Question

我需要将第一行中的日期（结束dt）与secord行中的日期（strt dt）进行比较，如果匹配，我需要选择第二行的strt dt。如果它匹配，我需要拿起第一排的strt dt。一个订阅和频道包可以有多个行。我发现在teradata中编写查询以获得所需的输出有困难。

以下是案例和预期结果。

案例1

Subscription    Channel package Start dt    End dt
11111112    Sports  7/3/2015    11/28/2015
11111112    Sports  4/1/2016    11/23/2016
11111112    Sports  11/23/2016  12/17/2017

需要输出

Subscription    Channel package Start dt    End dt
11111112    Sports  4/1/2016    12/17/2017

案例2

Subscription    Channel package Start dt    End dt
11111112    Sports  7/3/2015    11/28/2015
11111112    Sports  11/28/2015  4/1/2016
11111112    Sports  4/1/2016    11/23/2016
11111112    Sports  11/23/2016  12/17/2017

需要输出

Subscription    Channel package Start dt    End dt
11111112    Sports  7/3/2015    12/17/2017

我用

尝试了

MIN(Start dt) OVER (PARTITION BY Subscription   , Channel package   
                    ORDER BY Start dt ROWS BETWEEN 1 FOLLOWING  AND 1 FOLLOWING
                   ) AS NXT Start dt

但是当我在使用qualify时面临一些问题时，我无法继续前进。

Answer 1

我知道这可能是一个较晚的解决方案，也可能不是最快的解决方案。我会这样开始：

(
     SELECT
             TBL.SUBSCRIPTION
         ,   TBL.CHANNEL
         ,   TBL.SUBSCRIPTION || TBL.CHANNEL AS SUB_CHANNEL
         ,   CASE WHEN TBL.END_DT = TBL.START_DT THEN 1 ELSE 0 END AS SAME_DAY_FLAG
         ,   Dense_Rank() Over (ORDER BY SUB_CHANNEL ) AS GRP_RANK
         ,   TBL.START_DT
         ,   TBL.END_DT
         ,   PERIOD(TBL.START_DT, CASE WHEN TBL.END_DT > DATE THEN DATE ELSE TBL.END_DT + 1 END) AS COVERD_DTES
         
     FROM    <YOUR_TABLE> AS TBL
) WITH DATA
ON COMMIT PRESERVE ROWS;

根据需要创建索引。创建 SUB_CHANNEL 的原因是因为一些愚蠢的原因我无法解决，我无法为下面的某些功能拉出多于一列。


/* Create a table that will contain the Merged rows with proper start and end dates combining adjacent periods*/

CREATE VOLATILE TABLE VT_TBL_MERGED_ROWS AS
(
    WITH  CTE (GRP_RANK, COVERD_DTES) AS
    (
        SELECT 
                GRP_RANK, COVERD_DTES
        FROM VT_PRE_CTE 
        
    )
    SELECT * FROM TABLE
    (    
        TD_SYSFNLIB.Td_Normalize_Overlap_Meet(NEW VARIANT_TYPE( CTE.GRP_RANK),  CTE.COVERD_DTES)
        RETURNS (GRP_RANK INT, COVERD_DTES PERIOD(DATE))
        HASH BY  GRP_RANK
        LOCAL ORDER BY GRP_RANK, COVERD_DTES
     )  AS TBL_ROWS
    
)   WITH DATA
PRIMARY INDEX (GRP_RANK);

函数 TD_SYSFNLIB.Td_Normalize_Overlap_Meet 对重叠的结束日期和开始日期进行标准化，并分别为您提供最低和最后开始日期和结束日期的单行。由于我无法理解传递多个列的方法，因此我创建了 GRP_RANK，然后可以将其绑定。


CREATE VOLATILE TABLE VT_PRE_OUTPUT AS 
(SELECT 
                A.SUBSCRIPTION
            ,   A.CHANNEL
            ,   A.GRP_RANK
            ,   B.COVERD_DTES
            ,   Min(A.EFF_DT) AS EFF_DT
            ,   Max(A.EXP_DT) AS EXP_DT
            ,   End(B.COVERD_DTES)- Begin(B.COVERD_DTES) AS TOTAL_DAYS
        FROM    VT_PRE_CTE AS A
        JOIN    VT_TBL_MERGED_ROWS AS B
        ON      A.GRP_RANK = B.GRP_RANK 
        AND     A.COVERD_DTES OVERLAPS B.COVERD_DTES
        GROUP BY 1,2,3) 
        WITH DATA 
        PRIMARY INDEX(MTN,  EXP_DT)
        ON COMMIT PRESERVE ROWS;

如果 A.COVERD_DTES OVERLAPS B.COVERD_DTES 组合之间有中断，这个 SUBSCRIPTION/CHANNEL 实际上会给你多行。您可以决定如何处理多个片段，使用 QUALIFY Row_Number() OVER(PARTITION BY SUBSCRIPTION, CHANNEL ORDER BY MIN_DT ASC) = 1 获取第一次出现或使用 ORDER BY MIN_DT DESC) 获取最新出现。

Teradata / SQL：如何使用teradata SQL和结果

1 个答案: