我可以对多个 window 函数调用使用一个 PARTITION 定义吗？

Question

这是我的查询。我对 14 个结果列使用具有相同分区定义的相同 window 函数 sum()：

select id,weekly,
  sum(totalsteps) OVER (PARTITION BY id, weekly) as total_steps, 
  sum(totaldistance) OVER (PARTITION BY id, weekly) as total_distance,
  sum(veryactiveminutes) OVER (PARTITION BY id, weekly) as total_veryactive,  
  sum(fairlyactiveminutes) OVER (PARTITION BY id, weekly) as total_fairlyactive, 
  sum(lightlyactiveminutes) OVER (PARTITION BY id, weekly) as total_lightlyactive, 
  sum(totalsteps)  OVER (PARTITION BY id, weekly) as total_steps,
  sum(totaldistance) OVER (PARTITION BY id, weekly) as total_distance, 
  sum(veryactivedistance) OVER (PARTITION BY id, weekly) as total_veryactivedistance, 
  sum(moderatelyactivedistance) OVER (PARTITION BY id, weekly) as total_moderatelyactivedistance, 
  sum(lightactivedistance) OVER (PARTITION BY id, weekly) as total_lightactivedistance,  
  sum(sedentaryactivedistance) OVER (PARTITION BY id, weekly) as total_sedentaryactivedistance, 
  sum(calories) OVER (PARTITION BY id, weekly) as total_calories, 
  sum(totalminutesasleep) OVER (PARTITION BY id, weekly) as total_asleep, 
  sum(totaltimeinbed)  OVER (PARTITION BY id, weekly) as total_inbed
from (select *, date_trunc('week', activitydate) as weekly
  from activitysleep_merged 
  ) as weeklysum

每个金额都必须拼写 OVER (PARTITION BY id, weekly) 吗？
有没有更好的方法来重写我的查询？

Answer 1

你可以尝试使用WINDOW子句，可选的WINDOW子句有一般形式

WINDOW window_name AS ( window_definition ) [, ...]

然后用OVER window_name做你的聚合函数，可能更优雅

 select id,weekly,
    sum(totalsteps) over w as  total_steps, 
    sum(totaldistance) over w as  total_distance,
    sum(veryactiveminutes) over w as  total_veryactive, 
    sum(fairlyactiveminutes) over w as  total_fairlyactive, 
    sum(lightlyactiveminutes) over w as  total_lightlyactive, 
    sum(totalsteps) over w as  total_steps,
    sum(totaldistance) over w as  total_distance,
    sum(veryactivedistance) over w as  total_veryactivedistance, 
    sum(moderatelyactivedistance) over w as  total_moderatelyactivedistance, 
    sum(lightactivedistance) over w as  total_lightactivedistance, 
    sum(sedentaryactivedistance) over w as  total_sedentaryactivedistance, 
    sum(calories) over w as  total_calories, 
    sum(totalminutesover w as leep) over w as  total_over w as leep, 
    sum(totaltimeinbed) over w as  total_inbed 
from (
    select *, date_trunc('week', activitydate) as weekly
    from activitysleep_merged 
) WINDOW w AS ( PARTITION BY id, weekly );

更详细的可以参考WINDOW Clause

SQLfiddle

Answer 2

您可以使用 WINDOW clause.

SELECT id, weekly
     , sum(totalsteps)    OVER w AS total_steps
     , sum(totaldistance) OVER w AS total_distance
     , ...
FROM  (SELECT *, date_trunc('week', activitydate) AS weekly FROM activitysleep_merged ) AS weeklysum
WINDOW w AS (PARTITION BY id, weekly);   -- !

您仍然需要重复 OVER 关键字，但可以将分区的实际定义替换为 WINDOW 子句中声明的标识符。

这是语法快捷方式，对性能没有影响。在任何一种情况下，Postgres 都会 re-use 相同的分区。

聚合？

也就是说，您的查询看起来很可疑不想要 window 函数开始，但是 普通聚合 .在使用它时，您实际上 不需要 该子查询：

SELECT id, date_trunc('week', activitydate) AS weekly
     , sum(totalsteps)    AS total_steps
     , sum(totaldistance) AS total_distance
     , ...
FROM   activitysleep_merged
GROUP  BY 1, 2   -- !
ORDER  BY 1, 2   --  or BY 2, 1 ?

并且您需要订购结果。

这会为每个 (id, weekly) 生成一个（聚合的）行 - 与原始查询相反，returns 每个输入行一行。

我加入了位置引用作为语法快捷方式，因为这个问题是关于短语法的。相关：

Select first row in each GROUP BY group?

我可以对多个 window 函数调用使用一个 PARTITION 定义吗？

Can I use one PARTITION definition for multiple window function calls?

sql

postgresql

syntax

window-functions

partition-by

聚合？