gaogaotiantian commented on PR #53666: URL: https://github.com/apache/spark/pull/53666#issuecomment-3712290867
I have some detail suggestions on the code itself but I want to wait until we have a further discussion about this feature (just to avoid wasting our time). The current implementation (or the concept as its own) will introduce a non-trivial overhead for data conversion. For simple and common types like integer, bytes, strings, we introduced multiple python-level function calls to each element, which could result in a perf regression - to default behavior. The implementation is not complete either I think? We need to handle this in all container types I assume? I'd suggest that we discuss whether we want to implement this - if the benefit trumps the overhead in both perf and maintenance, before we start reviewing the code itself. -- 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] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
