Question

我有这个问题：

SELECT 
      RTPropertyUniqueIdentifier
      ,[FA Unique Listing Identifier - Ref ID]
      ,MLS.[Listing Tracking ID]
      , MLS.[Assessor's Parcel Identification Number] as MLS_PARCELID
      , PP.ParcelID
      , GEOID
      , CASE 
        WHEN PP.PP_ParcelID = MLS.MLS_PARCELID then 'PARCEL MATCH'
        Else 'ADDRESS MATCH' END as MATCHTYPE
      ,[Update Timestamp]
          INTO PROPERTY.DBO.PP_MLS_Bridge
  FROM STAGE.DBO.STAGE_MLS_BRG MLS
  join STAGE.DBO.STAGE_PP_BRG PP on 
    (
        PP.PP_ParcelID = MLS.MLS_PARCELID
    ) OR (
        lower(MLS.StreetNum) = lower(PP.AddNum)
        and (lower(PP.AddStreet) like '%'+lower(MLS.Street_Name)+'%' or lower(MLS.Street_Name) like CONCAT('%', lower(PP.AddStreet), '%'))
        and (lower(PP.AddUnitNum) like '%' + lower(MLS.Unit) + '%' or lower(MLS.Unit) like CONCAT('%', lower(PP.AddUnitNum), '%'))
        and lower(PP.AddCity) = lower(MLS.[Property City])
        and lower(PP.AddState) = lower(MLS.[Property State])
        and lower(PP.AddZip) = lower(MLS.[Property Zip])
);

MLS表强劲有5000万条记录。 PP表的实力为1.85亿。这个查询将连续运行8天以上，这意味着我不确定我是否也进行了优化。我正在寻找一种加快速度的方法。

由于

更新1：执行计划：

更新2：更新了SQL语句

INSERT PROPERTY.DBO.PP_MLS_Bridge
SELECT 
      RTPropertyUniqueIdentifier
      ,[FA Unique Listing Identifier - Ref ID]
      ,MLS.[Listing Tracking ID]
      , MLS.[Assessor's Parcel Identification Number] as MLS_PARCELID
      , PP.ParcelID
      , GEOID
      , CASE 
        WHEN PP.PP_ParcelID = MLS.MLS_PARCELID then 'PARCEL MATCH'
        Else 'ADDRESS MATCH' END as MATCHTYPE
      ,[Update Timestamp]
  FROM STAGE.DBO.STAGE_MLS_BRG MLS
  join STAGE.DBO.STAGE_PP_BRG PP on PP.PP_ParcelID = MLS.MLS_PARCELID
UNION
SELECT 
      RTPropertyUniqueIdentifier
      ,[FA Unique Listing Identifier - Ref ID]
      ,MLS.[Listing Tracking ID]
      , MLS.[Assessor's Parcel Identification Number] as MLS_PARCELID
      , PP.ParcelID
      , GEOID
      , CASE 
        WHEN PP.PP_ParcelID = MLS.MLS_PARCELID then 'PARCEL MATCH'
        Else 'ADDRESS MATCH' END as MATCHTYPE
      ,[Update Timestamp]
  FROM STAGE.DBO.STAGE_MLS_BRG MLS
  join STAGE.DBO.STAGE_PP_BRG PP on
        MLS.StreetNum = PP.AddNum
        and (PP.AddStreet like '%'+MLS.Street_Name+'%' or MLS.Street_Name like CONCAT('%', PP.AddStreet, '%'))
        and (PP.AddUnitNum like '%' + MLS.Unit + '%' or MLS.Unit like CONCAT('%', PP.AddUnitNum, '%'))
        and PP.AddCity = MLS.[Property City]
        and PP.AddState = MLS.[Property State]
        and PP.AddZip = MLS.[Property Zip];

执行计划：第1部分

第2部分

Answer 1

好的，你有2次可怕的表扫描。所以，我可以给你一些建议：

删除LOWER。 SQL Server比较忽略大小写的字符串
在PP.PP_ParcelID和MLS.MLS_PARCELID
使用OR将2个选项替换为您的联接中的UNION - 查询优化程序将选择更好的计划
尝试使用简单整数键（如lower(PP.AddZip) = lower(MLS.[Property Zip])）上的某些连接替换PP.AddZipId = MLS.PropertyZipId之类的字符串上的所有连接。比较整数更简单

尝试基于具有IN连接的四个字段来查询优化连接

1 个答案: