使用 UNPIVOT 将代码从 sql 重写为 redshift

Posted 2023-03-31

【中文标题】使用 UNPIVOT 将代码从 sql 重写为 redshift

【英文标题】：Rewrite code with UNPIVOT from sql to redshift

【发布时间】：2018-09-11 14:42:44

【问题描述】：

我有需要转换为 redshift 的 SQL 服务器脚本。

我转换了一部分

这是我遇到问题的脚本的一部分。

    SELECT  a.*,
        b.*

FROM    (

SELECT  u.ContactId,
        u.Description,
        CONVERT(Float,u.SeatCharge) AS SeatCharge
        --CAST(u.SeatCharge AS numeric (18,4)) AS SeatCharge,
        --CONVERT(INT, CASE WHEN IsNumeric(CONVERT(VARCHAR(12), u.SeatCharge)) = 1 then CONVERT(VARCHAR(12), u.SeatCharge) else 0 End)

FROM(

SELECT  md.contactid,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7172 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Core,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7182 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_RCM,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7192 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Advanced_RCM,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7183 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Payroll,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7184 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_DXM,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7185 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Messaging,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7186 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Tasks,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7173 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Core,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7187 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Data_Collection,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7189 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Auditing_Tools,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7190 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Group_Sessions,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7191 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_ABC_Data,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7211 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_AGA,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 4495 THEN CONVERT(float,Value) ELSE 0 END)) AS LMS_Core,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 7175 THEN CONVERT(float,Value) ELSE 0 END)) AS HRIS_PTO,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 8106 THEN CONVERT(float,Value) ELSE 0 END)) AS LMS_Course_Groups,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 8286 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Reach_Me,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 8999 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Redshift,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 9155 THEN CONVERT(float,Value) ELSE 0 END)) AS HRIS_Benefits,
                CONVERT(float, MAX(CASE WHEN md.fieldid= 9156 THEN CONVERT(float,Value) ELSE 0 END)) AS HRIS_Assets
        FROM    public.contact_meta md
        WHERE   md.fieldid IN (4495,7172,7182,7192,7183,7184,7185,7186,7173,7187,7189,7190,7191,7175,7211,7212,8106,8286,8999,9155,9156)
                --AND md.ContactId = 75337
        GROUP BY
                md.contactid
                )s

--UNPIVOT
--(
--SeatCharge
--FOR Description IN (
  --      PM_Core,
    --    PM_RCM,
      --  PM_Advanced_RCM,
       -- PM_Payroll,
       -- PM_DXM,
       -- PM_Messaging,
       -- PM_Tasks,

        --HRIS_Assets)
--)u



WHERE           u.SeatCharge <> ''

        )a

一切正常，而不是 UNPIVOT 功能。

据我所知，redshift 没有 UNPIVOT 功能。

所以现在我需要正确地重写它吗？

感谢您的帮助。

【参考方案1】：

考虑 UNNEST(ARRAY(...)) 方法，从 RedShift 最初派生的 PostgreSQL 方言中借用 approach by @Stew。具体来说，首先在列的字符串文字名称数组上调用UNNEST，然后再次在实际列的数组上调用以在单独的行中返回指示符和值。

