Hive进阶 分析窗口函数

需求：找出雇员（emp）表，各个部门工资前三个员工信息

来源：Oracle，db2,Sql server等数据库，这类型数据库被金融公司使用，做一些报表。
hive相关支出：
https://cwiki.apache.org/confluence/display/Hive/LanguageManual+WindowingAndAnalytics

select ename, deptno,sal,
       ROW_NUMBER() OVER(PARTITION BY deptno ORDER BY sal desc) row_num
from tb_emp

结果

ename	deptno	sal	row_num
KING		10	5000.0		1
CLARK	10	2450.0		2
MILLER	10	1300.0		3
SCOTT	20	3000.0		1
FORD		20	3000.0		2
JONES	20	2975.0		3
ADAMS	20	1100.0		4
SMITH		20	800.0		5
BLAKE	30	2850.0		1
ALLEN		30	1600.0		2
TURNER	30	1500.0		3
MARTIN	30	1250.0		4
WARD		30	1250.0		5
JAMES	30	950.0		6

问题：20号部门两人薪资一样
使用rank

select ename, deptno,sal,
       RANK() OVER(PARTITION BY deptno ORDER BY sal desc) row_num
from tb_emp

结果:

ename	deptno	sal	row_num
KING		10	5000.0		1
CLARK	10	2450.0		2
MILLER	10	1300.0		3
SCOTT	20	3000.0		1
FORD		20	3000.0		1
JONES	20	2975.0		3
ADAMS	20	1100.0		4
SMITH		20	800.0		5
BLAKE	30	2850.0		1
ALLEN		30	1600.0		2
TURNER	30	1500.0		3
MARTIN	30	1250.0		4
WARD		30	1250.0		4
JAMES	30	950.0		6

问题：直接从1跳到了3
解决：使用DENSE_RANK

select ename, deptno,sal,
       DENSE_RANK() OVER(PARTITION BY deptno ORDER BY sal desc) row_num
from tb_emp

结果:

ename	deptno	sal	row_num
KING		10	5000.0		1
CLARK	10	2450.0		2
MILLER	10	1300.0		3
SCOTT	20	3000.0		1
FORD		20	3000.0		1
JONES	20	2975.0		2
ADAMS	20	1100.0		3
SMITH		20	800.0		4
BLAKE	30	2850.0		1
ALLEN		30	1600.0		2
TURNER	30	1500.0		3
MARTIN	30	1250.0		4
WARD		30	1250.0		4
JAMES	30	950.0		5