mirror of
https://github.com/apache/impala.git
synced 2026-01-02 21:00:35 -05:00
This fixes two issues: * AddRowCustom() caused a performance regression when the function was heap-allocated. This is solved by splitting the API into two separate calls. This imposes an additional burden on the caller but it is easier to reason about its performance. * Allow re-reading streams with 'delete_on_read_' set so long as no rows were read from the stream. This is necessary for some spilling ExecNodes that prepare the stream for reading in order to acquire the buffer, but then need to spill the stream to free memory before they actually are able to read the stream. Change-Id: Ibab0d774f66be632f17376a56abf302821cca047 Reviewed-on: http://gerrit.cloudera.org:8080/7358 Reviewed-by: Tim Armstrong <tarmstrong@cloudera.com> Tested-by: Impala Public Jenkins