Question

为了从表中获取记录，我使用这个mysql查询：

SELECT 
    a.id as aid, a.data1 as adata1, a.data2 as adata2
    b.id as bid, b.data1 as bdata1, b.data2 as bdata2
FROM table AS a
JOIN table AS b ON ( a.id <> b.id ) 
WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100
ORDER BY RAND() 
LIMIT 1

此查询准确提取我需要的记录，但遗憾的是由于RAND()这个查询非常慢。

我找到了一些方法，如何避免使用RAND()功能，例如here。但我的问题是，我仍然找不到办法，如何在此查询中替换RAND()函数。在一些简单的查询中，替换RAND()并不是问题，但我不知道如何在上面的示例中执行此操作...因为WHERE子句中有更多条件。

Answer 1

由于您使用的是MySQL，因此可以尝试使用以下SQL查询，这些查询首先从表中获取计数，然后根据该计数选择随机偏移量。然后它准备一个语句，以便可以使用计算的偏移量并执行语句。

SELECT @count := COUNT(*) FROM table AS a JOIN table AS b ON ( a.id <> b.id ) WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100;
SET @offset = CONVERT(FLOOR(RAND() * @count), SIGNED);
PREPARE mystatement FROM "SELECT 
                          a.id as aid, a.data1 as adata1, a.data2 as adata2
                          b.id as bid, b.data1 as bdata1, b.data2 as bdata2
                          FROM table AS a
                          JOIN table AS b ON ( a.id <> b.id ) 
                          WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100 LIMIT ?, 1";
EXECUTE mystatement USING @offset;
DEALLOCATE PREPARE mystatement;

在大型数据集上的执行速度应该比ORDER BY RAND()快，请尝试让我知道......; - ）

修改

查询将无法在phpmyadmin上使用，因此使用MySQL控制台运行它们或编写一个php脚本，其中有两个选项，第一个是让mysql完成工作：

mysql_query('SELECT @count := COUNT(*) FROM table AS a JOIN table AS b ON ( a.id <> b.id ) WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100'); mysql_query('SET @offset = CONVERT(FLOOR(RAND() * @count), SIGNED)'); mysql_query('PREPARE mystatement FROM "SELECT a.id as aid, a.data1 as adata1, a.data2 as adata2 b.id as bid, b.data1 as bdata1, b.data2 as bdata2 FROM table AS a JOIN table AS b ON ( a.id <> b.id ) WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100 LIMIT ?, 1"'); $res = mysql_query('EXECUTE mystatement USING @offset'); $row = mysql_fetch_assoc($res); print_r($row);

可能更快的第二个选项包括使用MySQL完成部分工作，使用编程语言（在我们的例子中是PHP）中完成另一部分：

$res = mysql_query("SELECT COUNT(*) FROM table AS a JOIN table AS b ON ( a.id <> b.id ) WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100')"); $row = mysql_fetch_array($res); $offset = rand(0, $row[0]-1); $res = mysql_query("SELECT a.id as aid, a.data1 as adata1, a.data2 as adata2 b.id as bid, b.data1 as bdata1, b.data2 as bdata2 FROM table AS a JOIN table AS b ON ( a.id <> b.id ) WHERE (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100 LIMIT $offset, 1"); $row = mysql_fetch_assoc($res);

另一种加速我发现的ORDER BY RAND（）的替代方法包括如下查询：

SELECT a.id as aid, a.data1 as adata1, a.data2 as adata2 b.id as bid, b.data1 as bdata1, b.data2 as bdata2 FROM table AS a JOIN table AS b ON ( a.id <> b.id ) WHERE (RAND() < (SELECT ((1/COUNT(*))*10) FROM table AS a JOIN table AS b ON ( a.id <> b.id ) ) ) AND (a.data1=1 AND b.data1=1) AND ABS( a.rating - b.rating ) <100 ORDER BY RAND() LIMIT 1

别忘了告诉我你得到的结果;-)。

Answer 2

你的问题不是很具体。。。桌子有多大？究竟什么是“非常慢”？您正在尝试查找表中的所有记录对，其中data1 = 1且等级的差异小于100.在以下版本中，我将所有条件移至“ON”子句，因此它们更清晰地在一起：

SELECT a.id as aid, a.data1 as adata1, a.data2 as adata2
       b.id as bid, b.data1 as bdata1, b.data2 as bdata2
FROM table AS a join
     table AS b
     ON a.id <> b.id and
        a.data1 = b.data1 and
        a.data1 = 1 and b.data1 = 1 and
        ABS( a.rating - b.rating ) < 100
ORDER BY RAND() 
LIMIT 1

我还添加了附加条件a.data1 = b.data1，因为这有助于SQL引擎将其识别为equijoin，这应该有助于加入性能。

假设data1是选择性的（意味着相对较少的记录具有data1），那么您应该能够使用（data1，id）或（data1，rating）上的索引加快速度。

如果您知道每条记录至少有一个匹配（即每条记录都有另一条具有类似评级的记录），则以下变体应该更好：

SELECT a.id as aid, a.data1 as adata1, a.data2 as adata2
       b.id as bid, b.data1 as bdata1, b.data2 as bdata2
FROM (select *
      from table AS a
      where a.data1 = 1
      order by rand()
      limit 1
     ) a join
     table AS b
     ON a.id <> b.id and
        a.data1 = b.data1 and
        a.data1 = 1 and b.data1 = 1 and
        ABS( a.rating - b.rating ) < 100
ORDER BY RAND() 
LIMIT 1

首先选择随机记录，然后进行自我加入。

这让我觉得你可以对这个问题采取不同的方法，如下所示。首先计算您正在查看的数据的评级。然后选择一个随机的评级对，其中差异小于100，然后找到与那些匹配的随机记录。有了data1和rating的索引，这种方法可能是最快的。

Answer 3

如果你是O.K.在问题空间中分布较不均匀，您可以尝试：

SELECT a.id as aid, a.data1 as adata1, a.data2 as adata2
       b.id as bid, b.data1 as bdata1, b.data2 as bdata2
  FROM ( SELECT *
           FROM table
          WHERE data1 = 1
          ORDER
             BY RAND()
          LIMIT 1
       ) a
  JOIN table b
    ON b.data1 = 1
   AND b.rating BETWEEN a.rating - 100 AND a.rating + 100
 ORDER
    BY RAND()
 LIMIT 1
;

以上将随机选择一条记录为a，然后随机选择一条记录为b。因此，订购和加入的记录要少得多。这不太一致，因为这意味着a的所有选择的可能性将相等，而不是与b的可能相应选择的数量成比例，但也许它足够好目的

MySQL - 如何在更复杂的查询中替换ORDER BY？

3 个答案: