SQL 使用 Case 语句分组

Question

我有一个 table 看起来像这样

+-----------+-----------+---------------------+
| Report_id |  Status   |        Date         |
+-----------+-----------+---------------------+
|         1 | Completed | 2020-06-07 12:20:00 |
|         1 | Completed | 2020-06-07 12:22:00 |
|         2 | Running   | 2020-06-07 13:02:00 |
|         2 | Completed | 2020-06-07 13:10:00 |
|         3 | Completed | 2020-06-07 14:10:00 |
|         3 | Failed    | 2020-06-07 14:04:00 |
+-----------+-----------+---------------------+

我需要按 Report_id 对这些数据进行分组。因此，如果组内的所有状态值都等于已完成，则 Status 为已完成且 Date 是组内的最大值。但是，如果组内有一个运行或 Failed 值，则 Status 需要分别等于运行和 Failed 以及 Date 应该匹配这个值。

输出将如下所示。

+-----------+-----------+---------------------+
| Report_id |  Status   |        Date         |
+-----------+-----------+---------------------+
|         1 | Completed | 2020-06-07 12:22:00 |
|         2 | Running   | 2020-06-07 13:02:00 |
|         3 | Failed    | 2020-06-07 14:04:00 |
+-----------+-----------+---------------------+

我怀疑我需要在某处使用 Case 语句来获取此输出，但我不知道如何操作。请帮忙。

Answer 1

查看您的示例，您似乎需要每个报告的第一个状态（基于日期）

select m.* 
from my_table m
inner join (
    select Report_id,  min(date) min_date
    from my_table  
    group by Report_id
) t on t.Report_id = m. and t.min_date  = m.date

Answer 2

您可以尝试使用 row_number()

select * from
(
select report_id,status,date, row_number() over(partition by report_id order by 
case when status in ('Running','Failed') then 1 else 2 end asc,Date desc) as rn
from tablename
)A where rn=1

Answer 3

您可以使用 STRING_AGG 将状态列的文本组合成一个字符串，并检查该字符串是否包含 'running' 或 'failed'

select report_id, 
case when STRING_AGG(status, '') like '%Running%' then 'Running'
when STRING_AGG(status, '') like '%Failed%' then 'Failed'
else 'Completed' end Status,
max(date) Date
from abc
group by report_id

Answer 4

不确定这是否是您要查找的内容，也不确定这是否完全运行。

select report_id,
    case when q.failed_sum > 0 then 'Failed'
         when q.running_sum > 0 then 'Running'
        else  'Completed'
    end,
    max(date)
    from table inner join
        (
            select report_id, 
            sum(case when status = 'Failed' then 1 end) as failed_sum,
            sum(case when status = 'Running' then 1 end) as running_sum,
            sum(case when status = 'Completed' then 1 end)as completed_sum
            from table 
            group by report_id
        )q on report_id = q.report_id
    where status = (case when failed_sum > 0 then 'Failed'
        else when running_sum > 0 then 'Running'
        else then 'Completed'
    end) 
    group by report_id

Answer 5

您可以使用 distinct on:

select distinct on (record_id) t.*
from t
order by (case when status <> 'Completed' then 1 else 2 end),
         date desc;

SQL 使用 Case 语句分组

SQL group by with Case statement

sql

postgresql

group-by

greatest-n-per-group