--! qt:dataset:src
set hive.mapred.mode=nonstrict;
set hive.explain.user=false;
-- SORT_BEFORE_DIFF
-- union case: both subqueries are map-reduce jobs on same input, followed by reduce sink

explain 
  select count(1) FROM (select s1.key as key, s1.value as value from src s1 UNION  ALL  
                        select s2.key as key, s2.value as value from src s2) unionsrc;

select count(1) FROM (select s1.key as key, s1.value as value from src s1 UNION  ALL  
                      select s2.key as key, s2.value as value from src s2) unionsrc;