根据多列检查选择唯一记录

Question

我有一个postgres表fetches，其中包含以下列

id  |  status  |  company_id  |  created_at
-------------------------------------------
10  |  success |  2773        |  2017-03-14
11  |  error   |  190         |  2017-03-02
12  |  error   |  303         |  2017-03-01
13  |  error   |  286         |  2017-02-16
14  |  error   |  2773        |  2016-12-05

我试图达到以下结果：

id  |  status  |  company_id  |  created_at
-------------------------------------------
11  |  error   |  190         |  2017-03-02
12  |  error   |  303         |  2017-03-01
13  |  error   |  286         |  2017-02-16

这是我试图应用的逻辑：

1. 获取状态为error的所有提取。
2. 如果结果中包含company_id fetch，最新的created_at（success}为company_id = 2773，则将其从最终结果集中跳过。
3. success将被跳过，因为它的最新获取状态为select [ExtraDimension].[ExtraHierarchy].[ExtraLevel].[All] on 0, [Calendar].[Year].[Year].Member * {[Measures].[ValueA],[Measures.ValueB],[Measures.ValueC]} on 1 from [MyCube]

请告知我如何以SQL方式完成此任务。

Answer 1

尝试外连接：

with succ as (select company_id from fetches where status = 'success')
select fetches.* 
from fetches
left outer join succ on succ.company_id = fetches.company_id
where succ.company_id is null;

Answer 2

您可以使用窗口函数：

select f.*
from (select f.*,
             sum( (status = 'success')::int) over (partition by company_id) as isSuccess
      fetches f
     ) f
where isSuccess = 0 and status = 'error';

Answer 3

似乎你需要公司的行没有成功的状态

select * from my_table 
where company_id not in (select company_id from my_table where status = 'success')