On Android we have the `<http://Dispatchers.IO|Dispatchers.IO>` available, but that's not the case i...

# multiplatform

[Christian Würthenr]

On Android we have the <http://Dispatchers.IO|Dispatchers.IO> available, but that's not the case in KMM. What's the best way to replace it? I already have a expected actual with a Dispatcher.SharedIO which just points to <http://Dispatchers.IO|Dispatchers.IO> on Android, but what's the best solution on iOS?

[shahroz]

I am currently following this approach

https://github.com/customerio/mobile-shared/commit/838f4e6223fbcf7e422561d60313245c847146de

[Seb Jachec]

@shahroz Looking through Kotlin’s coroutines source code, isn’t your iOS dispatcher equivalent to Kotlin’s Dispatchers.Main implementation anyway? https://github.com/Kotlin/kotlinx.coroutines/blob/master/kotlinx-coroutines-core/nativeDarwin/src/Dispatchers.kt#L32-L43 (and https://github.com/Kotlin/kotlinx.coroutines/blob/master/kotlinx-coroutines-core/native/src/Dispatchers.kt)

[Christian Würthenr]

Dispatchers.Main seems to use the main dispatch queue, that would block the main thread I assume? (not an iOS dev)

[Christian Würthenr]

But I think the DarwinGlobalQueueDispatcher which is mapped to Dispatchers.Default does pretty much the same

[Seb Jachec]

@Christian Würthenr Yes Dispatchers.Main would block the main thread afaik

[Christian Würthenr]

I also assumed that I can replace <http://Dispatchers.IO|Dispatchers.IO> with Dispatchers.Default in KMM, but that lead to the app being dead locked on devices with low CPU core counts as on Android Dispatchers.Default is limited to one thread per core. Looking at the iOS implementation now, that seems not to be the case on iOS

[Christian Würthenr]

So I guess my approach is fine...

expect Dispatchers.SharedIO: CoroutineDispatcher

iOS is then

actual Dispatchers.SharedIO = Dispatchers.Default

and Android

actual Dispatchers.SharedIO = <http://Dispatchers.IO|Dispatchers.IO>

[shahroz]

There is a similar issue, for me Dispatchers.Default limited thread count seemed to be a problem.

But maybe, newFixedThreadPoolContext can fix this

[Christian Würthenr]

But it should not be fixed, right? Because the IO dispatcher is ment to be blocked for any IO. It should be allowed to create new threads as needed as the others will be "busy" waiting for IO

[shahroz]

Yup it won't fix it, but this is the best alternate we have for now

[Christian Würthenr]

Ok, great! Thanks for the input!!

[Christian Würthenr]

🙂