Question

我正在寻求帮助来优化这个mysql查询。需要超长时间才能运行，因为main_activity下的两个表都很大（每个超过1000万行！）。 main_db.members和main_db.customers分别约为400K和600K行。

编辑：

根据使用临时表的建议，只想添加我在只读数据库上运行查询，因此临时表可能是个问题。不使用临时表我可以做什么优化？

select distinct
    a.members_id,
    a.customer_id,
    a.subscription,
    a.buy_date,
    from_unixtime((max(m2.sales_date) / 1000), '%m/%d/%Y')  as sales_date,
    a.return_date,
    a.signup_date,
    from_unixtime((max(st.visit_date) / 1000), '%m/%d/%Y') as visit_date
from (select distinct
          m1.members_id,
          m1.customer_id,
          m1.subscription,
          from_unixtime((m1.buy_date / 1000), '%m/%d/%Y') as buy_date,
          from_unixtime((m1.return_date / 1000), '%m/%d/%Y')   as return_date,
          from_unixtime((c.signup_date / 1000), '%m/%d/%Y')   as signup_date
      from main_db.members m1
          join main_db.customer c on c.global_members_id = m1.members_id
     ) as a
    left join main_db.members m2 on m2.customer_id = a.customer_id
    left join main_activity.onlinevisit s on s.customer_id = a.customer_id
    left join main_activity.storevisit st on st.visit_id = s.visit_id

Answer 1

我们的想法是创建一个带有好密钥的临时表。我们可以从这开始：

create temporary table a (key(customer_id)) select distinct
          m1.members_id,
          m1.customer_id,
          m1.subscription,
          from_unixtime((m1.buy_date / 1000), '%m/%d/%Y') as buy_date,
          from_unixtime((m1.return_date / 1000), '%m/%d/%Y')   as return_date,
          from_unixtime((c.signup_date / 1000), '%m/%d/%Y')   as signup_date
      from main_db.members m1
          join main_db.customer c on c.global_members_id = m1.members_id;


select distinct
    a.members_id,
    a.customer_id,
    a.subscription,
    a.buy_date,
    from_unixtime((max(m2.sales_date) / 1000), '%m/%d/%Y')  as sales_date,
    a.return_date,
    a.signup_date,
    from_unixtime((max(st.visit_date) / 1000), '%m/%d/%Y') as visit_date
from a
    left join main_db.members m2 on m2.customer_id = a.customer_id
    left join main_activity.onlinevisit s on s.customer_id = a.customer_id
    left join main_activity.storevisit st on st.visit_id = s.visit_id;

您还需要确保在其他表中也有好键。

Answer 2

请提供SHOW CREATE TABLE。

我希望

上有索引

m2.customer_id
s.customer_id
st.visit_id

如果没有，那可能是一个重大的性能问题。

使用DISTINCT意味着JOINs乘以行数，您需要对其进行缩小。没有DISTINCT，每个查询都能正常工作吗？消除它会节省数据传递。

避免通货膨胀通缩开销的另一种可能性是取代

max(m2.sales_date)

与

( SELECT max(m2.sales_date)
      FROM main_db.members m2
      WHERE m2.customer_id = a.customer_id )

（等）

想要优化mysql查询

2 个答案: