SELECT *,row_number() over(partition by col1 order by col2) as rn1
这是什么意思呢？
简单的说就是进行分组排序并编号的过程。
row_number()很好理解，给每条记录分配一个行号，从1开始，连续的。那么over (partition by col1 order by col2 desc) 是什么意思呢？就是先按照col1进行分组，然后分别在每个组内按照col2进行降序排序，然后编号。
比如我们有一张表，里边有每个客户所领取的优惠券。现在我们需要拿到每个客户的最近一张优惠券，怎么写SQL呢？
customers_coupons_tbl
| CID | name | date |
| 1 | shoe | 20181130 |
| 1 | book | 20181212 |
| 2 | beauty | 20181201 |
| 2 | drug | 20181230 |
| 2 | book | 20181212 |
select *, row_number() over (partition by CID order by date desc) as rn1
from customers_couopns_tbl;
| CID | name | date | rn1 |
| 1 | book | 20181212 | 1 |
| 1 | shoe | 20181130 | 2 |
| 2 | beauty | 20181230 | 1 |
| 2 | drug | 20181215 | 2 |
| 2 | book | 20181212 | 3 |
然后我们再在上述表上进行过滤，把rn1不是1的都去掉：
select CID, name, date
from (
select *, row_number() over (partition by CID order by date desc) as rn1
from customers_couopns_tbl;
)
where rn1 = 1;
| CID | name | date |
| 1 | book | 20181212 |
| 2 | beauty | 20181230 |
好了，你现在肯定理解了什么是分组排序了。另外，rank()函数也可以替代row_number()，效果是一样的。
原文:http://blog.csdn.net/hongchangfirst/article/details/84773366
作者:hongchangfirst
hongchangfirst的主页:http://blog.csdn.net/hongchangfirst