SQL 2008 中临时表或子查询的多重联接

Question

我有以下 SQL，它获取一系列日期中每天的季节，然后按开始日期和结束日期以及夜晚数对每个季节进行分组。它的作用并不重要，但我的问题是哪个更好，我在下面完成的方式或每次在第二个查询中使用 @dateSeasons 时使用第一个 select 语句作为子查询。两种方式似乎运行相同，但这种方式看起来更整洁。

DECLARE @dateSeasons TABLE ([date] date, seasonID int)

INSERT INTO @dateSeasons
SELECT D.[date], S.ID
FROM @dates AS D
CROSS APPLY (

    SELECT TOP 1 ID
    FROM dbo.Seasons
    WHERE bookingID = @bookingID 
    AND D.[date] BETWEEN startDate AND endDate
    ORDER BY ID DESC

) AS S


SELECT MIN([date]), endDate, DATEDIFF(DAY, MIN([date]), DATEADD(DAY, 1, endDate)), seasonID
FROM (

    SELECT S1.seasonID, S1.[date], (

        SELECT MAX([date])
        FROM @dateSeasons S2
        WHERE S2.seasonID = S1.seasonID
        AND NOT EXISTS (

            SELECT NULL
            FROM @dateSeasons S3
            WHERE S3.[date] < S2.[date]
            AND S3.[date] > S1.[date]
            AND S3.seasonID <> S1.seasonID

        )

    ) AS endDate
    FROM @dateSeasons S1
) AS results
GROUP BY endDate, seasonID
ORDER BY MIN([date])

Answer 1

看起来整洁与编写 SQL 代码无关。从性能的角度来看，看起来优雅的方法通常是解决问题的最差方式。

确定哪种方法最好的唯一方法是首先确保您测试的两种方法返回相同的结果，然后对它们进行性能测试并检查执行计划（或在 mySQL 中解释）。使查询更好的技术也是特定于数据库的。在 SQL Server 中最适合性能调整的方法可能在 Oracle 中最不可行。

Answer 2

有时，使用公用表表达式 (CTE) 可以获得更好的性能：

WITH
dateSeasons ([date], [seasonID])
AS
(
    SELECT D.[date], S.ID
    FROM @dates AS D
    CROSS APPLY (

        SELECT TOP 1 ID
        FROM dbo.Seasons
        WHERE bookingID = @bookingID 
        AND D.[date] BETWEEN startDate AND endDate
        ORDER BY ID DESC

    ) AS S
)

SELECT MIN([date]), endDate, DATEDIFF(DAY, MIN([date]), DATEADD(DAY, 1, endDate)), seasonID
FROM (

    SELECT S1.seasonID, S1.[date], (

        SELECT MAX([date])
        FROM dateSeasons S2
        WHERE S2.seasonID = S1.seasonID
        AND NOT EXISTS (

            SELECT NULL
            FROM dateSeasons S3
            WHERE S3.[date] < S2.[date]
            AND S3.[date] > S1.[date]
            AND S3.seasonID <> S1.seasonID

        )

    ) AS endDate
    FROM dateSeasons S1
) AS results
GROUP BY endDate, seasonID
ORDER BY MIN([date])