为什么mysql索引中的基数值不等于列值的不同计数

时间:2013-05-24 10:34:01

标签: mysql indexing query-optimization cardinality

不久前我开始优化mysql db的查询。我创建了一些索引并决定使用 show index from 语法来查看它们的参数。并且看到基数不等于某些列的明显计数。

所以,我有桌子

CREATE TABLE `item_owners` (
  `uid` varchar(255) NOT NULL,
  `version` bigint(20) NOT NULL,
  `type_id` varchar(255) NOT NULL,
  `owner_name` varchar(255) NOT NULL,
  `item_id` varchar(255) NOT NULL,
  `status_id` varchar(255) NOT NULL,
  PRIMARY KEY (`uid`),
  KEY `FK181EADBC7346EE24` (`status_id`),
  KEY `FK181EADBC90094D43` (`type_id`),
  KEY `FK181EADBC499E38CA` (`item_id`),
  CONSTRAINT `FK181EADBC499E38CA` FOREIGN KEY (`item_id`) REFERENCES `items` (`uid`),
  CONSTRAINT `FK181EADBC7346EE24` FOREIGN KEY (`status_id`) REFERENCES `statuses` (`uid`),
  CONSTRAINT `FK181EADBC90094D43` FOREIGN KEY (`type_id`) REFERENCES `types` (`uid`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8

在那张表中我有2M行(实际上是2M uniq PK),500k uniq item_id,5 uniq status_id和20 uniq type_id

但是当我执行

show index from item_owners

我有

+-------------+------------+--------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table       | Non_unique | Key_name           | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+-------------+------------+--------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| item_owners |          0 | PRIMARY            |            1 | uid         | A         |     1893427 |     NULL | NULL   |      | BTREE      |         |               |
| item_owners |          1 | FK181EADBC7346EE24 |            1 | status_id   | A         |          17 |     NULL | NULL   |      | BTREE      |         |               |
| item_owners |          1 | FK181EADBC90094D43 |            1 | type_id     | A         |          17 |     NULL | NULL   |      | BTREE      |         |               |
| item_owners |          1 | FK181EADBC499E38CA |            1 | item_id     | A         |      148000 |     NULL | NULL   |      | BTREE      |         |               |
+-------------+------------+--------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

然后我决定执行

analyze table item_owners

此输出后:

+-------------+------------+--------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table       | Non_unique | Key_name           | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+-------------+------------+--------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| item_owners |          0 | PRIMARY            |            1 | uid         | A         |     2005419 |     NULL | NULL   |      | BTREE      |         |               |
| item_owners |          1 | FK181EADBC7346EE24 |            1 | status_id   | A         |          17 |     NULL | NULL   |      | BTREE      |         |               |
| item_owners |          1 | FK181EADBC90094D43 |            1 | type_id     | A         |          17 |     NULL | NULL   |      | BTREE      |         |               |
| item_owners |          1 | FK181EADBC499E38CA |            1 | item_id     | A         |     2005419 |     NULL | NULL   |      | BTREE      |         |               |
+-------------+------------+--------------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+

任何人都知道怎么可能?,因为我真的误解了为什么基数不等于uniq行的数量,即使对于PK索引也是如此。

1 个答案:

答案 0 :(得分:1)

引用the manual

  

<强>基数

     

索引中唯一值数量的估计值。这是   通过运行ANALYZE TABLEmyisamchk -a进行更新。基数是   根据存储为整数的统计信息计算,因此值不是   即使对于小桌子也必须完全