Question

我有一个SQL查询，必须在PySpark（DataBricks）中进行访问。由于查询复杂，PySpark无法读取相同内容。有人可以检查我的查询并帮助我在不使用“ WITH”语句的情况下以单个“ SELECT”语句编写此查询。

Stage:- 1
promotions="""
(WITH VCTE_Promotions as (SELECT v.Shortname, v.Employee_ID_ALT, v.Job_Level, 
                         v.Management_Level, CAST(sysdatetime() AS date) AS PIT_Date, v.Employee_Status_Alt as Employee_Status, 
                         v.Work_Location_Region, v.Work_Location_Country_Desc, v.HML, 
                         [DM_GlobalStaff].[dbo].[V_Worker_PIT].Is_Manager
FROM           [DM_GlobalStaff].[dbo].[V_Worker_CUR] as v 
LEFT OUTER JOIN
[DM_GlobalStaff].[dbo].[V_Worker_PIT] ON v.Management_Level = [DM_GlobalStaff].[dbo].[V_Worker_PIT].Management_Level),

VCTE_Promotion_v2_Eval as (
SELECT        Employee_ID_ALT,
                             ( SELECT max([pit_date]) AS prior_data 
                               FROM [DM_GlobalStaff].[dbo].[V_Worker_PIT] AS t
                               WHERE (employee_id_alt = a.Employee_ID_ALT) AND (PIT_Date < a.PIT_Date) AND (Is_Manager <> a.Is_Manager) OR
                                      (employee_id_alt = a.Employee_ID_ALT) AND (PIT_Date < a.PIT_Date) AND (Job_Level <> a.Job_Level)) AS prev_job_change_date, Is_Manager
FROM            VCTE_Promotions AS a)

SELECT  VCTE_Promotion_v2_Eval.Employee_ID_ALT, COALESCE (v_cur.Employee_Status_ALT, N'') AS Curr_Emp_Status, 
                         COALESCE (v_cur.Employee_Type, N'') AS Curr_Employee_Type, v_cur.Hire_Date_Alt AS Curr_Hire_Date, 
                         v_cur.Termination_Date_ALT  AS Curr_Termination_Date, COALESCE (v_cur.Termination_Action_ALT, N'') 
                         AS Curr_Termination_Action, cast (v_cur.Job_Level as int) AS Curr_Job_Level, 
                         COALESCE (v_cur.Management_Level, N'') AS Curr_Management_Level, 
                         COALESCE (VCTE_Promotion_v2_Eval.Is_Manager, N'') AS Curr_Ismanager, 
                         CASE WHEN v_m.Job_Level < v_cur.Job_Level OR
                         (VCTE_Promotion_v2_Eval.Is_Manager = 1 AND v_m.Is_Manager = 0 AND v_m.Job_Level <= v_cur.Job_Level) 
                         THEN 'Promotion' WHEN v_m.Job_Level <> v_cur.Job_Level OR
                         VCTE_Promotion_v2_Eval.Is_Manager <> v_m.Is_Manager THEN 'Other' ELSE '' END AS Promotion, v_cur.Tenure, 
                         v_cur.Review_Rating_Current
FROM            VCTE_Promotion_v2_Eval INNER JOIN
                         [DM_GlobalStaff].[dbo].[V_Worker_CUR] as v_cur ON VCTE_Promotion_v2_Eval.Employee_ID_ALT = v_cur.Employee_ID_ALT LEFT OUTER JOIN
                         [DM_GlobalStaff].[dbo].[V_Worker_PIT] as v_m ON VCTE_Promotion_v2_Eval.prev_job_change_date = v_m.PIT_Date AND 
                         VCTE_Promotion_v2_Eval.Employee_ID_ALT = v_m.employee_id_alt
) as pr """

stage-2
promotions = spark.read.jdbc(url=jdbcUrl, table=promotions, properties=connectionProperties)

stage-3
promotions.count()
promotions.show()

从Stage-2查询中获取以下错误：-

com.microsoft.sqlserver.jdbc.SQLServerException: Incorrect syntax near the keyword &apos;WITH&apos;.

---------------------------------------------------------------------------
Py4JJavaError                             Traceback (most recent call last)
<command-2532359884208251> in <module>()
----> 1 promotions = spark.read.jdbc(url=jdbcUrl, table=promotions, properties=connectionProperties)

/databricks/spark/python/pyspark/sql/readwriter.py in jdbc(self, url, table, column, lowerBound, upperBound, numPartitions, predicates, properties)
    533             jpredicates = utils.toJArray(gateway, gateway.jvm.java.lang.String, predicates)
    534             return self._df(self._jreader.jdbc(url, table, jpredicates, jprop))
--> 535         return self._df(self._jreader.jdbc(url, table, jprop))
    536 
    537

我的查询没有问题，这在我的SQL提示符下工作得很好。但是，一旦我在PYSPARK（DataBricks）中使用相同的查询，我就会收到语法错误。您也可以在PySpark语法上帮助我吗？

您的及时帮助将受到高度赞赏。

Answer 1

我无法进行测试，但是请尝试一下，然后比较结果以查看是否所有内容都匹配。

此外，我使用跨应用程序而不是相关子查询，因为没有简单的联接并且相关子查询效率不高，因此，Cross apply应该做得到

(
    SELECT
        VCTE_Promotion_v2_Eval.Employee_ID_ALT
       ,COALESCE(v_cur.Employee_Type, N'') AS Curr_Employee_Type
       ,v_cur.Review_Rating_Current
    (
    SELECT
    Employee_ID_ALT,
    pr.prev_job_change_date,
    IsManager        
    From 
        ( SELECT
            v.Shortname
           ,v.Employee_ID_ALT
           ,v.Job_Level
           ,v.Management_Level
           ,CAST(SYSDATETIME() AS DATE) AS PIT_Date
           ,v.Employee_Status_Alt AS Employee_Status
           ,v.Work_Location_Region
           ,v.Work_Location_Country_Desc
           ,v.HML
           ,dbo.T_Mngmt_Level_IsManager_Mapping.IsManager
        FROM Worker_CUR AS v
        LEFT OUTER JOIN dbo.T_Mngmt_Level_IsManager_Mapping
        ON v.Management_Level = dbo.T_Mngmt_Level_IsManager_Mapping.Management_Level
        ) as VCTE_Promotions a
    Cross APPLY ( 
                 SELECT
                    MAX(PIT_Date) AS prior_data
                 FROM dbo.V_Worker_PIT_with_IsManager AS t
                 WHERE (employee_id_alt = a.Employee_ID_ALT)
                 AND (PIT_Date < a.PIT_Date)
                 AND (IsManager <> a.IsManager)
                 OR (employee_id_alt = a.Employee_ID_ALT)
                 AND (PIT_Date < a.PIT_Date)
                 AND (Job_Level <> a.Job_Level)
                 )
                AS pr
            ) as VCTE_Promotion_v2_Eval
    INNER JOIN [DM_GlobalStaff].[dbo].[V_Worker_CUR] AS v_cur
            ON VCTE_Promotion_v2_Eval.Employee_ID_ALT = v_cur.Employee_ID_ALT
        LEFT OUTER JOIN dbo.V_Worker_PIT_with_IsManager AS v_m
            ON VCTE_Promotion_v2_Eval.prev_job_change_date = v_m.PIT_Date
                AND VCTE_Promotion_v2_Eval.Employee_ID_ALT = v_m.employee_id_alt ) as promotions

使用临时表时，pyspark中出现SQL查询错误

1 个答案: