How can we join multiple coroutines? Await waits only on a single coroutine

:information_source: Attention Topic was automatically imported from the old Question2Answer platform.
:bust_in_silhouette: Asked By pseudonymous_person

Hi all,
I’m very excited about the new await keyword and coroutines.

In the event where you need to fire off multiple coroutines, say to download 3 files, what’s the recommended pattern to do it simultaneously?

With just await, it has to be done sequentially:

await download_coroutine_1
await download_coroutine_2
await download_coroutine_3

Ideally, it would have the join feature in other languages:

join(download_coroutine_1, download_coroutine_2, download_coroutine_3)


await [download_coroutine_1, download_coroutine_2, download_coroutine_3]

Or even better, hack’s concurrent block: Asynchronous Operations: Concurrent

concurrent {
  first = await download_coroutine_1
  second = await download_coroutine_2
  third = await download_coroutine_3

If this isn’t implemented yet, please share a code pointer or two to get me started and I’ll implement it for us.

:bust_in_silhouette: Reply From: kienhoang

Hi, I made it by creating a small utility called Counter and using lambda functions to make different coroutine blocks.:

class_name Counter

signal completed

var _counter: int
var _total: int

func _init(p_total: int):
    _total = p_total
    _counter = 0

func submit():
    _counter = _counter + 1
    if _counter == _total:

In my main functions:

func main():
    var jobs = # I have 3 jobs here

    var job_1 = func():
        await download_coroutine_1

    var job_2 = func():
        await download_coroutine_2

    var job_3 = func():
        await download_coroutine_3

    await jobs.completed

Thanks for this feedback. I think that could work.

In your example, don’t you still need to invoke the lambdas? And do you do that without awaits so you only block at the bottom?

await jobs.completed

pseudonymous_person | 2023-02-26 14:21

Oh yes, that’s true. Thanks

kienhoang | 2023-02-26 14:46

:bust_in_silhouette: Reply From: jovialthunder

I liked kienhoang’s answer and made a version that fires off all of the jobs in quick succession after you queue them up. Create this object, queue up as many objects as you’d like (with their params, start method, and finished signal), and then run them all at once with run_all(). Await completed on the object for them all to finish.

If you want to do stuff with the objects’ return params as they complete, connect your own functions to their signals before calling run_all.


func http_and_timer(http_request:HTTPRequest, timer:Timer):
	var parallel :=
	parallel.append(http_request, [''], 'request', 'request_completed')
	parallel.append(timer, [1000], 'start', 'timeout')
	await parallel.completed


extends Object
class_name ParallelCoroutines

signal completed

var queued_coroutines:Array[Dictionary]
var total_count: int
var completed_count: int

func append(object:Object, params:Array, call_method:StringName, complete_signal:StringName):
	queued_coroutines.append({'object': object, 'params': params, 'call_method': call_method, 'complete_signal': complete_signal})

func run_all():
	total_count = queued_coroutines.size()
	for routine in queued_coroutines:
		var object:Object = routine['object']
		var params:Array = routine['params']
		var call_method:StringName = routine['call_method']
		var complete_signal:StringName = routine['complete_signal']
		object.callv(call_method, params)
		object.connect(complete_signal, _on_completed)

func _on_completed(a=null, b=null, c=null, d=null, e=null, f=null, g=null): # I don't know how else to allow but ignore parameters
	completed_count = completed_count + 1
	if completed_count == total_count: