Why there is prefetching in ReadUntilCancelledAsync? about open.channelextensions HOT 10 CLOSED

I picked up this technique from chatting Ben Adams on a project I published. It's an optimization that he had presented to me. It's not something you would consider always doing, but in this case the "pre-fetch" seemed a logical step as you are more likely to saturate threads this way.

If you don't think this is appropriate, I could introduce a parameter to allow for disabling it.
At the moment, I can't think of any reason you wouldn't want the optimization.

from open.channelextensions.

ends up in blocking at the end of channel

Not sure what you're suggesting here.
If you have a bounded channel, and it reaches it's configured limit of (lets say) 100. Then you really have 101.
In other cases, where you are either unbound or the channel is not full, you're simply preemptively acquiring an item.

from open.channelextensions.

All that said, I can see where this micro-optimization may not apply to Channels the same way it might in other situations.
It's meant for cases where you expect some amount of potential async delay and .TryRead there really is none.

from open.channelextensions.

I think you'll notice that I do this for writing to a channel as well.

from open.channelextensions.

I'm quite willing to change this. Would you look at the write version of this and tell me what you think there?

from open.channelextensions.

This has been corrected as per your suggestion.
https://www.nuget.org/packages/Open.ChannelExtensions/6.0.1

https://github.com/Open-NET-Libraries/Open.ChannelExtensions/blob/master/Open.ChannelExtensions/Extensions.Read.cs#L100-L105

from open.channelextensions.

@azhmur Thank you for pointing this out. :)

from open.channelextensions.

Thank you for fast and reasonable response. Great library, I wish this would be part of System.IO.Pipelines.

from open.channelextensions.

I wish this would be part of System.IO.Pipelines.

@azhmur Yeah, there are a large number of extensions I wish were simply built into the ecosystem.
Check out my database extensions library.
https://github.com/Open-NET-Libraries/Open.Database.Extensions (there's even a package for channels)
In either case, a large number of basic operations could be streamlined. I couldn't just sit by and not have best practices built in.

from open.channelextensions.

Thank you for fast and reasonable response.

@azhmur I love that someone took the time to scrutinize the code and post an issue. As I mentioned, I do think there is some amount of reasonable optimization for that pattern under specific cases, but this was not the case. I hadn't thought about it so deeply before. Thanks again.

from open.channelextensions.