我试图在查询中解释explain
的mysql(用两种不同的方式编写),这就是表:
create table text_mess(
datamess timestamp(3) DEFAULT 0,
sender bigint ,
recipient bigint ,
roger boolean,
msg char(255),
foreign key(recipient)
references users (tel)
on delete cascade
on update cascade,
primary key(datamess,sender)
)
engine = InnoDB
这是第一种类型的查询:
EXPLAIN
select /*!STRAIGHT_JOIN*/datamess, sender,recipient,roger,msg
from text_mess join (select max(datamess)as dmess
from text_mess
where roger = true
group by sender,recipient) as max
on text_mess.datamess=max.dmess ;
这是第二个:
EXPLAIN
select /*!STRAIGHT_JOIN*/datamess, sender,recipient,roger,msg
from (select max(datamess)as dmess
from text_mess
where roger = true
group by sender,recipient) as max
join
text_mess
on max.dmess = text_mess.datamess ;
两个查询要求相同的事情,唯一的区别是ref_table
(driving_table)的顺序,第一种情况是text_mess
,第二种情况是子查询。
第一个结果:
id select_type table type possible_keys key key_len ref rows Extra
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 PRIMARY text_mess ALL PRIMARY null null null 294225 null
1 PRIMARY derived2 ref auto_key0 auto_key0 7 inextremis.text_mess.datamess 10 Using index
2 DERIVED text_mess ALL recipient null null null 294225 Using where; Using temporary; Using filesort
第二个结果:
id select_type table type possible_keys key key_len ref rows Extra
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 PRIMARY <derived2> ALL null null null null 294225 Using where
1 PRIMARY text_mess ref PRIMARY PRIMARY 6 max.dmess 1 null
2 DERIVED text_mess ALL recipient null null null 294225 Using where; Using temporary; Using filesort
正如你所看到的差异是前两行的顺序,我的问题特别是在第二行(更快的查询)
第二行应该是内部表,但如果是这样,为什么列ref告诉我:max.dmess
,它应该是ref-table(子查询)的列。
最后一行是如何构建第一行的? 最后,你认为有一个更有效的查询?
答案 0 :(得分:0)
我认为答案是表扫描与主键。如果您看到,通过第一个查询,MySQL不使用任何键,而是准备好表格“text_mess”中的每一行:
id select_type table type possible_keys key key_len ref rows Extra
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 PRIMARY text_mess ALL PRIMARY null null null 294225 null
但是你在“derived2”表上使用“ON”语句,MySQL将创建一个自动键:
id select_type table type possible_keys key key_len ref rows Extra
----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 PRIMARY derived2 ref auto_key0 auto_key0 7 inextremis.text_mess.datamess 10 Using index
此密钥不存在,因此MySQL应该创建它。
如果您采用第二个示例,则在“derived2”表中进行全表扫描,并且正在使用“text_mess”中的PRIMARY KEY:
id select_type table type possible_keys key key_len ref rows Extra
------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
1 PRIMARY <derived2> ALL null null null null 294225 Using where
1 PRIMARY text_mess ref PRIMARY PRIMARY 6 max.dmess 1 null
答案是,在这种情况下,MySQL决定创建并使用索引,但通常它会进行全表扫描,而且速度更快。
想象一下, auto_key0 键再次包含此子查询中的唯一一列。这个额外的过程不是必需的。这就是为什么你的第二个查询更快。