SQL Server 2015中的查询导致问题

我的问题是：

select distinct 
    e.EmployeeId, EmployeeCode, EmployeeName, 
    Cast(d.LogDate as date) AttDate, 
    PunchTimes = Stuff((Select ', ' + convert(varchar, d.LogDate, 8) 
                        from DeviceLogs dl WITH (NoLock) 
                        where dl.UserId = e.EmployeeId 
                          and cast(dl.LogDate as date) = cast(d.LogDate as date) 
                        for XML path('') ), 1, 2, '') 
from 
    Employees e, DeviceLogs d
where 
    e.EmployeeId = d.UserId 
    --and cast(d.LogDate as date) between cast(@FromDate as date) and cast(@ToDate as date) 
    and cast(d.LogDate as date) between '1-Jun-2019' and '1-Jun-2019'
group by 
    e.EmployeeId, EmployeeCode, EmployeeName, Cast(d.LogDate as date)

我得到这个错误：

Msg 8120，16级，状态1，第2行 列“DeviceLogs.LogDate”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

您需要将castdl.logdateas date列添加到GROUPBY子句中

更正：误读子查询中的别名；问题是对子查询中使用了d.LogDate的相关引用。给定以下DDL

CREATE TABLE Employees (EmployeeId INT PRIMARY KEY
, EmployeeCode INT
, EmployeeName VARCHAR(255))

CREATE TABLE DeviceLogs (UserId INT FOREIGN KEY REFERENCES dbo.Employees(EmployeeId), LogDate DATETIME)

这项工作：-

select distinct e.EmployeeId, EmployeeCode, EmployeeName, Cast(d.LogDate as date) AttDate, 
    PunchTimes = Stuff((Select ', ' + convert(varchar, dl.LogDate, 8) 
                       from DeviceLogs dl WITH (NoLock) 
                        where dl.UserId = e.EmployeeId and cast(dl.LogDate as 
                        date) = cast(d.LogDate as date) for XML path('') ),1, 2, '') 
from Employees e, DeviceLogs d
where e.EmployeeId = d.UserId 

    and cast(d.LogDate as date) between '1-Jun-2019' and '1-Jun-2019'
group by e.EmployeeId, EmployeeCode, EmployeeName, Cast(d.LogDate as date)

但最初的问题不是：

select distinct e.EmployeeId, EmployeeCode, EmployeeName, Cast(d.LogDate as date) AttDate, 
    PunchTimes = Stuff((Select ', ' + convert(varchar, d.LogDate, 8) 
                       from DeviceLogs dl WITH (NoLock) 
                        where dl.UserId = e.EmployeeId and cast(dl.LogDate as 
                        date) = cast(d.LogDate as date) for XML path('') ),1, 2, '') 
from Employees e, DeviceLogs d
where e.EmployeeId = d.UserId 

    and cast(d.LogDate as date) between '1-Jun-2019' and '1-Jun-2019'
group by e.EmployeeId, EmployeeCode, EmployeeName, Cast(d.LogDate as date)

试试这个

    select distinct e.EmployeeId, EmployeeCode, EmployeeName, Cast(d.LogDate as date) AttDate, 
        PunchTimes = Stuff((Select ', ' + convert(varchar, d.LogDate, 8) from DeviceLogs dl WITH (NoLock) 
            where dl.UserId = e.EmployeeId and cast(dl.LogDate as date) = cast(d.LogDate as date) for XML path('') ),1, 2, '') 
    from Employees e, DeviceLogs d
    where e.EmployeeId = d.UserId 
        --and cast(d.LogDate as date) between cast(@FromDate as date) and cast(@ToDate as date) 
        and cast(d.LogDate as date) between '1-Jun-2019' and '1-Jun-2019'
    group by e.EmployeeId, EmployeeCode, EmployeeName, AttDate

any aggregate function is not allowed in group by

我建议避免分组方式，而是在子查询中使用select distinct：

select e.*, 
       stuff( (select ', ' + convert(varchar(255), d.LogDate, 8) 
               from DeviceLogs dl
               where dl.UserId = e.EmployeeId and 
                     cast(dl.LogDate as date) = e.AttDate
               for XML path('')
              ), 1, 2, ''
            ) as PunchTimes
from (select distinct e.EmployeeId, e.EmployeeCode, e.EmployeeName, 
             Cast(d.LogDate as date) as AttDate
            ) 
      from Employees e join
           DeviceLogs d
           on e.EmployeeId = d.UserId 
      where d.LogDate >= '2019-06-01' and
            d.LogDate < '2019-06-02'
     ) e;

注:

请勿在FROM子句中使用逗号。始终使用正确、明确、标准的联接语法。 使用标准日期格式，即“YYYYMMDD”或“YYYY-MM-DD”。 当一个表引用多个表时，限定所有列引用。 指定VARCHAR和类似类型时始终使用长度。默认值因上下文而异，可能会导致难以调试的错误。 只有当您真正理解NOLOCK的功能时，才能使用它。我猜你没有完全理解它。

---

使用GROUP BY子句而不使用任何聚合函数是个问题。错误很明显，但查询肯定不是。SELECT子句中似乎有一个子查询，筛选使用本地化的日期文本，这保证了转换问题。顺便说一下，NOLOCK表示读取脏数据，是否锁定，同时使用过多的锁定。看起来您遇到了性能问题，并试图用NOLOCK来掩盖它。这实际上会让事情变得更糟。至于错误，它是由子查询引起的。因为它只是一个子查询，而不是实际的聚合函数，所以它不能包含GROUP BY子句或聚合函数中没有的任何列。在SQL Server 2017中，您可以使用STRING_AGG进行此操作。在早期版本中，您必须使用SQLCLR方法进行字符串聚合没有SQL Server 2015版本-我们有2012年、2014年、2016年、2017年和2019年的版本-请选择已经存在的版本。我怀疑这个错误是关于子查询的，而不是外部查询。不，Castd.LogDate是日期，但castdl.LogDate不是日期。很容易错过，但它们都是GROUP BY中所必需的，或者对castdl.LogDate as date的引用更改为d.LogDate。将castdl.LogDate as date添加到GROUP BY子句中是不可能的，因为dl只出现在子查询中，它正试图执行字符串聚合，与SQL Server 2017的string_AGG或MySQL的GROUP_CONCAT一样。这在以前的版本中无法工作。Gordon Linoff通过完全删除GROUP by来展示这是如何工作的。毕竟，它只用于为字符串聚合提供不同的键。如果查询包含其他聚合，比如SUM，就不可能使用DISTINCTApologies，你是对的，我误读了子查询中的引用。实际的解决方案不需要重写，但另一个问题是在子查询中使用dl.LogDate而不是d.LogDate。PunchTimes=StuffSelect'，'+convertvarchar，dl.LogDate，8来自DeviceLogs dl和NoLock，其中dl.UserId=e.EmployeeId和castdl.LogDate as date=castd.LogDate as date作为XML路径的日期，1，2，GROUP BY中没有聚合函数。cast不是聚合函数