alamb commented on issue #3463: URL: https://github.com/apache/datafusion/issues/3463#issuecomment-3712034479
> Oh right, yes it will do that sorry, been years since I wrote that code (and it looks like there's some new PushDecoder anyway that might change all of this). FWIW the push decoder doesn't change the fundamental IO -- it is instead an explicit state machine (rather than using the implicit one created by the rust compiler with `await` calls). It is algorithmically the same thing > So yes it will behave the way you describe, which could make a difference for stores with high first-byte latencies. However, my understanding is Andrew is running on an NVMe drive where time spent doing IO will be bounded by bytes-read not number of fetches, which filter pushdown would if anything reduce... The idea that one reason it slows down is due to additional latency of multiple distinct IOs is an interesting one. I will ponder that and see if I can find some way to reduce it / prefetch -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
