hive同一张表union all的优化

insert overwrite table dwd.dwd_user_play_event_1d_delta_daily partitions (dt = '${DT}') 
select *
  from ods.user_event_log_ex 
 where dt = '${DT}' and event_name = 'play_start'
 union all 
select *
  from ods.user_event_log_ex 
 where dt = '${DT}' and event_name = 'play_end' and play_time > 0
 union all
select *
  from ods.user_event_log_ex
 where dt = '${DT}' and event_name = 'play_error' and error_code is not null

从上SQL中可以看出都是从同一张表中拉取数据，只是条件不一样，但上面的SQL需要对同一张表扫描三次，每次union all都会产生mapreduce，都是需要消耗资源和时间的，可对以上做如下优化，减少对表的扫描次数和mapreduce的产生。

from ods.user_event_log_ex 
inser overwrite table dwd.dwd_user_play_event_1d_delta_daily partition (dt = '${DT}')
select *
  where dt ='${DT}' and event_name = 'play_start'
insert overwrite table dwd.dwd_user_play_event_1d_delta_daily partition (dt = '${DT}')
select *
  where dt ='${DT}' and event_name = 'play_end' and play_time > 0
insert overwrite table dwd.dwd_user_play_event_1d_delta_daily partition (dt = '${DT}')
select *
  where dt ='${DT}' and event_name = 'play_error' and error_code is not null