1我們先回顧一下使用複合索引的幾個條件

點選(此處)摺疊或開啟

如果不是使用索引的最左列，則無法使用索引
不能跳過索引中的列，如select * from tab where a=** and c=***，則只能用到索引的第一列a
如果查詢中的莫個列有範圍查詢，則右邊的列都無法使用索引，如：select * from tab where a= and b> and c= 則索引只能用到a,b列

使用索引排序最好是滿足一下兩個條件

1.只有當索引的列順序與order by 子句的順序完全一致，並且所有列的排序方向（倒敘或者正序）都一樣時，mysql才能夠使用索引來對結果進行排序。
2.如果查詢需要關聯多張表的時候，則只有當order by子句引用的欄位全部為第一張表時，才能使用索引排序，order by子句和查詢型查詢的限制是一樣的：需要滿足索引的最左字首的要求，否則無法利用索引排序，有一種情況order by 子句可以不滿足索引的最左字首的要求，就是前導列為常量的時候，
修改：滿足的索引必須是執行計劃中的索引啊，，，

下面進行一些例子

背景：複合索引（a,b,c）

1.select * from tab where a=‘9878’ order by b； ---可以，因為最左字首提供了常量，ab組合就形成了最左字首，explain檢視只使用了複合索引的（a)列，但是排序時用到了b列

2.select * from tab where a> order by a,b; ---可以，（a,b）滿足最左字首，，，explain檢視只使用了複合索引的（a)列，但是排序時用到了b列

下面這些就不行了
select * from tab where a='asad' order by b desc,c asc; ---排序不一樣

select * from tab where a='adad' order by b,d; ----引用非索引欄位

select * from tab where a='adsad' order by c; ----（a,c)組合跳過b列

select * from tab where a> order b,c; -----a是範圍查詢，索引只能用到a列

select * from tab where a= and b in () order by c; -----b列範圍查詢，其右邊的欄位用不到了，只能用到 ab列

select * from tab where

一個面試題：

A：（c1,c2,c3,c4)
B:(c1,c2),排序用到c3
C:(c1)
D:(c1) 排序用到c2 c3
E:(c1,c2) 排序用到c3
題中的？如果代表的是非常量的意思，那麼我們就選B吧！

----這個理論上是可以使用索引進行關聯排序的，滿足第二個條件嘛：a既滿足來自第一張表，也滿足最左字首，但是最佳化器選擇時將table1表當作了第二張關聯表，所以實際上無法使用索引了！
select * from table1，table2 where table1.id=table2.id order by a;

下面舉一個生產上的例子

點選(此處)摺疊或開啟

mysql> explain SELECT payment.id,payment.money,payment.c_time,payment.card_type,b_member_card.active_time,b_users.name,b_users.tel,zone.zone_name,admin.name as admin_name,admin.role_id,b_card_info.card_name as ccname FROM `payment` LEFT JOIN b_member_card ON b_member_card.payid=payment.id LEFT JOIN b_users ON payment.xdy_id=b_users.uid LEFT JOIN zone ON payment.zone_id=zone.zone_id LEFT JOIN admin ON payment.sale_id=admin.id LEFT JOIN b_card_info ON payment.vip_card = b_card_info.id WHERE ( `sale_id` IN ('39','110','159','311','349','353','362','432','449','450','489','501','533','562') ) AND ( `payed` = 2 ) ORDER BY payment.id desc LIMIT 0,10;
+----+-------------+---------------+--------+---------------+---------------+---------+---------------------------------------+------+-----------------------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+---------------+--------+---------------+---------------+---------+---------------------------------------+------+-----------------------------+
| 1 | SIMPLE | payment | range | sale_id_payed | sale_id_payed | 5 | NULL | 40 | Using where; Using filesort | ---為什麼會出現filesort
| 1 | SIMPLE | b_member_card | ref | payid | payid | 5 | interface_hd_com0624.payment.id | 1 | |
| 1 | SIMPLE | b_users | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.xdy_id | 1 | |
| 1 | SIMPLE | zone | eq_ref | PRIMARY | PRIMARY | 3 | interface_hd_com0624.payment.zone_id | 1 | |
| 1 | SIMPLE | admin | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.sale_id | 1 | |
| 1 | SIMPLE | b_card_info | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.vip_card | 1 | |
+----+-------------+---------------+--------+---------------+---------------+---------+---------------------------------------+------+-----------------------------+
6 rows in set (0.00 sec)
。
檢視索引
mysql> show index from payment;
+---------+------------+---------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| Table | Non_unique | Key_name | Seq_in_index | Column_name | Collation | Cardinality | Sub_part | Packed | Null | Index_type | Comment | Index_comment |
+---------+------------+---------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
| payment | 0 | PRIMARY | 1 | id | A | 3173 | NULL | NULL | | BTREE | | |
| payment | 1 | sale_id_payed | 1 | sale_id | A | 137 | NULL | NULL | | BTREE | | |
| payment | 1 | sale_id_payed | 2 | payed | A | 244 | NULL | NULL | | BTREE | | |
+---------+------------+---------------+--------------+-------------+-----------+-------------+----------+--------+------+------------+---------+---------------+
3 rows in set (0.04 sec)
為什麼上面會出現filesort這種情況呢？id來之於執行計劃的第一張表，又是主鍵，完全滿足上面索引排序的要求啊，但是為啥沒用到索引排序呢？

