Quick js versus goja

[gedw99]

Proposed change

https://gitlab.com/cznic/quickjs

Use case

Perf !

see benchmarks ..

Perhaps another agent in the v3 branch is the best way .

Darwin and windows should be possible too

Contribution

No response

[hpvd]

See also this discussion #369

[MikeHawkesCapventis]

Not sure about this - complexity of build comes in to play. Goja has no C dependencies, and that makes the build significantly easier for us in the Golang world. Also, it's trivial to add new libraries to Goja, so we extend our model to 'wrap' all the network actions - this lets us have business-centric Javascript - we could have a GetWeather("home location") function that calls a NATS micro-service to grab the weather. The Javascript script has no knowledge of anything to do with the lower-level network calls and NATS callout etc. To my mind, trying to do that in QuickJS is horrible by comparison and breaks out of the Go ecosystem.

[jordan-rash]

Great discussion, thanks all. As it sits now, as we move closer to the point where the V3 branch lands on main (very soon), an actor SDK will be released as well. This will hopefully empower everyone to quickly experiment with all these different engines so that we can discover pros/cons together. As it sits now, as you can see from the branch, we are experimenting with goja, but I also hope to look at sobek before we decide which engine we are going to iterate on.

[MikeHawkesCapventis]

I've played with a few of these to build integration engines for our tech and clients - happy to share experiences, if that helps.