Hi there, how am i supposed to launch coroutines in a request handler? I want to instantly respond t...

# ktor

[Tobias Marschall]

Hi there, how am i supposed to launch coroutines in a request handler?
I want to instantly respond to the client and handle the request in the background using a coroutine. I tried to do it this way and it also works, but when i extract the request handler into an extra function, then intellij gives me a justified warning.
No warning:

get("/test") {
    launch {
        // do something
    }

    call.respond("Success")
}

Warning Ambiguous coroutineContext due to CoroutineScope receiver of suspend function when using:

suspend fun PipelineContext<Unit, ApplicationCall>.handleTest() {
    launch {
        // do something
    }

    call.respond("Success")
}

get("/test") {
    handleTest()
}

How am i supposed to do this?

[hfhbd]

PipelineContext already extend CoroutineScope. You should either use suspend fun handleTest(): String { return "Success"} or

fun PipelineContext.handleTest() {
call.respond("Success")
}

[Tobias Marschall]

I am afraid that the PipelineContext will get cancelled and thus the coroutine that i launched using it will stop and not complete the task. Can this be the case?

[Tobias Marschall]

And i need to be the function to suspend as i am calling other suspending functions in it.

[Tobias Marschall]

Created a SO question: https://stackoverflow.com/questions/69696479/launching-coroutine-in-request-handler

[darkmoon_uk]

This may be a case where it's appropriate to use GlobalScope; especially if you launch a parent supervisor Job inside GlobalScope first that groups together all such background tasks?

[Tobias Marschall]

What do you mean by parent supervisor job?

[Joost Klitsie]

What I did to handle a request and then send a socket update to other clients without blocking the response, is to launch the work in another scope. If you just use launch, you use the pipelines' scope. You should just define another scope (CoroutineScope(Dispatchers.Default) for example) and this worked like a charm!

[Tobias Marschall]

Is there something bad about using the pipeline’s scope actually?

[Joost Klitsie]

I am not sure, but I can imagine the pipeline scope to be scoped to the actual request/response. Triggering other (long running) work that falls outside of the request/response scope feels better to handle in a specific scope for that task

[Tobias Marschall]

Do you have an example that you can share with me?

[Joost Klitsie]

I'm not sure when the scope gets cancelled, I can imagine it to be terminated after the response been made successfully but I guess ktor professionals might know the answer to that

[Tobias Marschall]

It seems to keep existing even after the response finished successfully. At least i tried it out and id did…

[Aleksei Tirman [JB]]

The Ktor doesn't cancel jobs for routes handlers so your background job won't be canceled too. Also, you can make your job independent by creating a new one for the context of `launch`:

launch(coroutineContext + Job()) {
    repeat(10000) {
        println("Executing background task $it.")
        delay(1000)
    }
}

[Tobias Marschall]

Is it written somewhere in the documentation that ktor doesn’t cancel the jobs?

[Aleksei Tirman [JB]]

I don't think so.

[Tobias Marschall]

okay, thanks. What exactly is the benefit of using + Job() . As the jobs will not get cancelled anyways it should not be required, shouldn’t it?

[hfhbd]

But you don't want to cancel Ktor`s job if your job fails :)

[Tobias Marschall]

True story, learned something again. Thanks!