SQL子查询分析

在这里看到了一个帖子 http://www.iteye.com/topic/1122917，很简单，凑下热闹，因为Oracle，SQLServer的子查询和Join算法的选择很多，但是MySQL方面就没有那么多的选择了，所以很多子查询的性能都很差，这里分析几个帖子里MySQL的答案。

需求：使用SQL语句查出哪些人即会玩basketball又会玩badminton，找出这样的name-hobby组合

create table test (NAME varchar(20) not null, HOBBY varchar(20) not null);
insert into  test values('Adam','basketball');
insert into  test values('Bill','basketball');
insert into  test values('Bill','football');
insert into  test values('Cyper','basketball');
insert into  test values('Cyper','badminton');
insert into  test values('David','basketball');
insert into  test values('David','badminton');
insert into  test values('David','table tennis');

一个Groupby的写法

select name,group_concat(hobby),count(*) 
from test where HOBBY IN ('basketball','badminton')
group by name having count(*)]]]]>1;

一个exists的写法

select * from test a

WHERE hobby = 'basketball'

AND EXISTS (SELECT * from TEST b

WHERE a.name = b.name AND b.hobby = 'badminton');

但是有重复的情况如何处理？

insert into test values('Bill','basketball');
insert into test values('Bill','basketball');
insert into test values('David','badminton');
insert into test values('David','badminton');

Groupby的写法需要进化使用distinct

select name,group_concat(distinct(hobby)), count(distinct(hobby))

from test where hobby in ('basketball','badminton') group by name having count(distinct(hobby))>1

exists的写法的写法不用修改，但是执行计划是DEPENDENT SUBQUERY，这样效率很差，那么要优化成Derived Table写法，如下

select distinct(a.name) as name,concat(a.hobby, b.hobby) from test a,(SELECT distinct(name) as name , hobby from TEST WHERE hobby = 'badminton') as b

WHERE a.hobby = 'basketball'

AND a.name = b.name

--EOF--