窗口函数详细用法

仅做记录，原文：https://blog.csdn.net/scgaliguodong123_/article/details/60135385

窗口函数与分析函数
应用场景：
（1）用于分区排序
（2）动态Group By
（3）Top N
（4）累计计算
（5）层次查询

窗口函数
FIRST_VALUE：取分组内排序后，截止到当前行，第一个值
LAST_VALUE：取分组内排序后，截止到当前行，最后一个值
LEAD(col,n,DEFAULT) ：用于统计窗口内往下第n行值。第一个参数为列名，第二个参数为往下第n行（可选，默认为1），第三个参数为默认值（当往下第n行为NULL时候，取默认值，如不指定，则为NULL）
LAG(col,n,DEFAULT) ：与lead相反，用于统计窗口内往上第n行值。第一个参数为列名，第二个参数为往上第n行（可选，默认为1），第三个参数为默认值（当往上第n行为NULL时候，取默认值，如不指定，则为NULL）

OVER从句
1、使用标准的聚合函数COUNT、SUM、MIN、MAX、AVG
2、使用PARTITION BY语句，使用一个或者多个原始数据类型的列
3、使用PARTITION BY与ORDER BY语句，使用一个或者多个数据类型的分区或者排序列
4、使用窗口规范，窗口规范支持以下格式：
---------------------

(ROWS | RANGE) BETWEEN (UNBOUNDED | [num]) PRECEDING AND ([num] PRECEDING | CURRENT ROW | (UNBOUNDED | [num]) FOLLOWING)
(ROWS | RANGE) BETWEEN CURRENT ROW AND (CURRENT ROW | (UNBOUNDED | [num]) FOLLOWING)
(ROWS | RANGE) BETWEEN [num] FOLLOWING AND (UNBOUNDED | [num]) FOLLOWING

当ORDER BY后面缺少窗口从句条件，窗口规范默认是 RANGE BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW.

当ORDER BY和窗口从句都缺失, 窗口规范默认是 ROW BETWEEN UNBOUNDED PRECEDING AND UNBOUNDED FOLLOWING.

OVER从句支持以下函数，但是并不支持和窗口一起使用它们。
Ranking函数: Rank, NTile, DenseRank, CumeDist, PercentRank.
Lead 和 Lag 函数.

分析函数
ROW_NUMBER() 从1开始，按照顺序，生成分组内记录的序列,比如，按照pv降序排列，生成分组内每天的pv名次,ROW_NUMBER()的应用场景非常多，再比如，获取分组内排序第一的记录;获取一个session中的第一条refer等。
RANK() 生成数据项在分组中的排名，排名相等会在名次中留下空位
DENSE_RANK() 生成数据项在分组中的排名，排名相等会在名次中不会留下空位
CUME_DIST 小于等于当前值的行数/分组内总行数。比如，统计小于等于当前薪水的人数，所占总人数的比例
PERCENT_RANK 分组内当前行的RANK值-1/分组内总行数-1
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，如果切片不均匀，默认增加第一个切片的分布。NTILE不支持ROWS BETWEEN，比如 NTILE(2) OVER(PARTITION BY cookieid ORDER BY createtime ROWS BETWEEN 3 PRECEDING AND CURRENT ROW)。

Hive2.1.0及以后支持Distinct
在聚合函数（SUM, COUNT and AVG）中，支持distinct，但是在ORDER BY 或者窗口限制不支持。

1	`COUNT(DISTINCT` `a) OVER (PARTITION` `BY` `c)`

Hive 2.2.0中在使用ORDER BY和窗口限制时支持distinct

1	`COUNT(DISTINCT` `a) OVER (PARTITION` `BY` `c` `ORDER` `BY` `d` `ROWS` `BETWEEN` `1 PRECEDING` `AND` `1 FOLLOWING)`

Hive2.1.0及以后支持在OVER从句中支持聚合函数

SELECT rank() OVER (ORDER BY sum(b))
FROM T
GROUP BY a;

