使用 LEFT JOIN + ORDER BY 时如何避免 FileSort？

Posted 2023-04-19

【中文标题】使用 LEFT JOIN + ORDER BY 时如何避免 FileSort？

【英文标题】：How to avoid FileSort when using LEFT JOIN + ORDER BY?

【发布时间】：2016-03-10 03:25:13

【问题描述】：

mysql-5.6.24-win32.1432006610

我有两个简单的表，TUser(id, name) 和 TMessage(id, uid, message)。 TUser 保存用户，TMessage 保存用户消息。

SQL如下，另见：http://sqlfiddle.com/#!9/7f099

CREATE TABLE TUser(
    id     INT UNSIGNED     PRIMARY KEY    NOT NULL    AUTO_INCREMENT,
    name   VARCHAR(128) NOT NULL
);


CREATE TABLE TMessage(
    id      INT UNSIGNED    PRIMARY KEY    NOT NULL    AUTO_INCREMENT,
    uid     INT UNSIGNED    NOT NULL,
    message VARCHAR(256)    NOT NULL
);


CREATE INDEX TMessageIndexUid ON TMessage(uid);

插入一些数据：

INSERT INTO TUser (name) VALUES 
     ('jack')
    ,('rose')
    ,('peter');


INSERT INTO TMessage(uid, message) VALUES
     (1, 'Hello jack')
    , (1, 'Jack, how are you')
    , (1, 'Good morning jack')
    , (2, 'I love you, rose')
    , (3, 'Peter, please call back')
    , (3, 'What are you doing, Peter');

当我运行以下 LEFT JOIN + ORDER BY 查询时，FileSort 显示在 EXPLAIN 结果中：

EXPLAIN
SELECT *
  FROM        TUser
  LEFT JOIN   TMessage 
  ON          TUser.id=TMessage.uid
  WHERE       TUser.id=3
  ORDER BY    TMessage.id DESC;



id  select_type table    type  possible_keys    key              key_len ref   rows Extra
1   SIMPLE      TUser    const PRIMARY;         PRIMARY          4       const 1    Using temporary; Using filesort
1   SIMPLE      TMessage ref   TMessageIndexUid TMessageIndexUid 4       const 2 \N

有什么问题吗？

【参考方案1】：

文件排序由 ORDER BY 引入：

mysql> EXPLAIN
    -> SELECT *
    ->   FROM        TUser
    ->   LEFT JOIN   TMessage
    ->   ON          TUser.id=TMessage.uid
    ->   WHERE       TUser.id=3
    ->   ORDER BY    TMessage.id DESC;
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+---------------------------------+
| id | select_type | table    | partitions | type  | possible_keys    | key              | key_len | ref   | rows | filtered | Extra                           |
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+---------------------------------+
|  1 | SIMPLE      | TUser    | NULL       | const | PRIMARY          | PRIMARY          | 4       | const |    1 |   100.00 | Using temporary; Using filesort |
|  1 | SIMPLE      | TMessage | NULL       | ref   | TMessageIndexUid | TMessageIndexUid | 4       | const |    4 |   100.00 | NULL                            |
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+---------------------------------+
2 rows in set, 1 warning (0.00 sec)

mysql> EXPLAIN
    -> SELECT *
    ->   FROM        TUser
    ->   LEFT JOIN   TMessage
    ->   ON          TUser.id=TMessage.uid
    ->   WHERE       TUser.id=3;
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+-------+
| id | select_type | table    | partitions | type  | possible_keys    | key              | key_len | ref   | rows | filtered | Extra |
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+-------+
|  1 | SIMPLE      | TUser    | NULL       | const | PRIMARY          | PRIMARY          | 4       | const |    1 |   100.00 | NULL  |
|  1 | SIMPLE      | TMessage | NULL       | ref   | TMessageIndexUid | TMessageIndexUid | 4       | const |    4 |   100.00 | NULL  |
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+-------+
2 rows in set, 1 warning (0.00 sec)

由于 order by 对 LEFT JOIN 结果进行操作，我不知道如何避免这种文件排序。

如果将 LEFT JOIN 更改为 INNER JOIN 是可以接受的，这可能是一种绕过方式，同时会丢失一些与 TMessages 不匹配的用户信息。

mysql> EXPLAIN
    -> SELECT *
    ->   FROM        TUser
    ->   INNER JOIN   TMessage
    ->   ON          TUser.id=TMessage.uid
    ->   WHERE       TUser.id=3
    ->   ORDER BY    TMessage.id DESC;
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+-------------+
| id | select_type | table    | partitions | type  | possible_keys    | key              | key_len | ref   | rows | filtered | Extra       |
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+-------------+
|  1 | SIMPLE      | TUser    | NULL       | const | PRIMARY          | PRIMARY          | 4       | const |    1 |   100.00 | NULL        |
|  1 | SIMPLE      | TMessage | NULL       | ref   | TMessageIndexUid | TMessageIndexUid | 4       | const |    2 |   100.00 | Using where |
+----+-------------+----------+------------+-------+------------------+------------------+---------+-------+------+----------+-------------+
2 rows in set, 1 warning (0.00 sec)

【讨论】：

文件排序可以避免吗？

ORDER BY 有必要吗？

帖子已更新。更改为 INNER JOIN 是否可以接受？ @扎克

谢谢。是不是因为TMessage.id列可能使用LEFT JOIN得到NULL值，所以TMessage.id上的索引不会被使用？

我不太确定原因。正如我猜测的那样，使用 LEFT JOIN 后，行可能与原始左表不同，因此 id 上的原始索引将无法使用。在这种情况下，行号与左表相同是巧合。这取决于连接条件，例如，如果 'ON TUser. id != TMessage.uid' 用作连接条件，行数会有所不同。

【参考方案2】：

我也不知道“文件排序”。然后我发现了类似的东西，

"Anytime a sort can’t be performed from an index, it’s a filesort."

以下是参考链接：

https://www.percona.com/blog/2009/03/05/what-does-using-filesort-mean-in-mysql/ http://s.petrunia.net/blog/?p=24 Avoid filesort with INNER JOIN + ORDER BY