mysql查询速度非常慢：左联接多个表，每个表的where子句
[英]Very slow mysql query: left join multiple table and where clause for each table
问题描述
我有一个基本表（id表）和多个表（数据表），这些表具有与id相关的数据，每个表具有不同的类型数据。 我想从这些不同的表中按多种条件过滤ID。
我创建查询如下，但这太慢了，大约3分钟。 我检查了解释信息，但仍然找不到更好的方法。 有人能找到更快的方法吗？
询问
select jj.id, jj.imgtitle, jj.alias
from jjtable jj
inner join jjtable_catg jjc on jj.catid = jjc.cid
left join jjtable_map jjtm_name on jj.id = jjtm_name.picid
left join jjtable_tags jjts_name on jjtm_name.tid = jjts_name.tid
left join jjtable_map_ge jjtm_ge on jj.id = jjtm_ge.picid
left join jjtable_tags jjts_ge on jjtm_ge.tid = jjts_ge.tid
left join jjtable_map_ban jjtm_ban on jj.id = jjtm_ban.picid
left join jjtable_tags jjts_ban on jjtm_ban.tid = jjts_ban.tid
left join jjtable_map_per jjtm_per on jj.id = jjtm_per.picid
left join jjtable_tags jjts_per on jjtm_per.tid = jjts_per.tid
left join jjtable_map_fea jjtm_fea on jj.id = jjtm_fea.picid
left join jjtable_tags jjts_fea on jjtm_fea.tid = jjts_fea.tid
left join jjtable_map_ev jjtm_ev on jj.id = jjtm_ev.picid
left join jjtable_tags jjts_ev on jjtm_ev.tid = jjts_ev.tid
left join jjtable_map_ag jjtm_ag on jj.id = jjtm_ag.picid
left join jjtable_tags jjts_ag on jjtm_ag.tid = jjts_ag.tid
left join jjtable_map_fa jjtm_fa on jj.id = jjtm_fa.picid
left join jjtable_tags jjts_fa on jjtm_fa.tid = jjts_fa.tid
left join jjtable_map_im jjtm_im on jj.id = jjtm_im.picid
left join jjtable_tags jjts_im on jjtm_im.tid = jjts_im.tid where jj.published = 1
and jj.approved = 1
and jjts_fea.tid in(87,90)
and jjts_fea.delete_flag = 0
and jjtm_fea.delete_flag = 0
and jjc.cid in(4,10)
and jjts_name.tid in(77)
and jjts_name.delete_flag = 0
and jjtm_name.delete_flag = 0
and jjts_per.tid in(28,36)
and jjts_per.delete_flag = 0
and jjtm_per.delete_flag = 0
and jjts_ag.tid in(98,99)
and jjts_ag.delete_flag = 0
and jjtm_ag.delete_flag = 0
and jjts_fa.tid in(104,107)
and jjts_fa.delete_flag = 0
and jjtm_fa.delete_flag = 0
group by jj.id
order by case when date(jj.date) > date_add(date(now()), interval -14 day) then jj.view end DESC, jj.id DESC
编辑：谢谢您的评论。 首先，我添加信息作为您的请求。
为什么需要此查询；
该查询旨在从“表单输入”标签接收过帐数据，用户可以从不同类别中选择选项来过滤ID。 这就是为什么此查询包含此示例中未使用的表和数据的原因。
左联接之所以这么多，是因为每个“映射”表都有类别映射信息以过滤ID。 等等，用户希望找到与“ fea”，87、90和“ ag”，98、99映射的ID。
解释选择
+----+-------------+--------------+--------+-------------------+-----------------+---------+------------------------------------------------+------+---------------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+--------------+--------+-------------------+-----------------+---------+------------------------------------------------+------+---------------------------------+
| 1 | SIMPLE | jjts_name | const | PRIMARY | PRIMARY | 4 | const | 1 | Using temporary; Using filesort |
| 1 | SIMPLE | jjc | range | PRIMARY | PRIMARY | 4 | NULL | 2 | Using where |
| 1 | SIMPLE | jjts_pe | range | PRIMARY | PRIMARY | 4 | NULL | 2 | Using where; Using join buffer |
| 1 | SIMPLE | jjts_fea | range | PRIMARY | PRIMARY | 4 | NULL | 2 | Using where; Using join buffer |
| 1 | SIMPLE | jjts_ag | range | PRIMARY | PRIMARY | 4 | NULL | 2 | Using where; Using join buffer |
| 1 | SIMPLE | jjts_fa | range | PRIMARY | PRIMARY | 4 | NULL | 2 | Using where; Using join buffer |
