left join 的一些问题，不一定以左表为准，数据行数减少或者数据增多

 

left join 完，行数一定等于左表行数吗？ 不一定！！

实验如下：

SELECT * FROM hdd_test1 a LEFT JOIN hdd_test2 b ON a.a_id = b.b_id LIMIT 144

8 rows (134ms)

id	created_at	updated_at	a_id	id	created_at	updated_at	b_id
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1	1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1
2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2	2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1	3	2018-11-05 20:44:00.0	2018-11-05 20:44:00.0	1
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1	4	2018-11-05 20:44:04.0	2018-11-05 20:44:04.0	1
3	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	3
4	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	4
5	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	5
6	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	6

hdd_test1有6条记录， hdd_test2 有4条记录，直接left join ，产生6条记录，别高兴太早，看下面的查询：

SELECT * FROM hdd_test1 a LEFT JOIN hdd_test2 b ON a.a_id = b.b_id WHERE b.b_id > 1 LIMIT 144

1 row (37ms)查看CSV

id	created_at	updated_at	a_id	id	created_at	updated_at	b_id
2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2	2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2

 

如图所示，当where 条件在外，会先聚合后过滤，所以结果会变少

 

还有一种情况需要注意,left join 结果的行数也可能大于左表：

SELECT * FROM hdd_test1 LIMIT 144

2 rows (33ms)查看CSV

id	created_at	updated_at	a_id
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1
2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2

SELECT * FROM hdd_test2 LIMIT 144

4 rows (34ms)查看CSV

id	created_at	updated_at	b_id
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1
2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2
3	2018-11-05 20:44:00.0	2018-11-05 20:44:00.0	1
4	2018-11-05 20:44:04.0	2018-11-05 20:44:04.0	1

SELECT * FROM hdd_test1 a LEFT JOIN hdd_test2 b ON a.a_id = b.b_id LIMIT 144

4 rows (110ms)查看CSV

id	created_at	updated_at	a_id	id	created_at	updated_at	b_id
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1	1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1
2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2	2	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	2
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1	3	2018-11-05 20:44:00.0	2018-11-05 20:44:00.0	1
1	2018-10-09 10:10:10.0	2018-10-09 10:10:10.0	1	4	2018-11-05 20:44:04.0	2018-11-05 20:44:04.0	1

看到没？ 产生了四条记录，因为这个时候b 有重复的b_id，如果这时候计算a表值的count 或者sum ，结果会比预想的大。

虽然很简单，但是很容易犯错误