Question

如果我在表格中选择一个不是外键的列，我会找到索引：

mysql> explain SELECT id FROM stats_image_optimization_hourly WHERE stats_image_optimization_hourly.record_status=1;

+----+-------------+---------------------------------+-------+---------------+---------------------------+---------+------+---------+--------------------------+
| id | select_type | table                           | type  | possible_keys | key                       | key_len | ref  | rows    | Extra                    |
+----+-------------+---------------------------------+-------+---------------+---------------------------+---------+------+---------+--------------------------+
|  1 | SIMPLE      | stats_image_optimization_hourly | index | NULL          | image_time_record_status3 | 9       | NULL | 1824413 | Using where; Using index |
+----+-------------+---------------------------------+-------+---------------+---------------------------+---------+------+---------+--------------------------+
1 row in set (0.00 sec)

另一方面，如果选择一个外键列，则找不到索引：

mysql> explain SELECT server_id FROM stats_image_optimization_hourly WHERE stats_image_optimization_hourly.record_status=1;

+----+-------------+---------------------------------+------+---------------+------+---------+------+---------+-------------+
| id | select_type | table                           | type | possible_keys | key  | key_len | ref  | rows    | Extra       |
+----+-------------+---------------------------------+------+---------------+------+---------+------+---------+-------------+
|  1 | SIMPLE      | stats_image_optimization_hourly | ALL  | NULL          | NULL | NULL    | NULL | 1824413 | Using where |
+----+-------------+---------------------------------+------+---------------+------+---------+------+---------+-------------+
1 row in set (0.00 sec)

以下是表格的定义：

stats_image_optimization_hourly | CREATE TABLE `stats_image_optimization_hourly` (
  `id` bigint(20) NOT NULL AUTO_INCREMENT,
  `time_stamp` datetime NOT NULL,
  `site_id` int(11) NOT NULL,
  `server_id` int(11) NOT NULL,
  `device_class_id` int(11) NOT NULL,
  `congestion_level_id` int(11) NOT NULL,
  `network_type_id` int(11) NOT NULL,
  `image_format_id` int(11) NOT NULL,
  `hits` bigint(20) unsigned NOT NULL DEFAULT '1',
  `in_bytes` bigint(20) unsigned NOT NULL DEFAULT '0',
  `out_bytes` bigint(20) unsigned NOT NULL DEFAULT '0',
  `opt_pct` decimal(11,2) NOT NULL DEFAULT '0.00',
  `record_status` tinyint(3) unsigned NOT NULL DEFAULT '1',
  PRIMARY KEY (`id`),
  UNIQUE KEY `time_stamp` (`time_stamp`,`site_id`,`server_id`,`device_class_id`,`congestion_level_id`,`network_type_id`,`image_format_id`),
  KEY `fk_image_device_class_id3` (`device_class_id`),
  KEY `fk_image_congestion_level_id3` (`congestion_level_id`),
  KEY `fk_image_network_type_id3` (`network_type_id`),
  KEY `fk_image_site_id3` (`site_id`),
  KEY `fk_image_server_id3` (`server_id`),
  KEY `fk_image_format3` (`image_format_id`),
  KEY `image_time_record_status3` (`time_stamp`,`record_status`),
  CONSTRAINT `fk_image_congestion_level_id3` FOREIGN KEY (`congestion_level_id`) REFERENCES `congestion_level` (`id`) ON DELETE CASCADE ON UPDATE NO ACTION,
  CONSTRAINT `fk_image_device_class_id3` FOREIGN KEY (`device_class_id`) REFERENCES `device_class` (`id`) ON DELETE CASCADE ON UPDATE NO ACTION,
  CONSTRAINT `fk_image_format3` FOREIGN KEY (`image_format_id`) REFERENCES `image_format` (`id`) ON DELETE CASCADE ON UPDATE NO ACTION,
  CONSTRAINT `fk_image_network_type_id3` FOREIGN KEY (`network_type_id`) REFERENCES `network_type` (`id`) ON DELETE CASCADE ON UPDATE NO ACTION,
  CONSTRAINT `fk_image_server_id3` FOREIGN KEY (`server_id`) REFERENCES `server` (`id`) ON DELETE CASCADE ON UPDATE NO ACTION,
  CONSTRAINT `fk_image_site_id3` FOREIGN KEY (`site_id`) REFERENCES `site` (`id`) ON DELETE CASCADE ON UPDATE NO ACTION
) ENGINE=InnoDB AUTO_INCREMENT=2479561 DEFAULT CHARSET=latin1 |

任何人都可以解释原因吗？

Answer 1

它似乎与索引选择性有关。您在tinyint列上有一个索引，与记录数相比，它具有非常少的不同值。除非索引覆盖了查询（如第一个 - 它的innodb表，并且主键列隐式包含在二级索引中），优化器（基于统计信息）决定全表扫描比索引范围扫描更便宜+抬头。

Answer 2

您id是主键，因此，它是InnoDB表中每个索引的一部分。

当您运行第一个查询时，您需要的所有数据（过滤和选定字段）都包含在索引本身中，因此只需扫描索引即可运行查询。

您可以在解释输出中通过Using where; Using index告诉：using index部分表示它是仅索引扫描。

在您的第二个查询中，server_id不是record_status上索引的一部分。这意味着，如果MySQL选择了索引访问方法，则必须在嵌套循环中查找每个索引记录的相应表记录，以检索server_id的值。

这是一项非常昂贵的操作，并且大多数记录都有record_status = 1（MySQL知道这一点，因为它保留了索引键分布的直方图）。

因此优化器选择仅进行表格式扫描，因为表中的所有数据都存在，并且大多数表记录都必须被检索，但效率较低。

选择外键时，在mysql SELECT中找不到索引

2 个答案: