Mysql查询优化——范围搜索
Posted
技术标签:
【中文标题】Mysql查询优化——范围搜索【英文标题】：Mysql query optimisation - range searches 【发布时间】：2013-09-12 11:45:15 【问题描述】：希望找到一些关于如何优化此查询的想法 - 多次范围搜索非常尴尬
SELECT t1.id, t1.custom_track_id, t1.deleted, t1.last_modified, t1.client_id
FROM tracks t1
INNER JOIN tracks t2 ON t1.custom_track_id = t2.custom_track_id
AND t1.last_modified > t2.last_modified
AND t1.deleted !=0
AND t2.deleted =0
AND t2.client_id
IN ( 103, 115, 116 )
WHERE t1.client_id
IN ( 103, 115, 116 )
它要查找和加入的所有字段都是 INT 字段。
索引（是的，那里有一些狡猾的）：
+--------+------------+------------------------------------------------+--------------+----------------------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+--------+------------+------------------------------------------------+--------------+----------------------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| tracks | 0 | PRIMARY | 1 | id | A | 566045 | NULL | NULL | | BTREE | | |
| tracks | 1 | client_id | 1 | client_id | A | 589 | NULL | NULL | | BTREE | | |
| tracks | 1 | featured | 1 | featured | A | 16 | NULL | NULL | | BTREE | | |
| tracks | 1 | system_status | 1 | system_status | A | 20 | NULL | NULL | | BTREE | | |
| tracks | 1 | created_by | 1 | created_by | A | 225 | NULL | NULL | | BTREE | | |
| tracks | 1 | custom_track_id | 1 | custom_track_id | A | 283022 | NULL | NULL | | BTREE | | |
| tracks | 1 | custom_track_id | 2 | custom_artist_id | A | 283022 | NULL | NULL | | BTREE | | |
| tracks | 1 | counterpoint_id | 1 | counterpoint_id | A | 113209 | NULL | NULL | YES | BTREE | | |
| tracks | 1 | system_status_2 | 1 | system_status | A | 20 | NULL | NULL | | BTREE | | |
| tracks | 1 | composition | 1 | composition | A | 13 | NULL | NULL | YES | BTREE | | |
| tracks | 1 | published_start_date | 1 | published_start_date | A | 13 | NULL | NULL | YES | BTREE | | |
| tracks | 1 | published_end_date | 1 | published_end_date | A | 13 | NULL | NULL | YES | BTREE | | |
| tracks | 1 | deleted | 1 | deleted | A | 20 | NULL | NULL | | BTREE | | |
| tracks | 1 | restricted_access_required | 1 | restricted_access_required | A | 13 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_deleted_featured_restrictedaccess | 1 | client_id | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_deleted_featured_restrictedaccess | 2 | custom_track_id | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_deleted_featured_restrictedaccess | 3 | deleted | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_deleted_featured_restrictedaccess | 4 | last_modified | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_customtrackid_deleted_lastmodified | 1 | client_id | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_customtrackid_deleted_lastmodified | 2 | custom_track_id | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_customtrackid_deleted_lastmodified | 3 | deleted | A | 336 | NULL | NULL | | BTREE | | |
| tracks | 1 | mv_clientid_customtrackid_deleted_lastmodified | 4 | last_modified | A | 336 | NULL | NULL | | BTREE | | |
+--------+------------+------------------------------------------------+--------------+----------------------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
还有解释：
---+---------+-------------------------------------------------+-------+--------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+-------+--------------------------------------------------------------------------------------------------------------------------------+-----------------------------------------------+---------+-------------------------------------------------+-------+--------------------------+
| 1 | SIMPLE | t1 | range | client_id,custom_track_id,deleted,mv_clientid_deleted_featured_restrictedaccess,mv_clientid_customtrackid_deleted_lastmodified | mv_clientid_deleted_featured_restrictedaccess | 4 | NULL | 16018 | Using where; Using index |
| 1 | SIMPLE | t2 | ref | client_id,custom_track_id,deleted,mv_clientid_deleted_featured_restrictedaccess,mv_clientid_customtrackid_deleted_lastmodified | custom_track_id | 302 | synchtank_shared_application.t1.custom_track_id | 2 | Using where |
+----+-------------+-------+-------+--------------------------------------------------------------------------------------------------------------------------------+-----------------------------------------------+---------+-------------------------------------------------+-------+--------------------------+
因此寻找优化查询或索引的方法。我也很好奇为什么它可以匹配的 2 个复合索引，它选择没有 last_modified 的那个。
为斯蒂芬解释：
+----+-------------+-------+-------+-----------------------------------------------+-----------------------------------------------+---------+------+-------+---------------------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+-------+-------+-----------------------------------------------+-----------------------------------------------+---------+------+-------+---------------------------------------------+
| 1 | SIMPLE | t1 | range | client_id | client_id | 4 | NULL | 8111 | Using where |
| 1 | SIMPLE | t2 | range | mv_clientid_deleted_featured_restrictedaccess | mv_clientid_deleted_featured_restrictedaccess | 4 | NULL | 16018 | Using where; Using index; Using join buffer |
+----+-------------+-------+-------+-----------------------------------------------+-----------------------------------------------+---------+------+-------+---------------------------------------------+
【问题讨论】：
deleted 字段包含哪些值？
0 或时间戳整数
【参考方案1】：
你可以尝试强制mysql使用索引：
SELECT
t1.id,
t1.custom_track_id,
t1.deleted,
t1.last_modified,
t1.client_id
FROM
tracks t1 FORCE INDEX(`client_id`)
INNER JOIN tracks t2 FORCE INDEX(`mv_clientid_deleted_featured_restrictedaccess`)
ON t1.custom_track_id = t2.custom_track_id
AND t1.last_modified > t2.last_modified
AND t2.deleted = 0
AND t2.client_id IN ( 103, 115, 116 )
WHERE
t1.client_id IN ( 103, 115, 116 )
AND t1.deleted !=0
通常 mysql 不使用最有效索引的原因是由于基数中的错误统计信息，这就是为什么在具有大量列和索引的大表上运行 OPTIMIZE TABLE 是一种好习惯。
【讨论】：
很高兴知道 - 谢谢。将安排在优化中。我真的不想强制它使用什么索引 @DougMcK 至少在上面运行EXPLAIN 所以看看查询计划可能会更好
解释在上面。
@DougMcK 我的意思是我发布的查询的解释
道歉 - 添加了这一点。看起来该查询比原来的查询更糟糕，现在两者都使用范围搜索以上是关于Mysql查询优化——范围搜索的主要内容，如果未能解决你的问题，请参考以下文章