多列索引实验

多列索引经典题目:
http://www.zixue.it/thread-9218-1-4.html


假设某个表有一个联合索引（c1,c2,c3,c4）一下——只能使用该联合索引的c1,c2,c3部分
A where c1=x and c2=x and c4>x and c3=x 
B where c1=x and c2=x and c4=x order by c3
C where c1=x and c4= x group by c3,c2
D where c1=x and c5=x order by c2,c3

E where c1=x and c2=x and c5=? order by c2,c3

<span style="font-size:18px;">create table t4 (
c1 tinyint(1) not null default 0,
c2 tinyint(1) not null default 0,
c3 tinyint(1) not null default 0,
c4 tinyint(1) not null default 0,
c5 tinyint(1) not null default 0,index c1234(c1,c2,c3,c4)
);
insert into t4 values (1,3,5,6,7),(2,3,9,8,3),(4,3,2,7,5);</span>

对于A:
c1=x and c2=x and c4>x and c3=x  <==等价==> c1=x and c2=x and c3=x and c4>x
因此 c1,c2,c3,c4都能用上. 如下:

<span style="font-size:18px;">explain select * from t4 where c1=1 and c2=2 and c4>3 and c3=3 \G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: range
possible_keys: c1234
          key: c1234
      key_len: 4 #可以看出c1,c2,c3,c4索引都用上
          ref: NULL
         rows: 1
        Extra: Using where </span>

对于B: select * from t4 where c1=1 and c2=2 and c4=3 order by c3
c1 ,c2索引用上了,在c2用到索引的基础上,c3是排好序的,因此不用额外排序.
而c4没发挥作用.

注：因为c3是排好序的，就好像搭木板过河一样，c3已经排好序，就好比如搭了c1和c2的木板，但是c3已经默认有的了（其实是没有的，就好比如人家默认是做好的了，不提供你材料），所以c4就没有办法搭建了

<span style="font-size:18px;">explain select * from t4 where c1=1 and c2=2 and c4=3 order by c3 \G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: ref
possible_keys: c1234
          key: c1234
      key_len: 2
          ref: const,const
         rows: 1
        Extra: Using where
1 row in set (0.00 sec)

mysql> explain select * from t4 where c1=1 and c2=2 and c4=3 order by c5 \G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: ref
possible_keys: c1234
          key: c1234
      key_len: 2
          ref: const,const
         rows: 1
        Extra: Using where; Using filesort
1 row in set (0.00 sec)</span>

对于 C: 只用到c1索引,因为group by c3,c2的顺序无法利用c2,c3索引

注：group by 的原理：先对分组字段进行排序（order by)，然后再进行分组操作，即group by 等价于order by c2, c3

<span style="font-size:18px;">explain select * from t4 where c1=1 and c4=2 group by c3,c2 \G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: ref
possible_keys: c1234
          key: c1234
      key_len: 1 #只用到c1,因为先用c3后用c2分组,导致c2,c3索引没发挥作用
          ref: const
         rows: 1
        Extra: Using where; Using temporary; Using filesort
1 row in set (0.00 sec)</span>

<span style="font-size:18px;">explain select * from t4 where c1=1 and c4=2 group by c2,c3 \G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: ref
possible_keys: c1234
          key: c1234
      key_len: 1
          ref: const
         rows: 1
        Extra: Using where
1 row in set (0.00 sec)</span>

D语句: C1确定的基础上,c2是有序的,C2之下C3是有序的,因此c2,c3发挥的排序的作用.
因此,没用到filesort

<span style="font-size:18px;">explain select * from t4 where c1=1 and c5=2 order by c2,c3 \G  
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: ref
possible_keys: c1234
          key: c1234
      key_len: 1
          ref: const
         rows: 1
        Extra: Using where
1 row in set (0.00 sec)</span>

E: 这一句等价与 elect * from t4 where c1=1 and c2=3 and c5=2 order by c2, c3; 
因为c2的值既是固定的,参与排序时并不考虑

注：因为c1和c2是固定值，那么order by c2(一个常量), c3，这里c2已经是一个常量了，所以不需要做排序

即order by c2, c3 等价于order by c3

<span style="font-size:18px;">explain select * from t4 where c1=1 and c2=3 and c5=2 order by c2,c3 \G
*************************** 1. row ***************************
           id: 1
  select_type: SIMPLE
        table: t4
         type: ref
possible_keys: c1234
          key: c1234
      key_len: 2
          ref: const,const
         rows: 1
        Extra: Using where
1 row in set (0.00 sec)</span>