HIVE中IN的坑
问题:为什么HIVE中用了 NOT IN,结果集没了?
注:这个是原创,转载请注明,谢谢!
直接进实验室>>
> select * from a;
OK
1 a1
2 a2
3 a3
Time taken: 0.063 seconds, Fetched: 3 row(s)
hive> select * from b;
OK
1 b1
2 b2
NULL b3
Time taken: 0.063 seconds, Fetched: 3 row(s)
# 两表通过id匹配,求 A-B ,用 left join实现
hive> select t1.id,t1.name,t2.name from a t1
> left join b t2 on t1.id = t2.id
> where t2.name is null
OK
3 a3 NULL
Time taken: 34.123 seconds, Fetched: 1 row(s)
# 两表通过id匹配,求 A-B ,用 NOT IN 实现
select * from a where id not in ( select id from b );
OK
Time taken: 34.123 seconds, Fetched: 0 row(s)
这里有诡异了,为什么结果集没了呢? 不能啊??
原因:
在RMDB中, t1.id IN (select t2.id from b t2 ) 等价于 : t1 join b t2 on t1.id = t2.id and t1.id is not null
在hive中,虽然我们的版本已经高达2.0.0,但是对于IN的处理还是就比较简陋,没有对null值进行屏蔽,导致凡是子查询中有null值, 条件就会变成: id in ( null) , 当然, id in ( null) 这个条件是永远不会有结果的。
正确的用法:
# 两表通过id匹配,求 A-B ,用 NOT IN 实现
select * from a where id not in ( select id from b where id is not null );
OK
3 a3 NULL
Time taken: 34.123 seconds, Fetched: 1 row(s)
各位不妨可以做个试验:
--没结果
hive> select * from a where id not in (null);
OK
Time taken: 3.603 seconds