开窗函数over(partition by .. [order by ..])用法

---

作者：tang_xuming
来源：CSDN
原文：https://blog.csdn.net/tang_xuming/article/details/79197163
版权声明：本文为博主原创文章，转载请附上博文链接！

1、/** 创建表 /
create table test_orer_partition_by
(
prov_name varchar(20),
city_name varchar(50),
val_cnt int
) ;
2、/* 插入测试数据 /
insert into test_orer_partition_by values (‘GS’, ‘LZ’, 234) ;
insert into test_orer_partition_by values (‘GS’, ‘BY’, 34) ;
insert into test_orer_partition_by values (‘GS’, ‘DX’, 4) ;
insert into test_orer_partition_by values (‘GS’, ‘LN’, 67) ;
insert into test_orer_partition_by values (‘GS’, ‘GN’, 67) ;
insert into test_orer_partition_by values (‘GS’, ‘QY’, 67) ;
insert into test_orer_partition_by values (‘SX’, ‘XA’, 34) ;
insert into test_orer_partition_by values (‘SX’, ‘YA’, 85) ;
insert into test_orer_partition_by values (‘SX’, ‘BJ’, 254) ;
insert into test_orer_partition_by values (‘SX’, ‘HZ’, 99) ;
insert into test_orer_partition_by values (‘SX’, ‘XY’, 100) ;
insert into test_orer_partition_by values (‘SX’, ‘SL’, 23) ;
3、排序（row_number、rank、dense_rank）
3.1 /* row_number() over(partition by prov_name ORDER BY val_cnt) */
select prov_name, city_name, val_cnt,
row_number() over(partition by prov_name ORDER BY val_cnt) AS rst
from test_orer_partition_by ;

3.2 /** rank() over(partition by prov_name ORDER BY val_cnt) */
select prov_name, city_name, val_cnt,
rank() over(partition by prov_name) AS rst
from test_orer_partition_by ;

3.3 /** dense_rank() over(partition by prov_name ORDER BY val_cnt) */
select prov_name, city_name, val_cnt,
dense_rank() over(partition by prov_name ORDER BY val_cnt) AS rst
from test_orer_partition_by ;

4、聚合函数（sum、avg、count、max、min、first_value、last_value）
/** sum() over([partition by prov_name [ORDER BY val_cnt]]) */
有order by；按照排序连续累加；无order by，计算partition by后的和；over()中没有partition by，计算所有数据总和
select prov_name, city_name, val_cnt,
sum(val_cnt) over(partition by prov_name ORDER BY val_cnt) AS rst
from test_orer_partition_by ;

select prov_name, city_name, val_cnt,
sum(val_cnt) over(partition by prov_name) AS rst
from test_orer_partition_by ;

select prov_name, city_name, val_cnt,
sum(val_cnt) over() AS rst
from test_orer_partition_by ;

vg、count、max、min、first_value、last_value与sum用法一模一样，不在赘述。
5、分许函数（lag、lead）
/** lead() over([partition by prov_name [ORDER BY val_cnt]]) */
lag、lead有三个参数，第一个是表达式或字段，第二个是偏移量，第三个是为控制赋值
select prov_name, city_name, val_cnt,
lead(val_cnt, 1) over(partition BY prov_name ORDER BY val_cnt) AS rst
from test_orer_partition_by ;

select prov_name, city_name, val_cnt,
lead(val_cnt, 2) over(partition BY prov_name ORDER BY val_cnt) AS rst
from test_orer_partition_by ;

lag与lead用法相同，只是偏移顺序相反。