测试数据集：

## COUNT、SUM、MIN、MAX、AVG
select
    user_id,
    user_type,
    sales,
    --默认为从起点到当前行
    sum(sales) OVER(PARTITION BY user_type ORDER BY sales asc) AS sales_1,
    --从起点到当前行，结果与sales_1不同。
    sum(sales) OVER(PARTITION BY user_type ORDER BY sales asc ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS sales_2,
    --当前行+往前3行
    sum(sales) OVER(PARTITION BY user_type ORDER BY sales asc ROWS BETWEEN 3 PRECEDING AND CURRENT ROW) AS sales_3,
    --当前行+往前3行+往后1行
    sum(sales) OVER(PARTITION BY user_type ORDER BY sales asc ROWS BETWEEN 3 PRECEDING AND 1 FOLLOWING) AS sales_4,
    --当前行+往后所有行  
    sum(sales) OVER(PARTITION BY user_type ORDER BY sales asc ROWS BETWEEN CURRENT ROW AND UNBOUNDED FOLLOWING) AS sales_5,
    --分组内所有行
    SUM(sales) OVER(PARTITION BY user_type) AS sales_6                          
from
    order_detail
order by
    user_type,
    sales,
    user_id
 
+----------+------------+--------+----------+----------+----------+----------+----------+----------+--+
| user_id  | user_type  | sales  | sales_1  | sales_2  | sales_3  | sales_4  | sales_5  | sales_6  |
+----------+------------+--------+----------+----------+----------+----------+----------+----------+--+
| liiu     | new        | 1      | 2        | 2        | 2        | 4        | 22       | 23       |
| qibaqiu  | new        | 1      | 2        | 1        | 1        | 2        | 23       | 23       |
| zhangsa  | new        | 2      | 4        | 4        | 4        | 7        | 21       | 23       |
| wanger   | new        | 3      | 7        | 7        | 7        | 12       | 19       | 23       |
| lilisi   | new        | 5      | 17       | 17       | 15       | 21       | 11       | 23       |
| qishili  | new        | 5      | 17       | 12       | 11       | 16       | 16       | 23       |
| wutong   | new        | 6      | 23       | 23       | 19       | 19       | 6        | 23       |
| lisi     | old        | 1      | 1        | 1        | 1        | 3        | 6        | 6        |
| wangshi  | old        | 2      | 3        | 3        | 3        | 6        | 5        | 6        |
| liwei    | old        | 3      | 6        | 6        | 6        | 6        | 3        | 6        |
+----------+------------+--------+----------+----------+----------+----------+----------+----------+--+
 
注意:
结果和ORDER BY相关,默认为升序
如果不指定ROWS BETWEEN,默认为从起点到当前行;
如果不指定ORDER BY，则将分组内所有值累加;
 
关键是理解ROWS BETWEEN含义,也叫做WINDOW子句：
PRECEDING：往前
FOLLOWING：往后
CURRENT ROW：当前行
UNBOUNDED：无界限（起点或终点）
UNBOUNDED PRECEDING：表示从前面的起点 
UNBOUNDED FOLLOWING：表示到后面的终点
其他COUNT、AVG，MIN，MAX，和SUM用法一样。

## first_value与last_value
select
    user_id,
    user_type,
    ROW_NUMBER() OVER(PARTITION BY user_type ORDER BY sales) AS row_num,  
    first_value(user_id) over (partition by user_type order by sales desc) as max_sales_user,
    first_value(user_id) over (partition by user_type order by sales asc) as min_sales_user,
    last_value(user_id) over (partition by user_type order by sales desc) as curr_last_min_user,
    last_value(user_id) over (partition by user_type order by sales asc) as curr_last_max_user
from
    order_detail;
 
