窗口函数中的 order by 是否会添加更多框架？

Question

使用：

sum(Age) over(partition by Country) as Total

Id          FirstName       LastName   Age         Country         Total      
----------- --------------- ---------- ----------- --------------- -----------
          5 Betty           Doe                 28 UAE                      28
          6 Fred            XXX                  1 UK                    99731
          7 Angela          XXX                 37 UK                    99731
          8 Dave            XXX               2000 UK                    99731
          9 Tony            YYY               9000 UK                    99731
         10 Tony            ZZZ                323 UK                    99731
         11 Pete            ZZZ              88323 UK                    99731
          3 David           Robinson            22 UK                    99731
          4 John            Reinhardt           25 UK                    99731
          1 John            Doe                 31 USA                      53
          2 Robert          Luna                22 USA                      53

我得到了我所期望的。这是为了将所有行分为 3 个“帧”（阿联酋、英国和美国），并计算每个帧中的年龄总和。

但是：

sum(Age) over(partition by Country order by LastName) as Total

Id          FirstName       LastName   Age         Country         Total      
----------- --------------- ---------- ----------- --------------- -----------
          5 Betty           Doe                 28 UAE                      28
          4 John            Reinhardt           25 UK                       25
          3 David           Robinson            22 UK                       47
          6 Fred            XXX                  1 UK                     2085
          7 Angela          XXX                 37 UK                     2085
          8 Dave            XXX               2000 UK                     2085
          9 Tony            YYY               9000 UK                    11085
         10 Tony            ZZZ                323 UK                    99731
         11 Pete            ZZZ              88323 UK                    99731
          1 John            Doe                 31 USA                      31
          2 Robert          Luna                22 USA                      53

我不完全确定发生了什么。

看起来好像是使用“County”和“LastName”的组合来创建多个框架，所以英国的看起来是

UK,Reinhardt

UK,Robinson

UK,XXX

UK,YYY

UK,ZZZ

然后计算每个帧中的总数。几乎就像具有 2 个参数的 GroupBy，

Group by Country, LastName

，然后在遍历每一行时将它们相加。

如果是这样的话，下面到底发生了什么？为什么它不像上面那样总结每一帧？

sum(Age) over(partition by Country, LastName order by FirstName) as Total

Id          FirstName       LastName   Age         Country         Total      
----------- --------------- ---------- ----------- --------------- -----------
          5 Betty           Doe                 28 UAE                      28
          4 John            Reinhardt           25 UK                       25
          3 David           Robinson            22 UK                       22
          7 Angela          XXX                 37 UK                       37
          8 Dave            XXX               2000 UK                     2037
          6 Fred            XXX                  1 UK                     2038
          9 Tony            YYY               9000 UK                     9000
         11 Pete            ZZZ              88323 UK                    88323
         10 Tony            ZZZ                323 UK                    88646
          1 John            Doe                 31 USA                      31
          2 Robert          Luna                22 USA                      22

下面是测试代码

-- create
CREATE TABLE Customer (
  Id int,
  FirstName varchar(15),
  LastName varchar(10),
  Age int,
  Country varchar(15)
);

-- insert
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (1, 'John', 'Doe', 31, 'USA');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (2, 'Robert', 'Luna', 22, 'USA');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (3, 'David', 'Robinson', 22, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (4, 'John', 'Reinhardt', 25, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (5, 'Betty', 'Doe', 28, 'UAE');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (6, 'Fred', 'XXX', 1, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (7, 'Angela', 'XXX', 37, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (8, 'Dave', 'XXX', 2000, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (9, 'Tony', 'YYY', 9000, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (10, 'Tony', 'ZZZ', 323, 'UK');
INSERT INTO Customer(Id,FirstName,LastName,Age,Country) VALUES (11, 'Pete', 'ZZZ', 88323, 'UK');


SELECT 
    Id,
    FirstName,
    LastName,
    Age,
    Country,
    sum(Age) over(partition by Country) as Total
    --sum(Age) over(partition by Country order by LastName) as Total
    --sum(Age) over(partition by Country, LastName order by FirstName) as Total
FROM Customer

Answer 1

添加

order by LastName

返回的不是单独的帧，而是直到（并包括）当前行的所有行的累积总和。

使用

partition by Country, LastName order by FirstName

，您可以为每个国家/地区、姓氏组合创建一个框架，并返回该框架中内的累积总和。