SELECT  s.ContactId,
        UNNEST(ARRAY('PM_Core', 'PM_RCM', 'PM_Advanced_RCM', 'PM_Payroll', 'PM_DXM', 
                     'PM_Messaging', 'PM_Tasks', 'Clinical_Core', 'Clinical_Data_Collection',  
                     'Clinical_Auditing_Tools', 'Clinical_Group_Sessions', 'Clinical_Group_Sessions', 
                     'Clinical_ABC_Data', 'Clinical_AGA', 'LMS_Core', 'HRIS_PTO', 'LMS_Course_Groups',
                     'PM_Reach_Me', 'PM_Redshift', 'HRIS_Benefits', 'HRIS_Assets')) AS Description,
        UNNEST(ARRAY(PM_Core, PM_RCM, PM_Advanced_RCM, PM_Payroll, PM_DXM, 
                     PM_Messaging, PM_Tasks, Clinical_Core, Clinical_Data_Collection,  
                     Clinical_Auditing_Tools, Clinical_Group_Sessions, Clinical_Group_Sessions, 
                     Clinical_ABC_Data, Clinical_AGA, LMS_Core, HRIS_PTO, LMS_Course_Groups,
                     PM_Reach_Me, PM_Redshift, HRIS_Benefits, HRIS_Assets)) AS SeatCharge    
FROM
   (SELECT md.contactid,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7172 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Core,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7182 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_RCM,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7192 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Advanced_RCM,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7183 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Payroll,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7184 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_DXM,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7185 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Messaging,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7186 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Tasks,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7173 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Core,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7187 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Data_Collection,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7189 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Auditing_Tools,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7190 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_Group_Sessions,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7191 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_ABC_Data,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7211 THEN CONVERT(float,Value) ELSE 0 END)) AS Clinical_AGA,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 4495 THEN CONVERT(float,Value) ELSE 0 END)) AS LMS_Core,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 7175 THEN CONVERT(float,Value) ELSE 0 END)) AS HRIS_PTO,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 8106 THEN CONVERT(float,Value) ELSE 0 END)) AS LMS_Course_Groups,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 8286 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Reach_Me,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 8999 THEN CONVERT(float,Value) ELSE 0 END)) AS PM_Redshift,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 9155 THEN CONVERT(float,Value) ELSE 0 END)) AS HRIS_Benefits,
           CONVERT(float, MAX(CASE WHEN md.fieldid= 9156 THEN CONVERT(float,Value) ELSE 0 END)) AS HRIS_Assets
   FROM    public.contact_meta md
   WHERE   md.fieldid IN (4495,7172,7182,7192,7183,7184,7185,7186,7173,7187,7189,7190,7191,7175,7211,7212,8106,8286,8999,9155,9156)
   --AND   md.ContactId = 75337
   GROUP BY
           md.contactid
   ) s 

Random Data Demo

---

但是，请重新考虑先进行旋转，然后再取消旋转。只需使用CASE 有条件地创建您的指标列，然后聚合：

SELECT sub.contactid, sub.Description, MAX(sub."VALUE") As SeatCharge
FROM
       (SELECT md.contactid,
               CASE md.fieldid
                    WHEN 7172 THEN 'PM_Core'
                    WHEN 7182 THEN 'PM_RCM'
                    WHEN 7192 THEN 'PM_Advanced_RCM'
                    WHEN 7183 THEN 'PM_Payroll'
                    WHEN 7184 THEN 'PM_DXM'
                    WHEN 7185 THEN 'PM_Messaging'
                    WHEN 7186 THEN 'PM_Tasks'
                    WHEN 7173 THEN 'Clinical_Core'
                    WHEN 7187 THEN 'Clinical_Data_Collection'
                    WHEN 7189 THEN 'Clinical_Auditing_Tools'
                    WHEN 7190 THEN 'Clinical_Group_Sessions'
                    WHEN 7191 THEN 'Clinical_ABC_Data'
                    WHEN 7211 THEN 'Clinical_AGA'
                    WHEN 4495 THEN 'LMS_Core'
                    WHEN 7175 THEN 'HRIS_PTO'
                    WHEN 8106 THEN 'LMS_Course_Groups'
                    WHEN 8286 THEN 'PM_Reach_Me'
                    WHEN 8999 THEN 'PM_Redshift'
                    WHEN 9155 THEN 'HRIS_Benefits'
                    WHEN 9156 THEN 'HRIS_Assets'
                    ELSE NULL
               END AS Description,
               CONVERT(float, Value) AS VALUE
       FROM    public.contact_meta md
       WHERE   md.fieldid IN (4495, 7172, 7182, 7192, 7183, 7184,7185, 7186, 7173, 7187, 7189,
                              7190, 7191, 7175, 7211, 7212,8106, 8286, 8999, 9155, 9156)
       --AND   md.ContactId = 75337
       ) AS sub
GROUP BY sub.contactid, sub.Description

注意：将来，请务必合并一个 Description 值的查找表，您可以将这些 id 加入其中，并避免冗长的 CASE 逻辑计算，甚至是冗长的 IN 子句。

【讨论】：

嗯。我收到此错误[42601][500310] [Amazon](500310) Invalid operation: syntax error at or near "'PM_Core'" Position: 108;

@EugeneSukh Unsupported PostgreSQL Features: 数组构造函数 - 这是 Redshift 问题不应使用 postgresql 标记的原因之一

@EugeneSukh ... ANSI SQL 替代方案是每列有许多 UNION 语句。

那么我需要如何正确地重写它，使用 UNION？

好的。那么我需要如何重写这部分，使其在 Redashift 中工作？ @a_horse_with_no_name