在oracle中代码如下  select t2.home_county, count(distinct t1.user_id)
  from user_t2 t2
 inner join vpn_t1 t1 on t1.sum_month = 201212
                               and t1.inure_date <=
                                   to_date(20121231, 'yyyymmdd')
                               and t1.expire_date >=
                                   to_date(20121231, 'yyyymm')
                               and t1.user_id = t2.user_id
 inner join team_t3 t3 on t3.change_inure_date <=
                                       to_date(20121231, 'yyyymmdd')
                                   and t3.change_expire_date >
                                       to_date(20121231, 'yyyymmdd')
                                   and t3.inure_time <=
                                       to_date(20121231, 'yyyymmdd') + 1
                                   and t3.expire_time >=
                                       to_date(20121231, 'yyyymmdd')
                                   and t3.team_sub_type = 31
                                   and t3.team_id = t1.vpn_group_id
 where t2.sum_month = 201212
   and t2.user_status < 30
   and join_date <= to_date(20121231, 'yyyymmdd')
 group by t2.home_county;
 查看执行计划,当与team_t3表连接的时候是merge join cartestian
然后把上面代码中user_t2 建成临时表t2,vpn_t1 建成t1临时表, team_t3建成t3临时表,在查看执行计划就变成了hash join 了。建成临时表后的代码如下所示
select t2.home_county, count(distinct t1.user_id)
  from t2 
 inner join t1  on t1.user_id = t2.user_id
 inner join t3  on t3.team_id = t1.vpn_group_id
 where 1 = 1
 group by t2.home_county;
请问这是什么原理?