hive 编程指南

复习一下。

P107 函数列表

group by 生成分组之后，可以直接加 having 再进行过滤

order by 全局排序，sort by 仅在各个reducer里分别排序

distribute by s.symbol 相同 s.symbol 值的会落在相同的reducer里面

distribute by ...  sort by ...  可简写为 cluster by ...

Index 创建删除

嵌套查询

select t1.dkey,t1.nkey,t1.ipkey,t1.node_id,t1.time,t1.status,count(*)
from (
select dkey,nkey,ipkey,node_id,
substr(ts,1,13) as time,
CASE
  WHEN somev=-1 THEN 'bad'
  WHEN somev>3000 THEN 'bad'
  WHEN somev>500 THEN 'normal'
  ELSE 'good'
END as status
from
somedb.sometb
where date like '201506%' and dkey in ('keya','keyb')
) t1
group by t1.dkey,t1.nkey,t1.ipkey,t1.node_id,t1.time,t1.status