sql-server - MSSQL JOIN ON GROUP BY 太慢

Question

我在 MSSQL 中有以下查询

SELECT TOP 50 CustomerID FROM Ratings
WHERE CustomerID != 915
AND MovieID IN (SELECT DISTINCT MovieID FROM Ratings WHERE CustomerID = 915)
GROUP BY CustomerID
ORDER BY count(*) DESC

速度超快。当我尝试在这样的子查询中使用它时。

SELECT * FROM Ratings
WHERE MovieID = 1 AND
CustomerID IN (SELECT TOP 50 CustomerID FROM Ratings
    WHERE CustomerID != 915
    AND MovieID IN (SELECT DISTINCT MovieID FROM Ratings WHERE CustomerID = 915)
    GROUP BY CustomerID
    ORDER BY count(*) DESC)

关于为什么这么慢以及如何加快它的任何想法？我的主键是 (MovieID-CustomerID)，我在 CustomerID 上添加了一个索引

Answer 1

您需要在 (CustomerID, MovieID) 上创建一个额外的 UNIQUE 索引(按此顺序)以改进此查询。

有关性能详细信息，请参阅我的博客中的文章:

• Rating movies

由于您的子查询返回 UNIQUE 值集，因此可以将查询重写为 JOIN:

SELECT  r2.*
FROM    (
        SELECT  TOP 50 CustomerID
        FROM    (
                SELECT  MovieID
                FROM    Ratings
                WHERE   CustomerID = 915
                ) q
        JOIN    Ratings r
        ON      r.MovieID = q.MovieID
                AND CustomerID <> 915
        GROUP BY
                CustomerID
        ORDER BY
                COUNT(*) DESC
        ) ro
JOIN    Ratings r2
ON      r2.MovieID = 1
        AND r2.CustomerID = ro.CustomerID

要从对 Movie 1 进行评分的客户中选择 TOP 50 客户，请使用:

SELECT  r2.*
FROM    (
        SELECT  TOP 50 CustomerID
        FROM    (
                SELECT  MovieID
                FROM    Ratings
                WHERE   CustomerID = 915
                ) q
        JOIN    Ratings r
        ON      r.MovieID = q.MovieID
                AND CustomerID <> 915
                AND EXISTS
                (
                SELECT  1
                FROM    Ratings re
                WHERE   re.MovieID = 1
                        AND re.CustomerID = r.CustomerID
                )
        GROUP BY
                CustomerID
        ORDER BY
                COUNT(*) DESC
        ) ro
JOIN    Ratings r2
ON      r2.MovieID = 1
        AND r2.CustomerID = ro.CustomerID