.

filter {
    mutate { uppercase => [ "message" ] }
    mutate { lowercase => [ "message" ] }
    mutate { uppercase => [ "message" ] }
    mutate { lowercase => [ "message" ] }
}

@ph alrighty thanks, merging :)

Armin Braun merged this into the following branches!

needs manual backport to 6.0 and 5.6, will do later

Thank you for working on this :)

np :)