Runtime: Why PipeWriter doesn't have synchronous Flush?

[aliveli186]

System.IO.Pipelines

[carlossanlop]

Hello @aliveli186 , thank you for your contribution.

This was probably an oversight when the classes were added in this PR to corefx: https://github.com/dotnet/corefx/pull/27007 . I did not find any comments asking about adding a Flush() method.

Would you mind sharing your use case? Why would you prefer to use Flush instead of FlushAsync?

cc @JeremyKuhne

[stephentoub]

This was probably an oversight

I don't think so :wink:
cc: @davidfowl

[carlossanlop]

@stephentoub @davidfowl so what would be the reason behind not including Flush in the PR I mentioned above? I can't find an explanation in that PR, and there's no mention of the API approval issue for that PR, where I was hoping this would be explained. If it wasn't an oversight, then it would be great to provide an answer in this thread so we can close the issue.

[stephentoub]

System.IO.Pipelines is intended for high-throughput and scalable processing of data. In a producer/consumer scenario, a writer that flushes may need to wait for the reader to read before the flush can complete, and with a synchronous flush, that would mean blocking a thread (probably a thread pool thread) waiting for some other thread (probably also a thread pool thread) to do the necessary work. That's a recipe for being non-scalable.

[carlossanlop]

Clear as water, thanks @stephentoub . I'll close the issue.