| 1 | SIMPLE | jj | ref | PRIMARY,idx_catid | idx_catid | 4 | jjc.cid | 95 | Using where |
| 1 | SIMPLE | jjtm_ag | eq_ref | udx_picid_tid | udx_picid_tid | 8 | jj.id,jjts_ag.tid | 1 | Using where |
| 1 | SIMPLE | jjtm_fa | eq_ref | udx_picid_tid | udx_picid_tid | 8 | jj.id,jjts_fa.tid | 1 | Using where |
| 1 | SIMPLE | jjtm_fea | eq_ref | udx_picid_tid | udx_picid_tid | 8 | jj.id,jjts_fea.tid | 1 | Using where |
| 1 | SIMPLE | jjtm_pe | eq_ref | udx_picid_tid | udx_picid_tid | 8 | jjtm_fea.picid,jjts_per.tid | 1 | Using where |
| 1 | SIMPLE | jjtm_ev | ref | udx_picid_tid | udx_picid_tid | 4 | jjtm_fa.picid | 3 | Using index |
| 1 | SIMPLE | jjts_ev | eq_ref | PRIMARY | PRIMARY | 4 | jjtm_ev.tid | 1 | Using index |
| 1 | SIMPLE | jjtm_im | ref | udx_picid_tid | udx_picid_tid | 4 | jjtm_pe.picid | 6 | Using index |
| 1 | SIMPLE | jjts_im | eq_ref | PRIMARY | PRIMARY | 4 | jjtm_im.tid | 1 | Using index |
| 1 | SIMPLE | jjtm_ge | ref | udx_picid_tid | udx_picid_tid | 4 | jj.id | 23 | Using index |
| 1 | SIMPLE | jjts_ge | eq_ref | PRIMARY | PRIMARY | 4 | jjtm_ge.tid | 1 | Using index |
| 1 | SIMPLE | jjtm_ban | ref | udx_picid_tid | udx_picid_tid | 4 | jjtm_pe.picid | 9 | Using index |
| 1 | SIMPLE | jjts_ban | eq_ref | PRIMARY | PRIMARY | 4 | jjtm_ban.tid | 1 | Using index |
| 1 | SIMPLE | jjtm_name | eq_ref | udx_picid_tid | udx_picid_tid | 8 | jjtm_fea.picid,const | 1 | Using where |
+----+-------------+--------------+--------+-------------------+-----------------+---------+------------------------------------------------+------+---------------------------------+
20 rows in set (2 min 15.87 sec)
显示列
mysql> show columns from jjtables;
+--------------+------------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+--------------+------------------+------+-----+---------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| catid | int(11) | NO | MUL | 0 | |
| imgtitle | text | NO | | NULL | |
| alias | varchar(255) | NO | | | |
| date | datetime | NO | | NULL | |
| view | int(11) | NO | | 0 | |
| published | tinyint(1) | NO | | 0 | |
| approved | tinyint(1) | NO | | 0 | |
+--------------+------------------+------+-----+---------+----------------+
mysql> show columns from jjtable_catg;
+--------------+---------------------+------+-----+---------+----------------+
| Field | Type | Null | Key | Default | Extra |
+--------------+---------------------+------+-----+---------+----------------+
| cid | int(11) | NO | PRI | NULL | auto_increment |
| name | varchar(255) | NO | | | |
| alias | varchar(255) | NO | | | |
| parent | int(11) | NO | MUL | 0 | |
| desc | text | YES | | NULL | |
| order | int(11) | NO | | 0 | |
+--------------+---------------------+------+-----+---------+----------------+
mysql> show columns from jjtable_tags;
+--------------+-------------+------+-----+---------------------+----------------+
| Field | Type | Null | Key | Default | Extra |
+--------------+-------------+------+-----+---------------------+----------------+
| tid | int(11) | NO | PRI | NULL | auto_increment |
| name | varchar(60) | NO | UNI | NULL | |
| talias | varchar(60) | NO | | | |
| ttypeid | int(11) | NO | | 1 | |
