Is it ok to use ConcurrentHashMap with coroutines?...

# coroutines

[Anton Kapelyushok]

Is it ok to use ConcurrentHashMap with coroutines?
What dispatcher should I generally use with it?

[Anton Kapelyushok]

my use case:

I have potentially long reads from map - iterate over all entries doing something with each. During the reads there could be potential write requests, which I don't want to block

ConcurrentHashMap guarantees on concurrent reads are ok in my case. The only question is, as ConcurrentHashMap is lock based and blocking, should I use Dispatchers.IO with it?

[Patrick Steiger]

IMO default dispatcher is fine — only use IO for actual IO

Or use a normal map with suspending mutex lock

Not sure which would perform better

[Patrick Steiger]

On another topic, are you the reporter of https://github.com/Kotlin/kotlinx.coroutines/issues/3757?

Did you find an explanation for it?

[Anton Kapelyushok]

Yes, I am the reporter

[Anton Kapelyushok]

When calling cancel() on parent
1) ProducerCoroutine::onCancelled is called
2) ChannelCoroutine::cancel is called

[Anton Kapelyushok]

ChannelCoroutine::cancel is supposed to call channel.cancel(), but it does not happen because there isCanceled check

[Anton Kapelyushok]

And ProducerCoroutine::onCancelled calls channel.close(cause) instead of channel.cancel()

[Anton Kapelyushok]

In short I don't know what is happening, and I am not convinced it is the expected behaviour 🙂