我們更改一下sql語句

mysql> explain SELECT payment.id,payment.money,payment.c_time,payment.card_type,b_member_card.active_time,b_users.name,b_users.tel,zone.zone_name,admin.name as admin_name,admin.role_id,b_card_info.card_name as ccname FROM `payment` LEFT JOIN b_member_card ON b_member_card.payid=payment.id LEFT JOIN b_users ON payment.xdy_id=b_users.uid LEFT JOIN zone ON payment.zone_id=zone.zone_id LEFT JOIN admin ON payment.sale_id=admin.id LEFT JOIN b_card_info ON payment.vip_card = b_card_info.id WHERE ( `sale_id` IN ('39','110','159','311','349','353','362','432','449','450','489','501','533','562') ) AND ( `payed` = 2 ) ORDER BY payment.payed desc LIMIT 0,10
+----+-------------+---------------+--------+---------------+---------------+---------+---------------------------------------+------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+---------------+--------+---------------+---------------+---------+---------------------------------------+------+-------------+
| 1 | SIMPLE | payment | range | sale_id_payed | sale_id_payed | 5 | NULL | 40 | Using where | --排序條件改成payed就好了
| 1 | SIMPLE | b_member_card | ref | payid | payid | 5 | interface_hd_com0624.payment.id | 1 | |
| 1 | SIMPLE | b_users | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.xdy_id | 1 | |
| 1 | SIMPLE | zone | eq_ref | PRIMARY | PRIMARY | 3 | interface_hd_com0624.payment.zone_id | 1 | |
| 1 | SIMPLE | admin | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.sale_id | 1 | |
| 1 | SIMPLE | b_card_info | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.vip_card | 1 | |
+----+-------------+---------------+--------+---------------+---------------+---------+---------------------------------------+-----

觀察上面兩條sql你可以發現：要想使用索引排序你得滿足下面的條件
：在滿足最上面兩個條件的的基礎上
：附加：order by col的col還必須是滿足執行計劃中的用到的索引的最左字首條件

根據這些要求我們可以將sale_id_payed索引的範圍擴充一點至id 刪除原來的索引新建：create index sale_id_payed_id on payment(id,sale_id,payed);

mysql> explain SELECT payment.id,payment.money,payment.c_time,payment.card_type,b_member_card.active_time,b_users.name,b_users.tel,zone.zone_name,admin.name as admin_name,admin.role_id,b_card_info.card_name as ccname FROM `payment` LEFT JOIN b_member_card ON b_member_card.payid=payment.id LEFT JOIN b_users ON payment.xdy_id=b_users.uid LEFT JOIN zone ON payment.zone_id=zone.zone_id LEFT JOIN admin ON payment.sale_id=admin.id LEFT JOIN b_card_info ON payment.vip_card = b_card_info.id WHERE ( `sale_id` IN ('39','110','159','311','349','353','362','432','449','450','489','501','533','562') ) AND ( `payed` = 2 ) ORDER BY payment.id desc LIMIT 0,10;

+----+-------------+---------------+--------+---------------+---------+---------+---------------------------------------+------+-------------+
| id | select_type | table | type | possible_keys | key | key_len | ref | rows | Extra |
+----+-------------+---------------+--------+---------------+---------+---------+---------------------------------------+------+-------------+
| 1 | SIMPLE | payment | index | NULL | PRIMARY | 4 | NULL | 10 | Using where | ---比上面的最佳化更好了一下，用到主鍵索引了
| 1 | SIMPLE | b_member_card | ref | payid | payid | 5 | interface_hd_com0624.payment.id | 1 | |
| 1 | SIMPLE | b_users | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.xdy_id | 1 | |
| 1 | SIMPLE | zone | eq_ref | PRIMARY | PRIMARY | 3 | interface_hd_com0624.payment.zone_id | 1 | |
| 1 | SIMPLE | admin | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.sale_id | 1 | |
| 1 | SIMPLE | b_card_info | eq_ref | PRIMARY | PRIMARY | 4 | interface_hd_com0624.payment.vip_card | 1 | |
+----+-------------+---------------+--------+---------------+---------+---------+---------------------------------------+------+-------------+
6 rows in set (0.00 sec)

現在返回去思考為什麼會出現filesort排序
filesort排序定義：表示 MySQL 會對結果使用一個外部索引排序,而不是從表裡按索引次序讀到相關內容。可能在記憶體或者磁碟上進行排序。MySQL 中無法利用索引完成的排序操作稱為“檔案排序（filesort）
高效能mysqlP222頁
mysql架構設計與最佳化155頁

1.如果order by子句中的col都是來自於第一張表（但不滿足索引排序的條件），那麼mysql在關聯處理第一個表時就會進行檔案的排序。那麼我們在explain的extra欄位中就會看到useing filesort

使用索引掃描來進行排序

相關文章