jorisvandenbossche commented on PR #40721: URL: https://github.com/apache/arrow/pull/40721#issuecomment-2020811502
Based on a cprofile run and for this small example, around 35% comes from the actual conversion in our C++ layer (creating the (copied) numpy arrays), and the remainder is some overhead in our python code, and then mostly the overhead on the pandas' side constructing the objects. For example, a large part (ca 20-25%) is in the creation of the Index object for the column names. Looking at the profile, there looks to be some easy room for improvement here, but so that's something to do on the pandas side. -- 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]
