set hive.map.aggr = true;
-- SORT_BEFORE_DIFF
-- union case: all subqueries are a map-reduce jobs, 3 way union, same input for all sub-queries, followed by reducesink

explain 
  select unionsrc.key, count(1) FROM (select 'tst1' as key, count(1) as value from src s1
                                        UNION  ALL  
                                            select 'tst2' as key, count(1) as value from src s2
                                        UNION ALL
                                            select 'tst3' as key, count(1) as value from src s3) unionsrc group by unionsrc.key;


  select unionsrc.key, count(1) FROM (select 'tst1' as key, count(1) as value from src s1
                                        UNION  ALL  
                                            select 'tst2' as key, count(1) as value from src s2
                                        UNION ALL
                                            select 'tst3' as key, count(1) as value from src s3) unionsrc group by unionsrc.key;