內部聯接導致外部性能不足，有什么不同的方法？

Question

我在一個數據庫中有一個視圖，該視圖從一些表和視圖中檢索以前和當前的案件干事（思維人員）。 問題是這些記錄僅按結束日期（saoh.Date_TO）鏈接，該結束日期與另一個案件官員的開始日期（saoh.Date_FROM）相同。

為了在這些記錄之間創建聯接，我目前正在使用內部聯接進行外部聯接。 （可以在下面的腳本中看到）。 問題是該視圖有約300萬條記錄。 這意味着要查詢此視圖將花費非常長的時間（2-3小時）。

是否有人對如何改進下面的SQL基本設計有任何建議。

更多信息
環境： MSSQL Server 2008
其他信息： Snapshot_Period是用於報告的工具，未鏈接到案件負責人的日期。

   ALTER VIEW [dbo].[vw_Stage_Estate_Case_Officer_Source] AS
    SELECT  sp.SNAPSHOT_PERIOD_START_DATETIME,
            sp.SNAPSHOT_PERIOD_END_DATETIME,
            aes.APPLICATION_RID,
            aes.ESTATE_RID,
            aes.TRUSTEE_NUMBER,
            aes.TRUSTEE_TYPE,
            aes.TEAM_CODE,
            saoh.POSITION,
            saoh.DATE_FROM,
            saoh.DATE_TO,
            saoh.ERROR_CONDITION,
            saoch.USER_ID as PRIOR_CASE_OFFICER
    FROM    Stage_App_Estate_Statuses aes
            /*Standard snapshot period new join for MonthlyITS and yearlyTIS*/
            INNER JOIN  Stage_Snapshot_Period_New sp ON
                        change_date < sp.SNAPSHOT_PERIOD_END_DATETIME and
                        sp.SNAPSHOT_PERIOD_IS_FINALISED_INDICATOR = 'No'and 
                       (sp.SNAPSHOT_PERIOD_TYPE_NAME = 'MonthlyITS' or sp.SNAPSHOT_PERIOD_TYPE_NAME = 'YearlyITS')
            /*This should be inner joining to the staging table that links Case officers to team codes by region*/
            INNER JOIN  [DEV_STAGING].[dbo].[STAGE_STAF_ACTION_OFFICERS] saoh ON
                        aes.TEAM_CODE = saoh.POSITION AND LEFT(aes.ESTATE_RID,3) = LEFT(saoh.ACTION_OFFICER_RID,3)
            /*This should be inner joining to App_Estate_Statues again to get the previous UserID*/
            LEFT OUTER JOIN (SELECT staf.USER_ID,
                                staf.DATE_FROM, 
                                staf.DATE_TO,
                                a.TEAM_CODE,
                                a.ESTATE_RID
                        FROM [DEV_STAGING].[dbo].[STAGE_STAF_ACTION_OFFICERS] staf
                        INNER JOIN  Stage_App_Estate_Statuses a ON
                                    a.TEAM_CODE = staf.POSITION) saoch  ON
                         saoh.DATE_TO = saoch.DATE_FROM AND saoch.ESTATE_RID = aes.ESTATE_RID

    GO

表定義 快照期間：不相關/必須保持原樣。
階段應用程式狀態：

CREATE TABLE [dbo].[Stage_App_Estate_Statuses](
    [ESTATE_STATUS_RID] [nvarchar](20) NOT NULL,
    [ESTATE_RID] [nvarchar](30) NULL,
    [CATEGORY] [decimal](1, 0) NULL,
    [CHANGE_DATE] [datetime2](0) NULL,
    [CHANGE_TYPE] [nvarchar](3) NULL,
    [STATUS] [nvarchar](1) NULL,
    [TEAM_CODE] [nvarchar](4) NULL,
    [TRUSTEE_NUMBER] [decimal](22, 0) NULL,
    [TRUSTEE_TYPE] [nvarchar](10) NULL,
    [APPLICATION_RID] [nvarchar](30) NULL,
    [DML_TYPE] [nvarchar](1) NOT NULL,
    [AUDIT_KEY] [int] NOT NULL
) ON [PRIMARY]

Stage_STAF_ACTION_OFFICERS

CREATE TABLE [dbo].[Stage_STAF_ACTION_OFFICERS](
    [ACTION_OFFICER_RID] [nvarchar](15) NOT NULL,
    [POSITION] [nvarchar](13) NULL,
    [DATE_FROM] [date] NULL,
    [DATE_TO] [date] NULL,
    [USER_ID] [nvarchar](32) NULL,
    [ERROR_CONDITION] [nvarchar](100) NULL,
    [EXTRACTED_DATE] [date] NULL,
    [DML_TYPE] [nvarchar](1) NULL,
    [AUDIT_KEY] [int] NOT NULL
) ON [PRIMARY]

