表连接类型

前言

对于关系型数据库来说，对于复杂的查询表连接是必不可少的，下面妹子就来介绍下表连接的方式

内连接（Inner Join）

Inner指的是 intersect相交,顾名思义就是取两表的交集。

SELECT COUNT(a.per_id) FROM `b_borrow_list` AS a;/*63条*/
SELECT COUNT(b.per_id) FROM `b_borrow_deductions` AS b;/*10条*/
/*inner-join*/
SELECT COUNT(a.`per_id`) FROM `b_borrow_list` AS a 
INNER JOIN `b_borrow_deductions` AS b 
ON a.per_id=b.per_id;/*51条*/

表a有63条数据，表b有10条数据，两表都有键值per_id，用INNER JOIN取交集为51条

外连接（Outer Join）

Outer指的是 union集合,就是取两表的并集。
其中外连接分为3种：

Full outer join

full outer join 得到两表的乘集,也可叫做笛卡尔乘积。如果a表中的行在b表中没有对应的部分,b的部分将是 null, 反之亦然。

SELECT COUNT(a.per_id) FROM `b_borrow_list` AS a;/*63条*/
SELECT COUNT(b.per_id) FROM `b_borrow_deductions` AS b;/*10条*/
/*full-outer-join*/
SELECT COUNT(a.`per_id`) FROM `b_borrow_list` AS a 
JOIN `b_borrow_deductions` AS b /*630条*/

取两表的乘集，则a表的行数乘以b表的行数，63*10=630（条）

Left join

左连接，取表a的所有行，和b表中匹配的值，没有匹配的则以null值取代。

SELECT COUNT(a.per_id) FROM `b_borrow_list` AS a;/*63条*/
SELECT COUNT(b.per_id) FROM `b_borrow_deductions` AS b;/*10条*/
/*left-join*/
SELECT COUNT(a.`per_id`) FROM `b_borrow_list` AS a 
LEFT JOIN `b_borrow_deductions` AS b 
ON a.per_id=b.per_id;/*63条*/

a表有63行，b表中键值per_id在a表中都有。

Right join

右连接，取表b的所有行，和a表中匹配的值，没有匹配的则以null值取代。

SELECT COUNT(a.per_id) FROM `b_borrow_list` AS a;/*63条*/
SELECT COUNT(b.per_id) FROM `b_borrow_deductions` AS b;/*10条*/
/*right-join*/
SELECT COUNT(a.`per_id`) FROM `b_borrow_list` AS a 
RIGHT JOIN `b_borrow_deductions` AS b 
ON a.per_id=b.per_id;/*51条*/

b表有10行，a表中有重复键值per_id，所以有51条

反连接（Anti Join）

Anti指的是 英语词根反对的意思，与inner join正好相反。
多用于!= not in 等查询；如果找到满足条件（!= not in）的不返回，不满足条件（!= not in）的返回。

SELECT COUNT(a.per_id) FROM `b_borrow_list` AS a;/*63条*/
SELECT COUNT(b.per_id) FROM `b_borrow_deductions` AS b;/*10条*/
/*anti-join*/
SELECT COUNT(a.`per_id`) FROM `b_borrow_list` AS a ,`b_borrow_deductions` AS b 
WHERE a.per_id!=b.per_id;/*579条*/

相当于a表与b表的乘级减去a表与b表的交集，即(63*10)-51=579条

半连接（Semi Join）

半连接多在子查询exists中使用，找到存在的键值之后就不再查找内部其他的键值了。

SELECT COUNT(a.per_id) FROM `b_borrow_list` AS a;/*63条*/
SELECT COUNT(b.per_id) FROM `b_borrow_deductions` AS b;/*10条*/
/*semi-join */
SELECT COUNT(a.`per_id`) FROM `b_borrow_list` AS a 
WHERE EXISTS( 
SELECT NULL FROM `b_borrow_deductions` AS b  WHERE a.per_id=b.per_id)/*51条*/

星型连接（Star Join）

是指一张大的事实表和多个小表关联得出所需的数据，往四周发散，所以称为星型链接。
要使用星型转换，必须要求初始化参数star_transforation_enabled设为TRUE。

SELECT ...
FROM   really_big_fact f    -- 50 million rows
,      dim1 d1              -- 50 rows
,      dim2 d2              -- 100 rows
,      dim3 d3              -- 10 rows
,      dim4 d4              -- 15 rows
WHERE  f.key1 = d1.key1
AND    f.key2 = d2.key2
AND    f.key3 = d3.key3
AND    f.key4 = d4.key4

现在只有Oracle才有

表链接的方式

表链接的方式有：排序合并连接(Sort Merge join)、嵌套循环连接(Nested Logo join)、哈希连接(Mash join)、笛卡尔链接(Cross join)

参考资料

• http://www.orafaq.com/tuningguide/star%20join.html
• https://raoenhui.github.io/sql/2018/11/30/sqlLink/

Happy coding .. :)