paleolimbot commented on issue #38325: URL: https://github.com/apache/arrow/issues/38325#issuecomment-2020340273
> if a user actually wants the data to be copied, pandas cannot do this themselves, and it would be cudf that need to perform the copy. That is a great point! The producer almost certainly knows how to perform a copy in the most efficient way, but the consumer, particularly a CPU consumer, probably does not want to handle those details. More broadly, for all of the `__arrow_c_xxx__` protocol methods, the caller does not have any way to know how expensive calling the method will be (e.g., it may perform a copy or allocate quite a lot if transforming from item/rowwise to Arrow). Making all copies for all methods explicit seems safer but, as Joris mentioned, severely limits the utility of the protocol. -- 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]
