alamb commented on issue #10448: URL: https://github.com/apache/arrow-rs/issues/10448#issuecomment-5114019947
> However, the PageStore patch introducing a "zero-copying", which making the Page buffer staging in a Vec<Bytes>, and making the buffer staging with zstd's bound ( would is generally as large as original buffer ), occuping a large ratio before flushing the RowGroup Does that mean that https://github.com/apache/arrow-rs/pull/10020 made writing faster but use more memory? > I would dive into it, and so should v1 apply the same rule? Which gives control on both v1/v2 shrink_to_fit. And I don't know should I regard it as column-level feature ( In my scenerio, I've a html column, it's huge and zstd can have high compression ratio, and other columns is just lightweight, introducing some memory overhead is ok) Another thing we could consider would be to compare the len to the capacity (allocation size) and call shrink to fit if there was a large difference 🤔 -- 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]
