Hive窗口函数及练习

参考文章：Hive窗口函数详解—及3套案例练习（内容更详细）
1.窗口函数over()的语法结构

分析函数 over (partition by 列名 order by 列名 rows between 开始位置 and 结束位置）

over()函数包括三个函数：
分区 partition by 列名
排序 order by 列名
指定窗口范围 rows between 开始位置 and 结束位置

ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW（表示从起点到当前行）
ROWS BETWEEN 2 PRECEDING AND 1 FOLLOWING（表示往前2行到往后1行）
ROWS BETWEEN 2 PRECEDING AND 1 CURRENT ROW（表示往前2行到当前行）
ROWS BETWEEN CURRENT ROW AND UNBOUNDED FOLLOWING（表示当前行到终点）

这三个函数可以同时使用也可以不使用

over()函数中如果不使用这三个函数，窗口大小是针对查询产生的所有数据，如果指定了分区，窗口大小是针对每个分区的数据。

2.常与over()一起使用的分析函数
(1)聚合类
avg()、sum()、max()、min()

(2)排名类
row_number()、rank() 、dense_rank()

(3)其他类
lag(列名,往前的行数,[行数为null时的默认值，不指定为null])，可以计算用户上次购买时间，或者用户下次购买时间。
lead(列名,往后的行数,[行数为null时的默认值，不指定为null])
ntile(n) 把有序分区中的行分发到指定数据的组中，各个组有编号，编号从1开始，对于每一行，ntile返回此行所属的组的编号

针对一份电商客户订单明细表(orderinfo)进行分析

#1.统计支付订单总数
SELECT * , count(orderId) over() as total  from data.orderinfo
where isPaid = 1;
#最后生成的订单总数会填充成一个新的列
SELECT count(orderId) as total from data.orderinfo
where isPaid = 1;
#最后生成的一个临时表记录订单总数


#2.统计每天支付订单的总数
select *,count(*) over (partition by date_format(paidTime,'%Y-%m-%d')) as daytotal  from data.orderinfo
where isPaid =1 and paidTime <> 0;

#3计算从第一天到当前行的所有price大于4000的订单总数
select*,count(*) over (order by paidTime  rows between unbounded preceding and current row) as total from data.orderinfo
where isPaid =1 
and paidTime <> 0
and price >4000;


#4.查询在2016年4月份购买过的顾客信息及总订单数
select *,count(orderId) over () as total from data.orderinfo
where isPaid = 1
and substr(date_format(paidTime,'%Y-%m-%d'),1,7) = '2016-04'
order by userId;


#5.查询每个顾客的购买明细及月购买总金额
select *,round(sum(price) over (partition by userId,substr(paidTime,1,7)),2) as total_amount from data.orderinfo
where isPaid = 1
and paidTime <> 0;


#6.查询每个顾客的购买明细及到目前为止每个顾客购买总金额
select *,round(sum(price) over (partition by userId order by paidTime rows between unbounded preceding and current row)) 累计金额 from data.orderinfo
where isPaid = 1
and paidTime <> 0;


#7查询顾客上次的购买时间
select *,lag(paidTime,1) over (partition by userId order by paidTime) last_time from data.orderinfo
where isPaid = 1
and paidTime <> 0;


#8查询支付金额前20%的订单信息
select*
from(select *,ntile(5) over (order by price desc) group_num from data.orderinfo
     where isPaid = 1
     and paidTime <> 0) t
where group_num = 1;

针对员工工资信息表(emp)分析

#每个部门员工工资由高到低进行排序(三种排名方法)
SELECT * ,
SAL+ifnull(COMM,0) AS TSAL ,
row_number() over (partition by DEPTNO order by SAL+ifnull(COMM,0) desc) r1,
rank() over (partition by DEPTNO order by SAL+ifnull(COMM,0) desc) r2,
dense_rank() over (partition by DEPTNO order by SAL+ifnull(COMM,0) desc) r3
FROM bjpowernode.emp;

#每个部门工资前三的员工
select *
from (select *,
      SAL+ifnull(COMM,0) AS TSAL,
	  rank() over (partition by DEPTNO order by SAL+ifnull(COMM,0) desc) r
	  from bjpowernode.emp) t
where r <= 3;