不带聚合的SQL pivot

更新

我已经修改了SQL FIDLE，以使部门名称更能反映我想要达到的目标。如果你看小提琴，有3个select语句

从数据中按正确顺序列出部门 这是我尝试旋转的结果 假选择，以便显示我想要达到的目标 在我的透视尝试的结果中，第二行显示departmentLevel1为空，departmentLevel3不应在此行中。我尝试了行数和分区的各种组合，但我似乎无法解决这个问题。完全沮丧

这是我最新的小提琴

在你投票结束之前 请阅读此问题，然后阅读问题的副本。所提出的问题要简单得多。我的问题不仅是关于数据透视，而且是对数据透视结果进行分组，并比数据透视的开始和停止时间做更多的事情。我想我们只是在看标题。如果你不想阅读整个问题，这是可以的，但不要假设它是与其他一些标题类似的问题相同，请

我意识到我必须有一个聚合，所以我尝试使用max，但我真的只是试图以更表格的方式显示这些信息，以便我可以将其加入到另一个数据集

以下是起始数据：

select  departmentID, departmentParentID, departmentName, departmentGroupSortBy, departmentLevel, sortBy
from    vDepartmentList
where   regionID = 3
order by departmentGroupSortBy, departmentLevel, sortBy

departmentID    departmentParentID  departmentName  departmentGroupSortBy   departmentLevel sortBy
---------------------------------------------------------------------------------------------------
111                                 neth test dept  1                       1                1
115             111                 test sub1       1                       2               1
112                                 test2           2                       1                2
113             112                 new sub2        2                       2               1
114             112                 new sub 3a      2                       2                 2
116             114                 should L3       2                       3                 1

如果我运行此查询：

select  departmentParentID, [1] as departmentLevel1, [2] as departmentLevel2, [3] as departmentLevel3
from    (
    select  departmentParentID, departmentLevel, departmentName
    from    vDepartmentList
    where   regionID = 3
    ) p
pivot (
    max(departmentName)
    for departmentLevel in ([1],[2],[3])
    ) as  pvt
order by departmentParentID

我得到这个结果：

departmentParentID  departmentLevel1    departmentLevel2    departmentLevel3
----------------------------------------------------------------------------
                    test2       
111                                     test sub1   
112                                     new sub2    
114                                                          should L3

但我想得到的是：

departmentParentID  departmentLevel1    departmentLevel2    departmentLevel3
----------------------------------------------------------------------------
111                 neth test dept      test sub1   
112                 test2               new sub2    
112                 test2               new sub 3a          should L3

---

有人对我的错误有什么建议吗？

旋转在某种程度上就像分组一样。您需要原始源中的一列作为分组依据。在这种情况下，您没有它，而且他不知道“testsub1”以任何方式连接到“neth test dept”。这里我们需要的是为层次结构建立根。我们可以用一个小的递归CTE将t替换为vDepartmentList：

;with x as (
    select *, departmentID as root
    from #t
    where departmentParentID is null
    union all
    select #t.*, x.root
    from #t
    inner join x on #t.departmentParentID = x.departmentID
)
select * 
from x

现在我们有一个根，还有另一个问题：我们只有两个根，但你想在结果中有三行。Pivot运算符将基本上分组到根列上，并将“new sub2”、“new Sub3A”和“should L3”视为都属于“test2”，并生成具有这些值各自最大值的单行

应该有区别的方法，但我在你的数据中看不到。看来，你运气不好。

我根据你的视图sql创建了下面的sql。你几乎可以在那里得到结果

我使用@temp table删除order by子句。也许您可以在实现中删除@temp表

请尝试以下sql：

declare @temp table
(
  id int identity,
  departmentLevel1 int,
  departmentLevel2 int,
  departmentLevel3 int,
  xRow int
)

insert into @temp 
select 
    d.lvl1, d.lvl2, d.lvl3
    ,Rrow
from
(
select 
    a.lvl1, a.lvl2, a.lvl3
    ,row_number() over(partition by a.lvl1,a.lvl2 order by a.lvl1,a.lvl2) as Rrow
    ,departmentGroupSortby, departmentlevel, sortby
from 
    vDepartmentList a
) d
where (d.lvl2 is not null) 
order by departmentGroupSortby, departmentlevel, sortby
;

