tustvold commented on issue #8156: URL: https://github.com/apache/arrow-datafusion/issues/8156#issuecomment-1810436859
> FYI, my first exploration shows that this is not easy to do unless we perform row.clone(). Is that acceptable from a perfomance standpoint? FWIW the challenge here isn't the memory copy, that will be extremely cheap, however, what will hurt performance here are the additional allocations per row. RowConverter is currently setup to perform one allocation for an entire batch of rows with individual `Row` just borrowed slices of this one allocation. This amortises allocation overheads over potentially thousands of rows. That being said it is likely possible to get a trait based approach to work, it will just require some non-trivial lifetime-fu that may somewhat undermine any potential benefits from DRYing up the code. -- 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]
