Teradata SQL 获取最后一个非空值

Teradata SQL getting last not null value

我在 Teradata 中有一个 table,如下所示:

Account_Num Install_Due_Dt  Install_Num Install_Pay_Dt
12805196    12/08/2019                1     12/08/2019
12805196    10/09/2019                2              ?
12805196    10/10/2019                3              ?
12805196    11/11/2019                4     13/09/2019
12805196    10/12/2019                5              ?

我需要用第一个非空值填充 Install_Pay_Dt 列。例如,它应该是这样的:

Account_Num Install_Due_Dt  Install_Num Install_Pay_Dt
12805196    12/08/2019                1     12/08/2019
12805196    10/09/2019                2     12/08/2019
12805196    10/10/2019                3     12/08/2019
12805196    11/11/2019                4     13/09/2019
12805196    10/12/2019                5     13/09/2019

我使用的是 Teradata 15,所以无法使用延迟。我一直在搜索,但找不到解决方案。 ID 列是 Account_Num,order 列是 Install_num.

我试过这样做:

coalesce(Install_Pay_Dt, MAX(lag_) 
OVER(PARTITION BY 1 ORDER BY Install_Num asc
ROWS BETWEEN 1 PRECEDING AND 1 PRECEDING)) as lag

但它只填满了第二行。

我确定有一种更简洁的方法可以做到这一点,但这里有一个选择:

SELECT 
  acct.Account_Num, 
  acct.Install_Due_Dt, 
  acct.Install_Num, 
  COALESCE(
    acct.Install_Pay_Dt, -- Use original value, if not null
    dt.Install_Pay_Dt_Default -- Use default value if original is null
  ) AS Install_Pay_Dt
FROM MyTable acct
LEFT JOIN (
  SELECT Install_Pay_Dt AS Install_Pay_Dt_Default
  FROM MyTable
  WHERE Install_Pay_Dt IS NOT NULL -- Only get non-NULL values
  QUALIFY ROW_NUMBER() OVER(ORDER BY Install_Num) = 1 -- Only get first row
) dt ON 1=1 -- LEFT JOIN instead of CROSS JOIN in case there are no "install_pay_dt" values

这假设您希望将 "default value" 逻辑作为一个组应用于所有行,这是您发布的查询向我建议的,因为 PARTITION BY 1 将处理单个分区中的所有行。如果您想按 account_num 行组应用逻辑,则必须稍微修改一下。

last_valuelag非常相似,都支持IGNORE NULLS选项

last_value(Install_Pay_Dt IGNORE NULLS)
over (partition by Account_Num 
      order by Install_Num )