set hive.map.aggr = true;
-- SORT_BEFORE_DIFF
-- union case: 1 subquery is a map-reduce job, different inputs for sub-queries, followed by reducesink

explain 
  select unionsrc.key, count(1) FROM (select 'tst1' as key, cast(count(1) as string) as value from src s1
                                        UNION  ALL  
                                            select s2.key as key, s2.value as value from src1 s2
                                        UNION  ALL  
                                            select s3.key as key, s3.value as value from src1 s3) unionsrc group by unionsrc.key;

  select unionsrc.key, count(1) FROM (select 'tst1' as key, cast(count(1) as string) as value from src s1
                                        UNION  ALL  
                                            select s2.key as key, s2.value as value from src1 s2
                                        UNION  ALL  
                                            select s3.key as key, s3.value as value from src1 s3) unionsrc group by unionsrc.key;