平时在客户业务的数据分析中,经常有这么一种需求,那就是希望得到某个组里面某些字段最大或最小的记录.
比如雇员表:

[local]:5432 pg12@testdb=# \d employee
                       Table "public.employee"
   Column   |         Type          | Collation | Nullable | Default 
------------+-----------------------+-----------+----------+---------
 id         | integer               |           |          | 
 name       | character varying(30) |           |          | 
 department | character varying(30) |           |          | 
 salary     | double precision      |

我们通过 MockData生成测试数据,共1000行,department共有12个.

为开福等地区用户提供了全套网页设计制作服务，及开福网站建设行业解决方案。主营业务为成都网站建设、成都网站设计、开福网站设计，以传统方式定制建设网站，并提供域名空间备案等一条龙服务，秉承以专业、用心的态度为用户提供真诚的服务。我们深信只要达到每一位用户的要求，就会得到认可，从而选择与我们长期合作。这样，我们也可以走得更远！

[local]:5432 pg12@testdb=# select count(*) from employee;
 count 
-------
  1000
(1 row)
Time: 22.747 ms
[local]:5432 pg12@testdb=# select distinct department from employee;
        department        
--------------------------
 Marketing
 Training
 Sales
 Business Development
 Product Management
 Research and Development
 Support
 Legal
 Accounting
 Services
 Human Resources
 Engineering
(12 rows)
Time: 2.616 ms

下面希望得到每个department中salary中最大的employee.
常规的做法是使用分组求得最大值/最小值,然后进行关联查询:

[local]:5432 pg12@testdb=# select a.* from employee a,(select department,max(salary) as salary  from employee group by department) b 
pg12@testdb-# where a.department = b.department and a.salary = b.salary order by a.department;
 id  |        name        |        department        | salary  
-----+--------------------+--------------------------+---------
 453 | Ericha Hendrikse   | Accounting               |  9958.5
 307 | Kyle Hartegan      | Business Development     | 9754.93
 969 | Odelinda Marsden   | Engineering              |  9942.3
 201 | Glen Kasperski     | Human Resources          | 9559.54
 892 | Mirabelle Lesslie  | Legal                    | 9720.49
 214 | Chane Koschek      | Marketing                | 9943.86
 371 | Josy Ayliff        | Product Management       | 9975.48
 191 | Meir Alvaro        | Research and Development |    9870
 770 | Adoree de Guerre   | Sales                    | 9808.65
 370 | Benoite Overlow    | Services                 | 9884.79
 866 | Shirlee McIlherran | Support                  | 9884.08
 586 | Renae Jerromes     | Training                 | 9904.24
(12 rows)
Time: 8.256 ms
[local]:5432 pg12@testdb=#

这种方法有个问题是如果max salary有多条记录的话,上述查询的结果会有多条.
PostgreSQL提供了DISTINCT ON,可简单实现该需求

[local]:5432 pg12@testdb=# SELECT DISTINCT ON (department)
pg12@testdb-#     *
pg12@testdb-# FROM
pg12@testdb-#     employee
pg12@testdb-# ORDER BY
pg12@testdb-#     department,
pg12@testdb-#     salary DESC;
 id  |        name        |        department        | salary  
-----+--------------------+--------------------------+---------
 453 | Ericha Hendrikse   | Accounting               |  9958.5
 307 | Kyle Hartegan      | Business Development     | 9754.93
 969 | Odelinda Marsden   | Engineering              |  9942.3
 201 | Glen Kasperski     | Human Resources          | 9559.54
 892 | Mirabelle Lesslie  | Legal                    | 9720.49
 214 | Chane Koschek      | Marketing                | 9943.86
 371 | Josy Ayliff        | Product Management       | 9975.48
 191 | Meir Alvaro        | Research and Development |    9870
 770 | Adoree de Guerre   | Sales                    | 9808.65
 370 | Benoite Overlow    | Services                 | 9884.79
 866 | Shirlee McIlherran | Support                  | 9884.08
 586 | Renae Jerromes     | Training                 | 9904.24
(12 rows)
Time: 11.445 ms

Excellent Feature!

参考资料
The Many Faces of DISTINCT in PostgreSQL