Question

我的查询遇到一些困难，其中包含多个左连接以及按组和顺序组合。

text table和textdetails包含+ - 800k记录
复制表和copydetails包含+ - 200k记录
其他表格要小得多。

我为每个执行左连接的列都有外键。我还在每个列上都有索引，我在其中执行where语句。下面的MySQL查询仍然运行大约40秒。离开Group By会有所改善。抛弃Order By可以改善很多。

我做了一些研究，但我仍然对如何改进查询或索引感到困惑。

SELECT * FROM `copy` 
LEFT JOIN `domain` ON domain.domain_id = copy.copy_domain_id 
LEFT JOIN `domaincategory` ON copy.copy_domain_id = domaincategory.domaincategory_domain_id AND domaincategory.domaincategory_account_id = copy.copy_account_id 
LEFT JOIN `text` ON text.text_id = copy.copy_text_id LEFT JOIN `textdetails` ON textdetails.textdetails_text_id = text.text_id 
LEFT JOIN `channel` ON channel.channel_domain_id = domain.domain_id AND channel.channel_account_id = copy.copy_account_id 
LEFT JOIN `feed` ON feed.feed_id = text.text_feed_id 
WHERE (feed.feed_account_id = 96) AND (feed.feed_flag_delete IS NULL) AND (text.text_flag_delete IS NULL) AND (copy.copy_flag_delete IS NULL) AND (copy.copy_tracking_date_found IS NOT NULL) AND (channel.channel_active = 1) 
GROUP BY `copy`.`copy_id`
ORDER BY `copy`.`copy_tracking_date_found` DESC LIMIT 50

EXPLAIN选项的结果显示如下，但我无法弄清楚如何阅读并正确使用它

ID  : 1
Select_type : SIMPLE
Table : Feed
Type : Ref
Possible_Keys: PRIMARY,fk_feed_account_id,feed_flag_delete
Key: fk_feed_account_id
Key_len : 4:
Ref : const
Rows : 1
Extra: Using where; Using temporary; Using filesort


ID  : 1
Select_type : SIMPLE
Table : text
Type : Ref
Possible_Keys: PRIMARY,fk_text_feed_id,text_flag_delete
Key: text_flag_delete
Key_len : 2
Ref : const
Rows : 2628
Extra: Using where


ID  : 1
Select_type : SIMPLE
Table : textdetails
Type : Ref
Possible_Keys: fk_textdetails_text_id
Key: fk_textdetails_text_id
Key_len : 5
Ref : text.text_id
Rows : 1
Extra:


ID  : 1
Select_type : SIMPLE
Table : copy
Type : Ref
Possible_Keys: fk_copy_account_id,fk_copy_domain_id,fk_copy_text_...
Key: fk_copy_text_id
Key_len : 4
Ref : text.text_id
Rows : 1
Extra: Using where


ID  : 1
Select_type : SIMPLE
Table : domain
Type : eq_ref
Possible_Keys: PRIMARY
Key: PRIMARY
Key_len : 4
Ref : copy.copy_domain_id
Rows : 1
Extra: Using where


ID  : 1
Select_type : SIMPLE
Table : domaincategory
Type : eq_ref
Possible_Keys: fk_domaincategory_account_id,fk_domaincategory_dom
Key: fk_domaincategory_domain_id
Key_len : 4
Ref : domain.domain_id
Rows : 1
Extra:


ID  : 1
Select_type : SIMPLE
Table : channel
Type : ref
Possible_Keys: fk_channel_account_id,fk_channel_domain_id,channel...
Key: fk_channel_domain_id
Key_len : 4
Ref : copy.copy_domain_id
Rows : 2
Extra: Using where

或许我应该多解释一下这种关系？     feed：text = 1：n
    text：textdetails = 1：1
    text：copy = 1：n
    copy：domain = n：1
    channel：domain n：1

Answer 1

我会改变一些事情并更新查询以反映这一点。另外，对于索引。如果每列都有索引，但它们是单独的索引，那么WONT必然会对您有所帮助。如果可能，您需要复合（多个字段）索引以更好地匹配您的连接/标准和分组。

SELECT 
      * 
   FROM 
      feed f 
         JOIN text t
            ON feed.feed_id = t.text_feed_id
            LEFT JOIN textdetails td 
               ON t.text_id = td.textdetails_text_id 
            JOIN COPY c
               ON t.text_id = c.copy_text_id 
               LEFT JOIN domain d
                  ON c.copy_domain_id = d.domain_id
               LEFT JOIN domaincategory dc 
                  ON  c.copy_domain_id = dc.domaincategory_domain_id 
                  AND c.copy_account_id = cd.domaincategory_account_id
               JOIN channel ch 
                  ON  c.copy_account_id = ch.channel_account_id 
                  AND c.copy_domain_id = ch.channel_domain_id
                  AND ch.channel_active = 1
   WHERE 
          f.feed_account_id = 96
      AND f.feed_flag_delete IS NULL
      AND t.text_flag_delete IS NULL
      AND c.copy_flag_delete IS NULL
      AND c.copy_tracking_date_found IS NOT NULL
   GROUP BY 
      c.copy_id
   ORDER BY 
      c.copy_tracking_date_found DESC 
   LIMIT 
      50

Transitive Property
由于copy.copy_domain_id是domain.domain_id的连接，channel.channel_domain_id连接到domain.domain_id，我们只需更改为copy.copy_domain_id = channel.channel_domain_id，而不需要拆分连接到其他相同值的表。< / p>

第二个......你有LEFT JOINS，但是当你添加＆＃34; feed_account_id = 96＆＃34;时，你会自动将它转换为INNER JOIN，因为它是一个要求，因此使TEXT别名成为一个连接好。与具有Channel_Active = 1的CHANNEL表类似（我有更新查询以反映该情况）。

现在，由于根据特定帐户的FEED表进行限定，我已将其移至第一个FROM位置并连接到Copy表。

现在，索引有助于优化此

table        index
feed         ( feed_account_id, feed_id, feed_flag_delete )
text         ( text_feed_id, text_id, text_flag_delete )
textdetails  ( textdetails_text_id )
copy         ( copy_text_id, copy_domain_id, copy_account_id, copy_flag_delete, copy_tracking_date_found, copy_id )
domain       ( domain_id )
domaincategory ( domaincategory_domain_id, domaincategory_account_id )
channel       ( channel_domain_id, channel_account_id, channel_active )

具有多个左连接，分组依次和顺序的mysql优化

1 个答案: