Github user squito commented on the issue: https://github.com/apache/spark/pull/23058 @attilapiros yes, something like that would be possible. I was thinking you'd just use the existing serializer methods to do it, soemthing like: ```scala val buffer = getRemoteManagedBuffer() val valueItr = deserializeStream(buffer.createInputStream()) val result = valueItr.next() assert(!valueItr.hasNext()) // makes sure its closed too ``` my reluctance to bother with it is that you'd still be getting a `DirectTaskResult`, which has the data sitting in a `ByteBuffer` anyway. Also, its not that big a deal, as this is only for a single task result, which is not large in general. The change here is to avoid reading an entire partition into memory.
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org