Databricks SQL 将不接受相关列错误：非等式谓词中不允许相关列

Question

我有一个包含相关列的 T-SQL 查询。当我尝试使用 Databricks SQL 执行查询时，出现错误：

Error in SQL statement: AnalysisException: Correlated column is not allowed in a non-equality predicate:
Aggregate [max(DateOfChange#20975) AS max(DateOfChange)#20988]
+- Filter (((SiiAuditTypeID#20976 = 3) AND (SplitID#20973 = outer(SplitID#20825))) AND outer(StatusID#20829) IN (1,2))
   +- SubqueryAlias sia
      +- SubqueryAlias spark_catalog.dbo.SiiAudit
         +- Relation[SiiAuditID#20972,SplitID#20973,AccountID#20974,DateOfChange#20975,SiiAuditTypeID#20976,UserID#20977,UserName#20978,PortfolioID#20979,PortfolioSplit#20980,SiiAuditEntryKindID#20981,ModelPortfolioID#20982,primary_key_hash#20983,change_key_hash#20984,reject_reason#20985,reject_row#20986] parquet
;
Distinct
+- Project [AccountID#20826, CreatedDate#20827, StatusID#20829, scalar-subquery#20691 [SplitID#20825 && StatusID#20829] AS CancelledDate#20692]
   :  +- Aggregate [max(DateOfChange#20975) AS max(DateOfChange)#20988]
   :     +- Filter (((SiiAuditTypeID#20976 = 3) AND (SplitID#20973 = outer(SplitID#20825))) AND outer(StatusID#20829) IN (1,2))
   :        +- SubqueryAlias sia
   :           +- SubqueryAlias spark_catalog.dbo.SiiAudit

有人可以帮助重构 T-SQL，使其在没有相关/子查询的情况下执行吗？

我认为您不需要成为 Databricks 专家才能帮助解决这个问题。

查询如下：

SELECT DISTINCT
           s.AccountID,
           s.CreatedDate,
           s.StatusID,
           (
               SELECT MAX(DateOfChange) AS CancelledDate
               FROM CRM.InvestmentInstruction.SiiAudit sia
               WHERE sia.SiiAuditTypeID = 3 -- cancelled
                     AND sia.SplitID = s.SplitID
                     AND s.StatusID IN ( 1, 2 ) -- cancelled, completed
           ) AS CancelledDate
    FROM CRM.InvestmentInstruction.Split s
        INNER JOIN CRM.InvestmentInstruction.SplitPortfolio sp
            ON sp.SplitID = s.SplitID
        INNER JOIN CRM.InvestmentInstruction.InvestmentRequest ir
            ON ir.InvestmentRequestID = sp.InvestmentRequestID
        INNER JOIN CRM.dbo.ModelPortfolio mp
            ON mp.ModelPortfolioID = ir.ModelID
        INNER JOIN
        (
            SELECT DISTINCT
                   mh.ModelPortfolioID
            FROM CRM.dbo.modelHolding mh
                INNER JOIN Securities.dbo.Security sec
                    ON sec.SecurityID = mh.LinkSecurityId
            WHERE sec.IsCashSecurity = 0
        ) mh
            ON mh.ModelPortfolioID = mp.ModelPortfolioID
    WHERE s.TypeID = 0

Answer 1

你能用这种方式将条件移到子查询之外吗？:

case when s.StatusID in (1, 2) then (
           SELECT MAX(DateOfChange)
           FROM CRM.InvestmentInstruction.SiiAudit sia
           WHERE sia.SiiAuditTypeID = 3 -- cancelled
                 AND sia.SplitID = s.SplitID) end as CancelledDate