| pa_tid | int(11) | NO | | 0 | |
| delete_flag | tinyint(1) | NO | | 0 | |
| created | datetime | NO | | 0000-00-00 00:00:00 | |
| created_by | int(11) | NO | | 0 | |
| modified | datetime | NO | | 0000-00-00 00:00:00 | |
| modified_by | int(11) | NO | | 0 | |
| t_due | datetime | NO | | 0000-00-00 00:00:00 | |
+--------------+-------------+------+-----+---------------------+----------------+
mysql> show columns from jjtable_map;
+-------------+------------+------+-----+---------------------+----------------+
| Field | Type | Null | Key | Default | Extra |
+-------------+------------+------+-----+---------------------+----------------+
| id | int(11) | NO | PRI | NULL | auto_increment |
| picid | int(11) | NO | MUL | NULL | |
| tid | int(11) | NO | | NULL | |
| delete_flag | tinyint(1) | NO | | 0 | |
| created | datetime | NO | | 0000-00-00 00:00:00 | |
| created_by | int(11) | NO | | 0 | |
| modified | datetime | NO | | 0000-00-00 00:00:00 | |
| modified_by | int(11) | NO | | 0 | |
+-------------+------------+------+-----+---------------------+----------------+
对不起，我的解释不好。 希望此编辑有助于您更好地理解。 编辑结束
3 个解决方案
解决方案1
3 2014-06-25 15:15:59
一个大的改进是在联接上使用多个条件，而不是获得一个巨大的集合，然后将条件放在最后。 我的意思是，如果您同时添加条件，则加入集合的条件会更小，响应时间会更好：
select jj.id, jj.imgtitle, jj.alias
from jjtable jj
inner join jjtable_catg jjc on jj.catid = jjc.cid
left join jjtable_map jjtm_name on jj.id = jjtm_name.picid and jjtm_name.delete_flag = 0
...
解决方案2
3 已采纳 2014-06-25 15:28:08
尝试以下操作：将您的左联接转换为一堆WHERE EXISTS子查询，然后删除GROUP BY ，因为（这是我的怀疑），这正是您真正想要表达的。
select
jj.id, jj.imgtitle, jj.alias
from
jjtable jj
-- you could drop this join, you don't do anything with jjtable_catg
inner join jjtable_catg jjc on jjc.cid = jj.catid
where
jj.published = 1
and jj.approved = 1
and jjc.cid in(4, 10)
and exists (
select 1
from jjtable_map m inner join jjtable_tags t on m.tid = t.tid
where m.picid = jj.id m.delete_flag = 0 and t.tid in (77) and t.delete_flag = 0
)
and exists (
select 1
from jjtable_map_per m inner join jjtable_tags t on m.tid = t.tid
where m.picid = jj.id m.delete_flag = 0 and t.tid in (28,36) and t.delete_flag = 0
)
and exists (
select 1
from jjtable_map_fea m inner join jjtable_tags t on m.tid = t.tid
where m.picid = jj.id m.delete_flag = 0 and t.tid in (87,90) and t.delete_flag = 0
)
and exists (
select 1
from jjtable_map_ag m inner join jjtable_tags t on m.tid = t.tid
where m.picid = jj.id m.delete_flag = 0 and t.tid in (98,99) and t.delete_flag = 0
)
and exists (
select 1
from jjtable_map_fa m inner join jjtable_tags t on m.tid = t.tid
where m.picid = jj.id m.delete_flag = 0 and t.tid in (104,107) and t.delete_flag = 0
)
order by
case when date(jj.date) > date_add(date(now()), interval -14 day) then jj.view end DESC,
jj.id DESC
同时创建这些复合索引（如果缺少）：
-
jjtable_catg：(cid) -
jjtable_tags：(tid, delete_flag) -
jjtable_map和所有jjtable_map_*：（(picid, delete_flag, tid)
解决方案3
0 2014-06-25 15:05:05
该查询已经有很多表。 删除任何未使用的表。 有什么理由加入吗？
left join jjtable_map_im jjtm_im on jj.id = jjtm_im.picid
left join jjtable_tags jjts_im on jjtm_im.tid = jjts_im.tid where jj.published = 1
完全不使用jjtm_im和jjts_im。
mysql左边加入一个非常大的表 - 超级慢
声明:本站的技术帖子网页，遵循CC BY-SA 4.0协议，如果您需要转载，请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.