I have implemented a sleep sort algorithm using co...

# coroutines

[igor.wojda]

I have implemented a sleep sort algorithm using coroutines (don’t ask me why🤣) and now Vasya asks me if this implementation is thread-safe? And now I really wonder if it is thread-save to access sortedList like this? https://twitter.com/igorwojda/status/1255416485293502465

[araqnid]

(JVM) the mutableListOf returns a LinkedHashSet, which is not concurrent-modification safe (although it seems unlikely you’d hit that in this example, with the writes guaranteed to be 1sec apart)

[octylFractal]

if you use a single-threaded dispatcher like Dispatchers.Main (if you have it) or Executors.newSingleThreadExecutor().asDispatcher(), you could actually make this thread safe

[Evan R.]

Also, don’t use GlobalScope if you can avoid it! Errors won’t be thrown properly

[octylFractal]

err, rather, not "thread safe" but it wouldn't be cross-thread anymore

[Dominaezzz]

Cool sort though. 😁

[octylFractal]

and yes, the solution to not use GlobalScope here is to do:

coroutineScope {
  for (elm in list) {
    launch(properDispatcher) {
      delay(1000L * elm)
      sortedList.add(elm)
    }
  }
}

[Evan R.]

Instead, use the coroutineScope() builder and either a mutex or a channel to control access to the shared list lol

[Evan R.]

Here’s how you can do it with a channel:

[Dominaezzz]

Should actually be launch .

[octylFractal]

ah, yes

[Dominaezzz]

Should also leave the dispatcher to be inherited from context.

[octylFractal]

well, I was assuming a patch with the new dispatcher, so you'd either have to wrap the coroutineScope or the loop with withContext in that case

[Evan R.]

Untitled

[Evan R.]

The above snippet is thread-safe due to only the receiver coroutine mutating the list 🙂

[Dominaezzz]

Might as well go all the way and do channelFlow { ... }.toList() . 😁

[Evan R.]

That too lol

[Kroppeb]

Ideally, you'd have a way shorter delay, and add a insertion sort final step