有更有效的方法吗？ （从SQL创建XML文件）

Question

我用SQL创建了一个过程，该过程采用了大数据集并输出了大XML文件。由于内存问题，我试图对其进行调整，以使其在较小的子集上运行，然后将其添加到其他位置。我添加了名为“当前”的子查询，该子查询将数据过滤到单个IID实体，然后在该实体上运行相同的代码，但是由于某种原因，它花费的时间比我预期的长得多。

WITH 

    current AS (SELECT * FROM xml.MainDataset WHERE IID = '27'), 
    sources AS  (SELECT DISTINCT stage, IID, name FROM current ),
    pop AS  (SELECT DISTINCT stage, IID, pop, type, plat FROM current ),
    Singles AS  (SELECT DISTINCT stage, IID, pop, type, plat, stype, split1, split2  FROM current  Where split2 = 'N/A'),
    Measures AS     (SELECT DISTINCT stage, IID, pop, type, plat, stype, split1, split2, mType FROM current)


--SELECT stage, 

    SELECT IID as ukprn, name,

        (SELECT pop as pop_category, type, plat,

        -- SINGLES

            (SELECT stype as sCat, split1 as attribute,

                (SELECT mType,

                    (SELECT DISTINCT mDetail, stream1, stream2, stream3, stream4, stream5,

                    FROM current as da
                    WHERE da.stage  = mT.stage
                    AND da.IID      = mT.IID
                    AND da.pop      = mT.pop
                    AND da.type     = mT.type
                    AND da.plat     = mT.plat
                    AND da.stype    = mT.stype
                    AND da.split1   = mT.split1
                    AND da.split2   = mT.split2
                    AND da.mType    = mT.mType
                    FOR XML PATH ('data'), TYPE) [*]


                FROM Measures as mT
                WHERE mT.stage  = sST.stage
                AND mT.IID      = sST.IID
                AND mT.pop      = sST.pop
                AND mT.type     = sST.type
                AND mT.plat     = sST.plat
                AND mT.stype    = sST.stype
                AND mT.split1   = sST.split1
                --AND mT.split2 = sST.split2
                AND mT.split2 = 'N/A'
                FOR XML PATH ('measures'), TYPE) [*]


            FROM Singles as sST
            WHERE sST.stage = Po.stage
            AND sST.IID     = Po.IID
            AND sST.pop     = Po.pop
            AND sST.type    = Po.type
            AND sST.plat    = Po.plat
            FOR XML PATH ('singles'), TYPE) [*],



        FROM pop as Po
        WHERE Po.stage  = Pr.stage
        AND Po.IID      = '27'
        FOR XML PATH ('pop'), TYPE) [*]


    FROM Sources as Pr
    WHERE Pr.stage = 'Primary'
    FOR XML PATH ('source'),



root('ReleaseData')

数据集（xml.MainDataset）非常大，并且具有数百个唯一的IID值，但是，如果将查询分为两部分，它们将非常快：将主数据集过滤到单个IID需要一秒钟。在仅包含一个IID的数据集上运行查询只需要一秒钟。

这是怎么回事？是否会多次创建“当前”数据集？