over partition by与group by的区别:
group by是对检索结果的保留行进行单纯分组,一般和聚合函数一起使用例如max、min、sum、avg、count等一块用。partition by虽然也具有分组功能,但同时也具有其他的高级功能。
- CUME_DIST 小于等于当前值的行数/分组内总行数
create table windows_dept
(
deptno string,
pno string,
sum_insured decimal(18,2)
)
ROW FORMAT DELIMITED
FIELDS TERMINATED BY ‘,’
stored as textfile;
数据:
G01,P001,10
G01,P002,20
G01,P003,30
G02,P004,40
G02,P005,50
G03,P006,60
G03,P007,70
G03,P008,80
G03,P009,90
G03,P010,100
测试:
SELECT
deptno,
pno,
sum_insured,
CUME_DIST() OVER(ORDER BY sum_insured) AS rn1,
CUME_DIST() OVER(PARTITION BY deptno ORDERBY
sum_insured) AS rn2
FROM windows_dept;
结果:
deptno pno sum_insured rn1 rn2
G01 P001 10 0.1 0.3333333333333333
G01 P002 20 0.2 0.6666666666666666
G01 P003 30 0.3 1.0
G02 P004 40 0.4 0.5
G02 P005 50 0.5 1.0
G03 P006 60 0.6 0.2
G03 P007 70 0.7 0.4
G03 P008 80 0.8 0.6
G03 P009 90 0.9 0.8
G03 P010 100 1.0 1.0
分析:
rn1: 没有partition,所有数据均为1组,总行数为10,
第1行:小于等于10的行数为1,因此,1/10=0.1
第5行:小于等于50的行数为5,因此,5/10=0.5
rn2: 按照部门分组,deptno=G03的行数为5,
第6行:小于等于60的行数为1,因此,1/5=0.2
第8行:小于等于80的行数为3,因此,3/5=0.6
- PERCENT_RANK 分组内当前行的RANK值-1/分组内总行数-1
SELECT
deptno,
pno,
sum_insured,
PERCENT_RANK() OVER(ORDER BY sum_insured) AS rn1, --分组内
RANK() OVER(ORDER BY sum_insured) AS rn11, --分组内RANK值
SUM(1) OVER(PARTITION BY NULL) AS rn12, --分组内总行数
PERCENT_RANK() OVER(PARTITION BY deptno ORDER BY sum_insured) AS rn2
FROM windows_dept;
结果:
deptno pno sum_insured rn1 rn11 rn12 rn2
G01 P001 10 0.0 1 10 0.0
G01 P002 20 0.1111111111111111 2 10 0.5
G01 P003 30 0.2222222222222222 3 10 1.0
G02 P004 40 0.3333333333333333 4 10 0.0
G02 P005 50 0.4444444444444444 5 10 1.0
G03 P006 60 0.5555555555555556 6 10 0.0
G03 P007 70 0.6666666666666666 7 10 0.25
G03 P008 80 0.7777777777777778 8 10 0.5
G03 P009 90 0.8888888888888888 9 10 0.75
G03 P010 100 1.0 10 10 1.0
分析:
rn1: rn1 =(rn11-1) / (rn12-1)
第一行,(1-1)/(10-1)=0/9=0
第二行,(2-1)/(10-1)=1/9=0.1111111111111111
第四行,(4-1)/(10-1)=3/4=0.3333333333333333
rn2: 按照deptno分组,deptno=G01的总行数为3
第一行,(1-1)/(3-1)=0
第二行,(2-1)/(3-1)=0.5
第三行,(3-1)/(3-1)=1