当前位置：首页 > news >正文

摸鱼大数据——Hive函数14

news 2025/7/13 10:42:15

14、开窗(开列)函数

官网链接：Window Functions - Apache AsterixDB - Apache Software Foundation

14.1 基础使用

开窗函数格式: 开窗函数 over(partition by 分组字段名 [order by 排序字段名 asc|desc] [rows between 开窗开始 and 开窗结束])

partition by: 按照谁进行分组
order by: 对分组后的数据进行排序
rows between and: 限定窗口统计数据范围

开窗函数分类:第一类: 编号相关。row_number(): 123456。不管数据有没有重复，单调递增往后进行编号rank(): 123446。如果遇到相同数据，那么会重复编号，并且会占用后续的编号dense_rank(): 123445。如果遇到相同数据，那么会重复编号，但是不会占用后续的编号第二类: 聚合函数。count()、sum()、avg()、max()、min()....第三类: 取值函数。ntile()、lag()、lead()、first_value()、last_value()

示例：

use day09;

-- 创建表
create table pv_tb(cookieid string,datestr string,pv int
)row format delimited fields terminated by ',';

-- 导入数据
load data inpath '/dir/website_pv_info.txt' into table pv_tb;

-- 验证数据
select * from pv_tb;

-- 编号相关的窗口函数
selectcookieid,datestr,pv,-- row_number：用的最多。单调递增的进行编号，不管重复数据row_number() over(partition by cookieid order by pv asc) as rs1,-- rank：单调递增的进行编号，如果遇到重复数据，编号是相同，同时会占用后面的编号资格rank() over(partition by cookieid order by pv asc) as rs2,-- dense_rank：单调递增的进行编号，如果遇到重复数据，编号是相同，同时不会占用后面的编号资格dense_rank() over(partition by cookieid order by pv asc) as rs3
from pv_tb;


selectcookieid,datestr,pv,row_number() over(partition by cookieid order by pv asc) as rn,-- 如果有order by那么窗口的大小是慢慢逐渐放大的sum(pv) over(partition by cookieid order by pv asc) as sum_result,-- 如果没有order by那么窗口的大小直接彻底放大到最大sum(pv) over(partition by cookieid) as sum_result2
from pv_tb;

窗口的运行原理：

针对sum(pv) over(partition by cookieid order by pv asc) as sum_result语句

14.2 控制数据范围

开窗函数控制范围: rows between 范围开始 and 范围结束

具体的语法含义:1- 范围开始unbounded preceding: 从窗口开始数字 preceding: 前几行数据2- 范围结束unbounded following: 到窗口结束数字 following: 后几行数据3- 特殊的，既能够作为范围开始，也能够作为范围结束current row: 当前行

示例：

-- 控制窗口统计的数据范围
selectcookieid,datestr,pv,sum(pv) over(partition by cookieid order by pv rows between unbounded preceding and current row) as rs1,sum(pv) over(partition by cookieid order by pv rows between 2 preceding and current row) as rs2,sum(pv) over(partition by cookieid order by pv rows between unbounded preceding and unbounded following) as rs3,sum(pv) over(partition by cookieid order by pv rows between 2 preceding and unbounded following) as rs4,sum(pv) over(partition by cookieid order by pv rows between 2 preceding and 2 following) as rs5,sum(pv) over(partition by cookieid order by pv rows between current row and unbounded following) as rs6
from pv_tb;

14.3 其他开窗函数

ntile(n): 将窗口内的数据分配到n个桶里面去，返回的结果是桶的编号。可以使用在数据抽样中

lag: 取窗口中上一行的数据
lead: 取窗口中下一行的数据

first_value: 取窗口中第一行的数据
last_value : 取窗口中最后一行的数据

示例：

-- 其他函数
selectcookieid,datestr,pv,row_number() over(partition by cookieid order by pv asc) as rn,-- ntile(n)：将窗口内的数据分配到n个桶里面去，返回的结果是桶的编号ntile(3) over(partition by cookieid order by pv asc) as rs1,-- 取窗口中上一行的数据lag(pv) over(partition by cookieid order by pv asc) as rs2,-- 取窗口中下一行的数据lead(pv) over(partition by cookieid order by pv asc) as rs3,-- 取窗口中第一行的数据first_value(pv) over(partition by cookieid order by pv asc) as rs4,-- 取窗口中最后一行的数据last_value(pv) over(partition by cookieid order by pv asc) as rs5
from pv_tb;