开窗函数的使用详解(聚合函数图文详解)

练习数据：

create table order_tab(
    order_id int,
    user_no varchar(3),
    amount int,
    create_date date
);

insert into order_tab values
(1,'001',100,'2019-01-01'),
(2,'001',300,'2019-01-02'),
(3,'001',500,'2019-01-02'),
(4,'001',800,'2019-01-03'),
(5,'001',900,'2019-01-04'),
(6,'002',500,'2019-01-03'),
(7,'002',600,'2019-01-04'),
(8,'002',300,'2019-01-10'),
(9,'002',800,'2019-01-16'),
(10,'002',800,'2019-01-22');

一、排序开窗函数

① row_number() – 相同值排名顺延，返回结果1、2、3、4
② rank() --相同结果排名相同，后续排名不连续，返回结果为 1、2、2、4
③ dense_rank() – 相同结果排名相同，后续排名顺延，返回结果为1、2、2、3
④ ntile(n) – 分组排名，将数据分为n组并返回对应组号1、2…n NTILE(n)，用于将分组数据按照顺序切分成n片，返回当前切片值。将一个有序的数据集划分为多个桶(bucket)，并为每行分配一个适当的桶数（切片值，第几个切片，第几个分区等概念）。它可用于将数据划分为相等的小切片，为每一行分配该小切片的数字序号。如果切片不均匀，默认增加第一个切片的分布。

select order_id,create_date,user_no,amount,
       row_number() over (partition by user_no order by amount) as rowNumber,
       rank() over (partition by user_no order by amount) as ranKs,
       dense_rank() over (partition by user_no order by amount) as denseRank,
       ntile(3) over (partition by user_no order by amount) as ntiles
from order_tab;

二、聚合开窗函数

① sum() – 分组求和
② count() – 分组求总数
③ min() – 分组求最小值
④ max() – 分组求最大值
⑤ avg() --分组求均值

select order_id,create_date,user_no,amount,
       sum(amount) over (partition by user_no order by amount) as sum_num,
       count(amount) over (partition by user_no order by amount) as count_num,
       min(amount) over (partition by user_no order by amount) as min_num,
       max(amount) over (partition by user_no order by amount) as max_num,
       avg(amount) over (partition by user_no order by amount) as avg_num
from order_tab;

三、其他开窗函数

① lag(字段名,n,0) – 移位开窗函数，表示返回向上第n行指定字段对应数据。其中n代表向上偏移n行，0代表若偏移行数超出表范围则返回0也可以改成其他值，若不写则默认null
② lead(字段名,n,0) – 移位开窗函数，与lag()相反，表示返回向下第n行指定字段对应数据
③ first_value()-- 取分组内排序后，截止到当前行，第一个值
④ last_value() – 取分组内排序后，截止到当前行，最后一个值

select order_id,create_date,user_no,amount,
       lag(amount,2,0) over (partition by user_no order by amount) as lag_num,
       lead(amount,2,0) over (partition by user_no order by amount) as lead_num,
       first_value(amount) over (partition by user_no order by amount) as first_value_num,
       last_value(amount) over (partition by user_no order by amount) as last_value_num
from order_tab;