如何在索引列的查询中使用限制而不扫描所有行？

这是我的桌子：

在我的桌子上

• Clustering_key（主键和自增量）
• ID（索引栏）
• Data（文本数据类型列）
• Position（索引栏）维持顺序Data

我的表有 90,000 行，内容相同ID等于 5。我想要前 3 行ID等于 5 我的查询是这样的

Select * from mytable where ID=5 Limit 3;

ID列是索引列所以我认为mysql只扫描前3行但mysql扫描大约42000行。

这里解释一下查询：

任何避免所有行扫描的可能性。

请给我一些解决方案

提前致谢

我模拟了这个场景。

• 使用创建表

   CREATE TABLE mytable (
        Clustering_key INT NOT NULL AUTO_INCREMENT,
        ID INT NOT NULL,
        Data text NOT NULL,
        Position INT NOT NULL,
        PRIMARY KEY (Clustering_key),
        KEY(ID),
        KEY(Position)
    )

  

• 插入的数据与

    INSERT INTO mytable (ID,Data,Position) VALUES (5,CONCAT("Data-",5), 7);
    INSERT INTO mytable (ID,Data,Position) VALUES (5,CONCAT("Data-",5), 26);
    INSERT INTO mytable (ID,Data,Position) VALUES (5,CONCAT("Data-",51), 27);
    INSERT INTO mytable (ID,Data,Position) VALUES (5,CONCAT("Data-",56), 28);
    INSERT INTO mytable (ID,Data,Position) VALUES (5,CONCAT("Data-",57), 31);

  

• Explain

    mysql> explain Select * from mytable where ID=5 Limit 3
    +----+-------------+---------+------------+------+---------------+------+---------+-------+------+----------+-------+
    | id | select_type | table   | partitions | type | possible_keys | key  | key_len | ref   | rows | filtered | Extra |
    +----+-------------+---------+------------+------+---------------+------+---------+-------+------+----------+-------+
    |  1 | SIMPLE      | mytable | NULL       | ref  | ID            | ID   | 4       | const |    5 |   100.00 | NULL  |
    +----+-------------+---------+------------+------+---------------+------+---------+-------+------+----------+-------+
    1 row in set, 1 warning (0.00 sec)

  

是的explain显示检查的行是 5，而不是 3。 但似乎这只是一个误导性的信息。 可以通过以下步骤为所有查询启用慢速日志（设置 long_query_time=0）来验证运行时 rows_examined 的确切数量。

注意：您必须仅在您自己的测试数据库中设置 long_query_time=0。并且测试后必须将参数重置回之前的值。

     - set GLOBAL slow_query_log=1;
     - set global long_query_time=0;
     - set session long_query_time=0;
     mysql> show variables like '%slow%';
    +---------------------------+-------------------------------------------------+
    | Variable_name             | Value                                           |
    +---------------------------+-------------------------------------------------+
    | log_slow_admin_statements | OFF                                             |
    | log_slow_slave_statements | OFF                                             |
    | slow_launch_time          | 2                                               |
    | slow_query_log            | ON                                              |
    | slow_query_log_file       | /usr/local/mysql/data/slow.log                  |
    +---------------------------+-------------------------------------------------+
    5 rows in set (0.10 sec)
    mysql> select @@long_query_time;
    +-------------------+
    | @@long_query_time |
    +-------------------+
    |          0.000000 |
    +-------------------+
      

And then in the terminal, executing the query
<pre>
mysql> Select * from mytable where ID=5 Limit 3;
+----------------+----+---------+----------+
| Clustering_key | ID | Data    | Position |
+----------------+----+---------+----------+
|              5 |  5 | Data-5  |        7 |
|          26293 |  5 | Data-5  |       26 |
|          26294 |  5 | Data-51 |       27 |
+----------------+----+---------+----------+
3 rows in set (0.00 sec)

mysql> Select * from mytable where ID=5 Limit 1;

通过检查来检查慢速日志slow_query_log_file上面印有/usr/local/mysql/data/slow.log

您可以通过以下方式了解信息。

    # Time: 2019-04-26T01:48:19.890846Z
    # User@Host: root[root] @ localhost []  Id:  5124
    # Query_time: 0.000575  Lock_time: 0.000146 Rows_sent: 3  Rows_examined: 3 
    SET timestamp=1556243299;
    Select * from mytable where ID=5 Limit 3;
    # Time: 2019-04-26T01:48:34.672888Z
    # User@Host: root[root] @ localhost []  Id:  5124
    # Query_time: 0.000182  Lock_time: 0.000074 Rows_sent: 1  Rows_examined: 1 
    SET timestamp=1556243314;
    Select * from mytable where ID=5 Limit 1;

  

运行时Rows_exmained值等于limit范围。 测试在MySQL 5.7.18上完成。

----------------------------------另一种验证方式------------------------ ----------------------

    mysql> show status like '%Innodb_rows_read%';
    +------------------+-------+
    | Variable_name    | Value |
    +------------------+-------+
    | Innodb_rows_read | 13    |
    +------------------+-------+
    1 row in set (0.00 sec)

    mysql> Select * from mytable where ID=5 Limit 1;
    +----------------+----+--------+----------+
    | Clustering_key | ID | Data   | Position |
    +----------------+----+--------+----------+
    |              5 |  5 | Data-5 |        7 |
    +----------------+----+--------+----------+
    1 row in set (0.00 sec)

    mysql> show status like '%Innodb_rows_read%';
    +------------------+-------+
    | Variable_name    | Value |
    +------------------+-------+
    | Innodb_rows_read | 14    |
    +------------------+-------+
    1 row in set (0.00 sec)

  

您可以看到Innodb_rows_read只需将限制 1 增加 1 即可。 如果执行全表扫描查询，您可以看到该值将增加表的计数。

    mysql> select count(*) from mytable;
    +----------+
    | count(*) |
    +----------+
    |   126296 |
    +----------+
    1 row in set (0.05 sec)

    mysql> show status like '%Innodb_rows_read%';
    +------------------+--------+
    | Variable_name    | Value  |
    +------------------+--------+
    | Innodb_rows_read | 505204 |
    +------------------+--------+
    1 row in set (0.00 sec)

    mysql> Select * from mytable where Data="Data-5";
    +----------------+----+--------+----------+
    | Clustering_key | ID | Data   | Position |
    +----------------+----+--------+----------+
    |              5 |  5 | Data-5 |        7 |
    |          26293 |  5 | Data-5 |       26 |
    |          26301 |  5 | Data-5 |        7 |
    +----------------+----+--------+----------+
    3 rows in set (0.09 sec)

    mysql> show status like '%Innodb_rows_read%';
    +------------------+--------+
    | Variable_name    | Value  |
    +------------------+--------+
    | Innodb_rows_read | 631500 |
    +------------------+--------+
    1 row in set (0.00 sec)

  

两种方式都证实了explainfor limit 似乎提供了有关所检查行的误导性信息。