[ https://issues.apache.org/jira/browse/HIVE-20278?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16570713#comment-16570713 ]
Ashutosh Chauhan commented on HIVE-20278: ----------------------------------------- +1 What will it take to have RecordReaders other than Scan to return rows in order. No reason to overhead in that case either. Can you please create a follow-up for that. > Druid Scan Query avoid copying from List -> Map -> List > ------------------------------------------------------- > > Key: HIVE-20278 > URL: https://issues.apache.org/jira/browse/HIVE-20278 > Project: Hive > Issue Type: Improvement > Reporter: Nishant Bangarwa > Assignee: Nishant Bangarwa > Priority: Major > Labels: PERFORMANCE > Attachments: HIVE-20278.patch > > > DruidScanQueryRecordReader gets a compacted List<Object> from druid. It then > converts that list into a Map<String,Object> as DruidWritable where key is > the column name. > At the second stage DruidSerde takes this DruidWritable and creates a List > out out of the map again. We can avoid the map creation part by reading the > list sent by druid directly in the DruidSerde.deserialize() method. -- This message was sent by Atlassian JIRA (v7.6.3#76005)