+----------+------------+----------+-----------------+-----------------+---------------------+---------------------+--+
| user_id  | user_type  | row_num  | max_sales_user  | min_sales_user  | curr_last_min_user  | curr_last_max_user  |
+----------+------------+----------+-----------------+-----------------+---------------------+---------------------+--+
| wutong   | new        | 7        | wutong          | qibaqiu         | wutong              | wutong              |
| lilisi   | new        | 6        | wutong          | qibaqiu         | qishili             | lilisi              |
| qishili  | new        | 5        | wutong          | qibaqiu         | qishili             | lilisi              |
| wanger   | new        | 4        | wutong          | qibaqiu         | wanger              | wanger              |
| zhangsa  | new        | 3        | wutong          | qibaqiu         | zhangsa             | zhangsa             |
| liiu     | new        | 2        | wutong          | qibaqiu         | qibaqiu             | liiu                |
| qibaqiu  | new        | 1        | wutong          | qibaqiu         | qibaqiu             | liiu                |
| liwei    | old        | 3        | liwei           | lisi            | liwei               | liwei               |
| wangshi  | old        | 2        | liwei           | lisi            | wangshi             | wangshi             |
| lisi     | old        | 1        | liwei           | lisi            | lisi                | lisi                |
+----------+------------+----------+-----------------+-----------------+---------------------+---------------------+--+
 
## lead与lag
select
    user_id,device_id,
    lead(device_id) over (order by sales) as default_after_one_line,
    lag(device_id) over (order by sales) as default_before_one_line,
    lead(device_id,2) over (order by sales) as after_two_line,
    lag(device_id,2,'abc') over (order by sales) as before_two_line
from
    order_detail;
 
+----------+-------------+-------------------------+--------------------------+-----------------+------------------+--+
| user_id  |  device_id  | default_after_one_line  | default_before_one_line  | after_two_line  | before_two_line  |
+----------+-------------+-------------------------+--------------------------+-----------------+------------------+--+
| qibaqiu  | fds         | fdsfagwe                | NULL                     | 543gfd          | abc              |
| liiu     | fdsfagwe    | 543gfd                  | fds                      | f332            | abc              |
| lisi     | 543gfd      | f332                    | fdsfagwe                 | dfsadsa323      | fds              |
| wangshi  | f332        | dfsadsa323              | 543gfd                   | hfd             | fdsfagwe         |
| zhangsa  | dfsadsa323  | hfd                     | f332                     | 65ghf           | 543gfd           |
| liwei    | hfd         | 65ghf                   | dfsadsa323               | fds             | f332             |
| wanger   | 65ghf       | fds                     | hfd                      | dsfgg           | dfsadsa323       |
| qishili  | fds         | dsfgg                   | 65ghf                    | 543gdfsd        | hfd              |
| lilisi   | dsfgg       | 543gdfsd                | fds                      | NULL            | 65ghf            |
| wutong   | 543gdfsd    | NULL                    | dsfgg                    | NULL            | fds              |
+----------+-------------+-------------------------+--------------------------+-----------------+------------------+--+

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115

116

117

118

119

120

121

122

123

124

125

126

127

## RANK、ROW_NUMBER、DENSE_RANK
select
    user_id,user_type,sales,
    RANK() over (partition by user_type order by sales desc) as r,
    ROW_NUMBER() over (partition by user_type order by sales desc) as rn,
    DENSE_RANK() over (partition by user_type order by sales desc) as dr
from
    order_detail;   
 
 
+----------+------------+--------+----+-----+-----+--+
| user_id  | user_type  | sales  | r  | rn  | dr  |
+----------+------------+--------+----+-----+-----+--+
| wutong   | new        | 6      | 1  | 1   | 1   |
| qishili  | new        | 5      | 2  | 2   | 2   |
| lilisi   | new        | 5      | 2  | 3   | 2   |
| wanger   | new        | 3      | 4  | 4   | 3   |
| zhangsa  | new        | 2      | 5  | 5   | 4   |
| qibaqiu  | new        | 1      | 6  | 6   | 5   |
| liiu     | new        | 1      | 6  | 7   | 5   |
| liwei    | old        | 3      | 1  | 1   | 1   |
| wangshi  | old        | 2      | 2  | 2   | 2   |
| lisi     | old        | 1      | 3  | 3   | 3   |
+----------+------------+--------+----+-----+-----+--+  
 
## NTILE
 
select
    user_type,sales,
    --分组内将数据分成2片
    NTILE(2) OVER(PARTITION BY user_type ORDER BY sales) AS nt2,
    --分组内将数据分成3片    
    NTILE(3) OVER(PARTITION BY user_type ORDER BY sales) AS nt3,
    --分组内将数据分成4片    
    NTILE(4) OVER(PARTITION BY user_type ORDER BY sales) AS nt4,
    --将所有数据分成4片
    NTILE(4) OVER(ORDER BY sales) AS all_nt4
from
    order_detail
order by
    user_type,
    sales
 