Answer 1

很難沒有任何可測試的東西，所以僅以此為起點

SELECT
    sp.SNAPSHOT_PERIOD_START_DATETIME,
    sp.SNAPSHOT_PERIOD_END_DATETIME,
    aes.APPLICATION_RID,
    aes.ESTATE_RID,
    aes.TRUSTEE_NUMBER,
    aes.TRUSTEE_TYPE,
    aes.TEAM_CODE,
    saoh.POSITION,
    saoh.DATE_FROM,
    saoh.DATE_TO,
    saoh.ERROR_CONDITION,
    (select USER_ID
        from [DEV_STAGING].[dbo].[STAGE_STAF_ACTION_OFFICERS] sao
        where sao.DATE_FROM = saoh.DATE_TO
            and (select ESTATE_RID from Stage_App_Estate_Statuses where TEAM_CODE = sao.POSITION) = aes.ESTATE_RID
    )as PRIOR_CASE_OFFICER --Even if this approach doesnt help, keep the original join as outer because there might not be a prior_case_officer
FROM
    Stage_App_Estate_Statuses aes
    /This should be inner joining to the staging table that links Case officers to team codes by region/
    INNER JOIN [DEV_STAGING].[dbo].[STAGE_STAF_ACTION_OFFICERS] saoh 
        ON aes.TEAM_CODE = saoh.POSITION 
        AND LEFT(aes.ESTATE_RID,3) = LEFT(saoh.ACTION_OFFICER_RID,3)
            /Standard snapshot period new join for MonthlyITS and yearlyTIS/
    INNER JOIN (select

                    SNAPSHOT_PERIOD_START_DATETIME,
                    SNAPSHOT_PERIOD_END_DATETIME,

                from Stage_Snapshot_Period_New
                where 
                    SNAPSHOT_PERIOD_IS_FINALISED_INDICATOR = 'No'
                    and SNAPSHOT_PERIOD_TYPE_NAME in ('MonthlyITS','YearlyITS')
                ) sp
        ON change_date < sp.SNAPSHOT_PERIOD_END_DATETIME 

Answer 2

在這種情況下，您可以將“外部聯接中的內部聯接”轉換為多個左聯接，盡管您的where子句將需要處理子查詢中內部聯接正在處理的情況，例如：

... query same up to the left join ...
            /*This should be inner joining to App_Estate_Statues again to get the previous UserID*/
            LEFT OUTER JOIN [dbo].[STAGE_STAF_ACTION_OFFICERS] saoch
               ON saoh.DATE_TO = saoch.DATE_FROM
                 AND saoch.POSITION = aes.TEAM_CODE
            LEFT OUTER JOIN Stage_App_Estate_Statuses a ON
                 a.TEAM_CODE = saoch.POSITION
                 AND a.ESTATE_RID = aes.ESTATE_RID
    WHERE (saoch.ACTION_OFFICER_RID IS NULL)
       OR (saoch.ACTION_OFFICER_RID IS NOT NULL AND a.ESTATE_STATUS_RID IS NOT NULL)

我簡化了表以刪除不相關的位，並添加了一些虛擬數據。 對於我使用的虛擬數據，此查詢返回的結果與原始查詢相同，但查詢計划有所不同-這意味着您必須對其進行測試，以確定其性能是否更好。

這是我使用的SQLFiddle 。

Answer 3

從以上答案中汲取靈感后，我的最終工作代碼。

WITH curr AS
    (SELECT est.ESTATE_RID
        ,est.TEAM_CODE
        ,staff.POSITION
        ,ISNULL(staff.USER_ID, su.System_User_ID) as USERID
        ,est.CHANGE_DATE
        ,ROW_NUMBER() OVER (ORDER BY est.ESTATE_RID DESC, est.CHANGE_DATE ASC) AS RowNum
        FROM Stage_App_Estate_Statuses est
            LEFT OUTER JOIN Stage_STAF_ACTION_OFFICERS staff ON
            est.TEAM_CODE = staff.POSITION and
            est.CHANGE_DATE BETWEEN staff.DATE_FROM AND staff.DATE_TO
            LEFT OUTER JOIN Stage_System_User_Mapping su on
            est.TEAM_CODE = su.System_User_ID
        WHERE CHANGE_TYPE LIKE '%T%'
    )
SELECT  curr.ESTATE_RID, 
        curr.TEAM_CODE,
        ISNULL(curr.USERID,curr.TEAM_CODE) AS Current_UserID,
        prev.USERID AS Previous_UserID,
        curr.CHANGE_DATE as Date_From,
        helper.CHANGE_DATE as Date_To
FROM curr
LEFT OUTER JOIN curr prev ON    curr.RowNum = (prev.RowNum + 1) and
                                curr.ESTATE_RID = prev.ESTATE_RID                           
LEFT OUTER JOIN curr helper ON curr.RowNum = (helper.RowNum - 1) and
                                curr.ESTATE_RID = helper.ESTATE_RID

ORDER BY ESTATE_RID ASC, curr.CHANGE_DATE ASC