Getting started with window functions

1. When to use windowing function ？

Windowing functions are often used in conjunction with aggregate functions , Generally speaking, the number of rows after aggregation is less than that before aggregation , But sometimes we want to show the data before aggregation , And display the aggregated data , Now we introduce the window function .

for example ：

+-------+-------------+-------+---------------+--+
| name  |  orderdate  | cost  | sum_window_0  |
+-------+-------------+-------+---------------+--+
| jack  | 2017-01-01  | 10    | 205           |
| jack  | 2017-01-08  | 55    | 205           |
| tony  | 2017-01-07  | 50    | 205           |
| jack  | 2017-01-05  | 46    | 205           |
| tony  | 2017-01-04  | 29    | 205           |
| tony  | 2017-01-02  | 15    | 205           |
| jack  | 2017-02-03  | 23    | 23            |
| mart  | 2017-04-13  | 94    | 341           |
| jack  | 2017-04-06  | 42    | 341           |
| mart  | 2017-04-11  | 75    | 341           |
| mart  | 2017-04-09  | 68    | 341           |
| mart  | 2017-04-08  | 62    | 341           |
| neil  | 2017-05-10  | 12    | 12            |
| neil  | 2017-06-12  | 80    | 80            |
+-------+-------------+-------+---------------+--

2. Syntax of window function ：

UDAF() over (PARTITION By col1,col2 order by col3 Window clause （rows between .. and ..）) AS Column alias

Be careful ：PARTITION By It can be followed by multiple fields ,order By Just follow one field .

3.over() The role of

over() Determines the aggregation range of the aggregation function , By default, the data in the whole window is aggregated , The aggregate function calls each piece of data once .

for example ：
select name, orderdate, cost, sum(cost) over()
from business; 

4.partition by Clause ：

Use Partiton by Clause to partition data , It can be used paritition by Aggregate the in the area .

for example ：
select name, orderdate, cost, sum(cost) over(partition by name)
from business;

5.order by Clause ：

effect ：
（1） Sort the data in the partition ;
（2） Determine which rows to aggregate （ By default, the aggregation from the starting point to the current row ）

for example ：
select name, orderdate, cost, sum(cost) over(partition by name order by orderdate)
from business;

6. Window clause

CURRENT ROW： Current row
n PRECEDING： forward n Row data
n FOLLOWING： Back up n Row data
UNBOUNDED： The starting point ,UNBOUNDED PRECEDING From the front , UNBOUNDED FOLLOWING To the end

By using partition by Clause partitions the data . If you want to make a more detailed dynamic division of windows ,
We need to introduce window clauses .

for example ：
select name, orderdate,cost, sum(cost) 
over(partition by name order by orderdate rows between UNBOUNDED PRECEDING and CURRENT ROW)
from business;

7. Some attention ：

（1）order by Must be with the partition by after ;
（2）Rows Must be with the Order by Son ;
（3）（partition by .. order by） Can be replaced by （distribute by .. sort by ..）