 
+------------+--------+------+------+------+----------+--+
| user_type  | sales  | nt2  | nt3  | nt4  | all_nt4  |
+------------+--------+------+------+------+----------+--+
| new        | 1      | 1    | 1    | 1    | 1        |
| new        | 1      | 1    | 1    | 1    | 1        |
| new        | 2      | 1    | 1    | 2    | 2        |
| new        | 3      | 1    | 2    | 2    | 3        |
| new        | 5      | 2    | 2    | 3    | 4        |
| new        | 5      | 2    | 3    | 3    | 3        |
| new        | 6      | 2    | 3    | 4    | 4        |
| old        | 1      | 1    | 1    | 1    | 1        |
| old        | 2      | 1    | 2    | 2    | 2        |
| old        | 3      | 2    | 3    | 3    | 2        |
+------------+--------+------+------+------+----------+--+
 
 
求取sale前20%的用户ID
 
select
    user_id
from
(
    select
        user_id,
        NTILE(5) OVER(ORDER BY sales desc) AS nt
    from
        order_detail
)A
where nt=1;
 
## CUME_DIST、PERCENT_RANK 
 
select
user_id,user_type,sales,
--没有partition,所有数据均为1组
CUME_DIST() OVER(ORDER BY sales) AS cd1,
--按照user_type进行分组
CUME_DIST() OVER(PARTITION BY user_type ORDER BY sales) AS cd2 
from
order_detail;   
 
+----------+------------+--------+------+----------------------+--+
| user_id  | user_type  | sales  | cd1  |         cd2          |
+----------+------------+--------+------+----------------------+--+
| liiu     | new        | 1      | 0.3  | 0.2857142857142857   |
| qibaqiu  | new        | 1      | 0.3  | 0.2857142857142857   |
| zhangsa  | new        | 2      | 0.5  | 0.42857142857142855  |
| wanger   | new        | 3      | 0.7  | 0.5714285714285714   |
| lilisi   | new        | 5      | 0.9  | 0.8571428571428571   |
| qishili  | new        | 5      | 0.9  | 0.8571428571428571   |
| wutong   | new        | 6      | 1.0  | 1.0                  |
| lisi     | old        | 1      | 0.3  | 0.3333333333333333   |
| wangshi  | old        | 2      | 0.5  | 0.6666666666666666   |
| liwei    | old        | 3      | 0.7  | 1.0                  |
+----------+------------+--------+------+----------------------+--+
 
 
select
user_type,sales
--分组内总行数      
SUM(1) OVER(PARTITION BY user_type) AS s, 
--RANK值  
RANK() OVER(ORDER BY sales) AS r,    
PERCENT_RANK() OVER(ORDER BY sales) AS pr,
--分组内     
PERCENT_RANK() OVER(PARTITION BY user_type ORDER BY sales) AS prg 
from
order_detail;   
 
+----+-----+---------------------+---------------------+--+
| s  |  r  |         pr          |         prg         |
+----+-----+---------------------+---------------------+--+
| 7  | 1   | 0.0                 | 0.0                 |
| 7  | 1   | 0.0                 | 0.0                 |
| 7  | 4   | 0.3333333333333333  | 0.3333333333333333  |
| 7  | 6   | 0.5555555555555556  | 0.5                 |
| 7  | 8   | 0.7777777777777778  | 0.6666666666666666  |
| 7  | 8   | 0.7777777777777778  | 0.6666666666666666  |
| 7  | 10  | 1.0                 | 1.0                 |
| 3  | 1   | 0.0                 | 0.0                 |
| 3  | 4   | 0.3333333333333333  | 0.5                 |
| 3  | 6   | 0.5555555555555556  | 1.0                 |
+----+-----+---------------------+---------------------+--+

posted @ 2020-04-14 13:34 星朝阅读(2568) 评论(0) 编辑收藏举报

刷新页面返回顶部

星朝

窗口函数详细用法

公告