Sql 如何根据案例分组

我有一个脚本，它可以整理两个日期之间的登录日志，并将它们分类为三个时间段之一。我需要它将它们组合在一起，统计在这些时间段内登录到系统的唯一用户的数量。例如：

SQL:

选择 案例 在castdatediffsecond、'01-01-1970 00:00:00'、'2018-07-23 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2018-12-09 23:59:59'作为varchar之间创建lsl.TIME时，2018年第二学期：2018年7月23日至2018年12月9日 然后是“2018年第二学期” 当在castdatediffsecond、'01-01-1970 00:00:00'、'2018-12-10 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2018-02-03 23:59:59'作为varchar之间创建lsl.time时，2018-19:10/12/2018至03/02/2019 然后是“圣诞假期，2018-19” 在castdatediffsecond、'01-01-1970 00:00:00'、'2019-02-04 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2019-06-23 23:59:59'作为varchar之间创建lsl.TIME时，2019年第1学期：2019年2月4日至2019年6月23日 然后是“2019年第一学期” else'圣诞假期，2018-19' 以“登录日期”结束， countdistinct lsl.userid为“唯一登录名”， lsl.行动 来自前缀\u日志存储\u标准\u日志lsl 其中lsl.action='loggedin'和 lsl.time在castdatediffsecond、'01-01-1970 00:00:00'、'2018-07-23 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2019-06-23 23:59:59'作为varchar之间创建 按lsl.timecreated、lsl.action分组 按lsl.timecreated asc订购 这将按预期列出各行上的数据：

------------------------------------------------------------
| Login Date               | # of Unique Logins | action   |
------------------------------------------------------------
| Semester 2, 2018         | 1                  | loggedin |
| Semester 2, 2018         | 1                  | loggedin |
| Semester 2, 2018         | 1                  | loggedin |
| Semester 2, 2018         | 1                  | loggedin |
| Christmas Break, 2018-19 | 1                  | loggedin |
| Christmas Break, 2018-19 | 1                  | loggedin |
| Christmas Break, 2018-19 | 1                  | loggedin |
| Semester 1, 2019         | 1                  | loggedin |
| Semester 1, 2019         | 1                  | loggedin |
------------------------------------------------------------

我需要将登录日期分组如下。我该怎么做

------------------------------------------------------------
| Login Date               | # of Unique Logins | action   |
------------------------------------------------------------
| Semester 2, 2018         | 4                  | loggedin |
| Christmas Break, 2018-19 | 3                  | loggedin |
| Semester 1, 2019         | 2                  | loggedin |
------------------------------------------------------------

我试图通过将相同的案例添加到组中，但它抛出了一个我不理解的错误：

... 分组 案例 在castdatediffsecond、'01-01-1970 00:00:00'、'2018-07-23 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2018-12-09 23:59:59'作为varchar之间创建lsl.TIME时，2018年第二学期：2018年7月23日至2018年12月9日 然后是“2018年第二学期” 当在castdatediffsecond、'01-01-1970 00:00:00'、'2018-12-10 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2018-02-03 23:59:59'作为varchar之间创建lsl.time时，2018-19:10/12/2018至03/02/2019 然后是“圣诞假期，2018-19” 在castdatediffsecond、'01-01-1970 00:00:00'、'2019-02-04 00:00:00'作为varchar和castdatediffsecond、'01-01-1970 00:00:00'、'2019-06-23 23:59:59'作为varchar之间创建lsl.TIME时，2019年第1学期：2019年2月4日至2019年6月23日 然后是“2019年第一学期” 终止 lsl.行动 按lsl.timecreated asc订购 错误：

SQLState:42000 错误代码：8120 列前缀_logstore_standard_log.timecreated在选择列表中无效，因为它未包含在聚合函数或GROUP BY子句中

---

您可以使用外部查询进行聚合

select "Login Date" , count("# of unique logins") , action
from (
    select
        case
            when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-07-23 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2018-12-09 23:59:59') as varchar) -- Semester 2, 2018: 23/07/2018 to 09/12/2018
                then 'Semester 2, 2018'
            when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-12-10 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2018-02-03 23:59:59') as varchar) -- Christmas Break, 2018-19: 10/12/2018 to 03/02/2019
                then 'Christmas Break, 2018-19'
            when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2019-02-04 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2019-06-23 23:59:59') as varchar) -- Semester 1, 2019: 04/02/2019 to 23/06/2019
                then 'Semester 1, 2019'
            else 'Christmas Break, 2018-19'
        end as 'Login Date',
        count(distinct lsl.userid) as '# of unique logins',
        lsl.action
    from prefix_logstore_standard_log lsl
    where lsl.action = 'loggedin' and
        (lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-07-23 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2019-06-23 23:59:59') as varchar))
    group by lsl.timecreated, lsl.action
    order by lsl.timecreated asc
) as a
group by "Login Date" , action

---

使用CTE公共表表达式，它应该可以解决您的问题

不幸的是，我没有您的数据要测试，但是，以下内容应该可以工作：

with cte ([Login Date],[# of Unique Logins],[action])
as
(
select
    case
        when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-07-23 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2018-12-09 23:59:59') as varchar) -- Semester 2, 2018: 23/07/2018 to 09/12/2018
            then 'Semester 2, 2018'
        when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-12-10 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2018-02-03 23:59:59') as varchar) -- Christmas Break, 2018-19: 10/12/2018 to 03/02/2019
            then 'Christmas Break, 2018-19'
        when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2019-02-04 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2019-06-23 23:59:59') as varchar) -- Semester 1, 2019: 04/02/2019 to 23/06/2019
            then 'Semester 1, 2019'
        else 'Christmas Break, 2018-19'
    end as 'Login Date',
    count(distinct lsl.userid) as '# of unique logins',
    lsl.action
from prefix_logstore_standard_log lsl
where lsl.action = 'loggedin' and
    (lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-07-23 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2019-06-23 23:59:59') as varchar))
group by lsl.timecreated, lsl.action
order by lsl.timecreated asc
)

select
*
from    cte
group by [Login Date],[# of Unique Logins],[action]

---

让我知道它是否有效。

您是否意外回复了错误的帖子？我看不出这段代码有什么关系。您也没有对代码所做的事情提供任何解释。我只是给出了如何对用例进行分组的示例

with cte ([Login Date],[# of Unique Logins],[action])
as
(
select
    case
        when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-07-23 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2018-12-09 23:59:59') as varchar) -- Semester 2, 2018: 23/07/2018 to 09/12/2018
            then 'Semester 2, 2018'
        when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-12-10 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2018-02-03 23:59:59') as varchar) -- Christmas Break, 2018-19: 10/12/2018 to 03/02/2019
            then 'Christmas Break, 2018-19'
        when lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2019-02-04 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2019-06-23 23:59:59') as varchar) -- Semester 1, 2019: 04/02/2019 to 23/06/2019
            then 'Semester 1, 2019'
        else 'Christmas Break, 2018-19'
    end as 'Login Date',
    count(distinct lsl.userid) as '# of unique logins',
    lsl.action
from prefix_logstore_standard_log lsl
where lsl.action = 'loggedin' and
    (lsl.timecreated between cast(datediff(second,'01-01-1970 00:00:00','2018-07-23 00:00:00') as varchar) and cast(datediff(second,'01-01-1970 00:00:00','2019-06-23 23:59:59') as varchar))
group by lsl.timecreated, lsl.action
order by lsl.timecreated asc
)

select
*
from    cte
group by [Login Date],[# of Unique Logins],[action]