andishgar commented on issue #46177: URL: https://github.com/apache/arrow/issues/46177#issuecomment-2838305341
> Thanks for looking into these [@andishgar](https://github.com/andishgar) ! We could certainly make better use of preallocation. > > Of note: for the `*View` output types, it might even be reasonable to use [`ScalarKernel::can_write_into_slices`](https://github.com/apache/arrow/blob/c2ec1cda88918a3920b7b29eca8d2e3ca7a49fb2/cpp/src/arrow/compute/kernel.h#L564C8-L564C30) which in turn can enable [contiguous preallocation](https://github.com/apache/arrow/blob/0580deeab4731e41e5c993d2f5f068e462118c3a/cpp/src/arrow/compute/exec.cc#L995-L998). StringViews can refer to multiple data buffers, so it'd even be possible to append to different data buffers from different threads My current understanding is limited to ScalarFunction. However, I have two questions: 1-Why is no buffer data allocated for string data types? For compute functions like `ascii_upper`, can the output data buffer length be inferred from the input? 2-Are Scalar compute functions executed using multi-threading? (I’m aware of other compute functions but haven’t studied them yet.) -- 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: github-unsubscr...@arrow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
