使用窗口函数计算运行总和时缺少的记录。

huangapple go评论52阅读模式
英文:

Missing records for a running sum using window functions

问题

以下是您要的翻译部分:

SELECT Account,  
       Date,  
       SUM(Amount) OVER (PARTITION BY Account ORDER BY Date) AS RunningTotal
FROM tab1;

但它不提供如上所示的数据。如何获得上述输出,以显示截止到最新日期不再存在的帐户的数据?

英文:
Account date Amount
X1 20230501 10
X2 20230501 11
X3 20230501 12
X1 20230502 20
X3 20230502 21
X1 20230503 30
X4 20230503 10

I want to retrieve a running total of "Amount" using windows function to get the following output:

Account date Running Total
X1 20230501 10
X2 20230501 11
X3 20230501 12
X1 20230502 30
X2 20230502 11
X3 20230502 33
X1 20230503 60
X2 20230503 11
X3 20230503 33
X4 20230503 10

I've tried the following SQL approach:

SELECT Account,  
       Date,  
       SUM(Amount) OVER (PARTITION BY Account ORDER BY Date) AS RunningTotal
FROM tab1;

But it does not provide data as above. How do I get output as above to show data for accounts which are no longer as of latest date?

答案1

得分: 3

以下是翻译好的部分:

一种解决此问题的可能方法是:

  • 从您的表中选择所有可能的账户
  • 生成一个在您表格日期范围内的日历表(介于最小日期和最大日期之间)
  • 将所有可能的账户与所有可能的日期进行交叉连接,然后将其左连接到您的表格,并执行您的累积求和
WITH cte_accounts AS(
    SELECT DISTINCT Account FROM tab
), cte_dates AS (
    SELECT MIN(Date) AS date_val, MAX(Date) AS last_date_val FROM tab
    UNION ALL 
    SELECT DATEADD(DAY, 1, date_val), last_date_VAL FROM cte_dates WHERE last_date_val > date_val 
)
SELECT a.Account,  
       d.date_val,  
       SUM(tab.Amount) OVER (PARTITION BY a.Account ORDER BY d.date_val) AS RunningTotal
FROM cte_accounts a
CROSS JOIN cte_dates d
LEFT JOIN tab 
       ON tab.Account = a.Account
      AND tab.date = d.date_val
ORDER BY d.date_val, a.Account

输出

Account date_val RunningTotal
X1 2023-05-01 10
X2 2023-05-01 11
X3 2023-05-01 12
X4 2023-05-01 null
X1 2023-05-02 30
X2 2023-05-02 11
X3 2023-05-02 33
X4 2023-05-02 null
X1 2023-05-03 60
X2 2023-05-03 11
X3 2023-05-03 33
X4 2023-05-03 10

在这里检查演示 链接

最后的 ORDER BY 子句是不必要的,仅用于可视化目的。另外,如果您不希望有空值,可以在子查询中将它们筛选掉。

英文:

One possible way of addressing this problem is to:

  • select all possible accounts from your table
  • generate a calendar table in range (minimum_date, maximum_date) belonging to your table dates
  • cross join all possible accounts with all possible dates, then left join this to your table, and fire your running sum
WITH cte_accounts AS(
    SELECT DISTINCT Account FROM tab
), cte_dates AS (
    SELECT MIN(Date) AS date_val, MAX(Date) AS last_date_val FROM tab
    UNION ALL 
    SELECT DATEADD(DAY, 1, date_val), last_date_VAL FROM cte_dates WHERE last_date_val > date_val 
)
SELECT a.Account,  
       d.date_val,  
       SUM(tab.Amount) OVER (PARTITION BY a.Account ORDER BY d.date_val) AS RunningTotal
FROM       cte_accounts a
CROSS JOIN cte_dates    d
LEFT JOIN tab 
       ON tab.Account = a.Account
      AND tab.date = d.date_val
ORDER BY d.date_val, a.Account

Output

Account date_val RunningTotal
X1 2023-05-01 10
X2 2023-05-01 11
X3 2023-05-01 12
X4 2023-05-01 null
X1 2023-05-02 30
X2 2023-05-02 11
X3 2023-05-02 33
X4 2023-05-02 null
X1 2023-05-03 60
X2 2023-05-03 11
X3 2023-05-03 33
X4 2023-05-03 10

Check the demo here.

The last ORDER BY clause is not necessary: it's just for visualization purposes. On a side note, if you do mind about having the null values, you can filter them out in a subquery.

答案2

得分: 1

以下是翻译好的代码部分:

with AccDates as (
  select Account, min(date) minDate
  from tab1
  group by account
),
Dates as (select distinct date from tab1)
select a.Account, d.date, sum(t.Amount) over (partition by a.Account order by d.Date) RunningTotal
from AccDates a 
left join Dates d on d.date >= a.minDate
left join tab1 t on t.Account = a.Account and t.date = d.date
order by d.date, a.account;

请注意,这只是翻译的代码部分,不包括其他内容。

英文:

Another example, similar to @lemon answer. Added checking for first date for Account. Without full calendar.

with AccDates as(
  select  Account, min(date) minDate
  from tab1
  group by account
)
,Dates as( select distinct date from tab1)

select a.Account
  ,d.date
  ,sum(t.Amount)over(partition by a.Account order by d.Date) RunningTotal
from AccDates a 
left join Dates d on  d.date>=a.minDate
left join tab1 t on t.Account=a.Account and t.date=d.date
order by d.date,a.account
;

Result

Account date RunningTotal
X1 2023-05-01 10
X2 2023-05-01 11
X3 2023-05-01 12
X1 2023-05-02 30
X2 2023-05-02 11
X3 2023-05-02 33
X1 2023-05-03 60
X2 2023-05-03 11
X3 2023-05-03 33
X4 2023-05-03 10

huangapple
  • 本文由 发表于 2023年6月2日 01:30:43
  • 转载请务必保留本文链接:https://go.coder-hub.com/76384349.html
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定