Question

我继承了一个客户端的应用程序，并且一直在处理以下查询，该查询需要187秒并搜索304M表行以发回444个结果行。

我是否有权想要摆脱子选择并用连接替换它们？我找不到正确的方法。任何优化此查询的帮助都将非常受欢迎。感谢...

SELECT Business.name, Business.primary_city, Count(click.id) as clicks, Count(DISTINCT email_leads.parent_message_id) as tot_email_leads, Count(bc.business_id) as county_no, Count(br.business_id) as region_no, count(reveals.id) as reveals_no

FROM businesses as Business

LEFT JOIN business_clickthroughs as click ON ( Business.id = click.business_id  AND (click.created BETWEEN '2014-04-01 00:00:00' and '2014-04-30 23:59:59'))
LEFT JOIN users as U ON Business.id = U.business_id
LEFT JOIN messages as email_leads ON (U.id = email_leads.from_to AND (email_leads.parent_message_id is null OR email_leads.parent_message_id = email_leads.id )  AND (email_leads.created BETWEEN '2014-04-01 00:00:00' and '2014-04-30 23:59:59'))
LEFT JOIN business_counties as bc ON Business.id = bc.business_id                   
LEFT JOIN businesses_business_types as bt ON Business.id = bt.business_id 
LEFT JOIN business_reveals as reveals ON (reveals.business_id = Business.id  AND (reveals.created BETWEEN '2014-04-01 00:00:00' and '2014-04-30 23:59:59'))
LEFT JOIN business_regions as br ON Business.id = br.business_id 

WHERE 1=1 

Group By Business.id;

Answer 1

从您正在编写的查询来看，子选择可能是正确的方法。您没有显示实际的原始查询，因此这只是猜测。

您的查询正在加入数据，这些数据很容易就是单独的维度 - 用户，消息，县，“揭示”（无论是什么）。然后，连接将为每个业务生成这些维度的笛卡尔积，进一步放大数据并减慢查询速度。

而且，如果你的较大的表有3亿多行，那么300秒来总结 all 8个左右表中的数据似乎并不合理。您的查询没有任何过滤数据的where条件。 join s都是left join s，这会减少过滤。

如果性能问题，请询问其他问题。包括实际查询，查询的explain计划以及表的布局（尤其是索引结构）。

优化包含联接和子查询的查询

1 个答案: