hive窗口函数使用

hive窗口函数基本面试面sql必考项目，这里画一下重点猴,那现在就操练起来~
窗口函数格式：

function_name(field) over (partition by field order by field rows between unbounded preceding and current row)

窗口说明：
n PRECEDING：往前n行数据
n FOLLOWING：往后n行数据
CURRENT ROW：当前行
UNBOUNDED：起点，UNBOUNDED PRECEDING 表示从前面的起点 UNBOUNDED FOLLOWING：表示到后面的终点
默认是自起点到当前行

数据准备：

create table study_test
(
student_id bigint comment '学生',
class_name string comment '课程名称',
score double comment '成绩'
);
insert into table study_test values(1,'语文',66.6)
,(1,'数学',33.6) ,(1,'英语',55.6) 
,(2,'语文',66.6)
,(2,'数学',88.6) ,(2,'英语',35.6) 
,(3,'语文',46.6)
,(3,'数学',58.6) ,(3,'英语',95.6) 
,(4,'语文',96.6)
,(4,'数学',98.6) ,(4,'英语',85.6)
; 

1.常用的排名函数 row_number,rank,dense_rank
1.1 row_number(),排名从1开始增加，无重复并列情况
1.2 rank(),有并列，排名相等会在名次中留下空位
1.3 dense_rank(),有并列，排名相等会在名次中不留空位
测试case:

select 
*,
row_number()over(partition by class_name order by score desc) rnk_1,
rank()over(partition by class_name order by score desc) rnk_1,
dense_rank()over(partition by class_name order by score desc) rnk_1
from study_test

结果：

排序名次

2.sum,avg,min,max等窗口函数
测试case:

select 
*,
row_number()over(partition by class_name order by score desc) rnk , --排名
sum(score)over(partition by class_name order by score desc) as sum_all, -- 分组所有行
sum(score)over(partition by class_name order by score desc) as sum_1, --默认起点到当前行
sum(score)over(partition by class_name order by score desc rows between unbounded preceding and current row)  as sum_2, --同上
sum(score)over(partition by class_name order by score desc rows between 1 preceding and current row) as sum_3, -- 往前一行到当前行
sum(score)over(partition by class_name order by score desc rows between current row and unbounded following) as sum_4, -- 当前行到往后所有行
sum(score)over(partition by class_name order by score desc rows between 1 preceding and 1 following ) as sum_5 -- 往前一行+当前行+往后一行
from study_test
where class_name='数学'

结果：

3.lead(),lag()函数,first_value(),last_value()

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）；
同理，lag(col,n,default) 用于统计窗口内往上第n行值。
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

测试 case:

select 
*,
row_number()over(partition by class_name order by score desc) rnk , --排名
lead(score,1,0.999)over(partition by class_name order by score desc) as lead_1, -- 往后一行，默认0.999
lead(score,1)over(partition by class_name order by score desc) as lead_2, -- 往后一行，默认null
lag(score,1,999.999)over(partition by class_name order by score desc) as lag_1, -- 往前一行，默认999.999
lag(score,1)over(partition by class_name order by score desc) as lag_2, -- 往前一行，默认null
first_value(score)over(partition by class_name order by score desc) as first_data,
last_value(score)over(partition by class_name order by score desc) as last_data,
last_value(score)over(partition by class_name) as last_data_2
from study_test
where class_name='数学'

测试结果：