AngersZhuuuu commented on issue #24124: [SPARK-27186][SQL] mapPartition to replace map to speedUp Dataset's toLocalIterator process URL: https://github.com/apache/spark/pull/24124#issuecomment-474649973 > Please fix the title. > But, why does this improve speed? you're just doing the decoding on the executor vs driver side, without more parallelism. I'm not sure we can/should move the UnsafeRow from executor to driver; may be OK but I'd be wary of changing the semantics here. In old code, it will get the second value for each kv pair parallelism and the transform to localIterator then decode record in driver one by one, what I want to do is to let it decode in dirver paralleism and just cycle one time
---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