select
    c.departmentLevel1 as departmentParentID
    ,Case
        when c.departmentLevel1 = d.departmentid then d.DepartmentName
        else 
            null
     end as DepartmentLevel1
    ,Case
        when c.departmentLevel2 = e.departmentid then e.DepartmentName
        else 
            null
     end as DepartmentLevel2
    ,Case
        when c.departmentLevel3 = f.departmentid then f.DepartmentName
        else 
            null
     end as DepartmentLevel3
    --,d.departmentId   ,e.departmentid ,f.departmentid
from
(
select 
    a.id, a.departmentLevel1,a.departmentLevel2,a.departmentLevel3
    ,a.xRow
from @temp a
inner join
(
select 
    cast(departmentLevel1 as nvarchar(5)) + cast(departmentLevel2 as nvarchar(5)) as xrow, count(cast(departmentLevel1 as nvarchar(5)) + cast(departmentLevel2 as nvarchar(5))) as xcount
from @temp
group by cast(departmentLevel1 as nvarchar(5)) + cast(departmentLevel2 as nvarchar(5))
) b on cast(a.departmentLevel1 as nvarchar(5)) + cast(a.departmentLevel2 as nvarchar(5)) = b.xrow and a.xRow = b.xcount
) c
left join department d on c.departmentLevel1 = d.departmentId
left join department e on c.departmentLevel2 = e.departmentid
left join department f on c.departmentLevel3 = f.departmentid

结果

---

希望这对您有所帮助

因为您需要返回层次结构数据结构的3个级别，所以使用联接而不是枢轴可能会更容易。首先返回父行：

select p.departmentName
from dbo.department p
where p.departmentParentID is null;

这将获得所有顶级行。然后，您开始为您需要的每个附加级别添加一个联接，在您的情况下，您总共需要三个级别，因此您将添加两个联接。最终查询将如下所示：

select p.departmentName department1,
    c.departmentName department2,
    gc.departmentName department3
from dbo.department p         -- parent level
left join dbo.department c    -- child level 
    on p.departmentID = c.departmentParentID
left join dbo.department gc   -- grandchild level
    on c.departmentID = gc.departmentParentID
where p.departmentParentID is null;

看

似乎使用联接而不是枢轴更容易得到结果。这将使用源数据，而不是您创建的递归视图。这产生了以下结果：

| DEPARTMENT1 | DEPARTMENT2 |  DEPARTMENT3 |
|-------------|-------------|--------------|
|     depart1 |     d1 sub1 |       (null) |
|     depart2 |      d2 sub |       (null) |
|     depart2 |     d2 sub2 | d2s2 subSub1 |

如果要使用递归查询获得结果，则可以稍微更改当前视图，以返回3列名称，而不是级别号：

;with cte as
(
    select departmentID, 
        departmentParentID,  
        departmentName as Department1,
        cast(null as varchar(100)) Department2,
        cast(null as varchar(100)) Department3,
        1 as Level
    from dbo.department
    where departmentParentID is null
    union all
    select d.departmentID, 
        d.departmentParentID,
        Department1,
        case when Level + 1 = 2 then d.departmentName else Department2 end,
        case when Level + 1 = 3 then d.departmentName else Department3 end,
        Level + 1
    from dbo.department d
    inner join cte h
        on d.departmentParentID = h.departmentID
)
select *
from cte;

---

看。然后，您可以使用WHERE子句进行一些筛选，以返回包含您需要的所有部门值的行

在子查询中使用row_number为parentid级别中的每个项创建唯一的值。这将允许返回多行。可能重复的行不是重复的行。部门级别的复杂性使其与简单地旋转开始和停止时间大不相同。也许这是我的问题的一部分，我试着去做很多事情once@bluefeet感谢您添加了一个行号，现在上面看到的结果至少显示了所有记录，但仍然没有将所有记录都放在正确的行上。我不知道如何正确地对这个结果进行分组。我对这个结果做了sql处理，因为我不容易解释。递归CTE是我试图改变观点的方式，但可能是我做错了。如果你有机会看看小提琴，看看我搞砸了什么，我很感激它。我投你一票，因为你是对的。我将为BlueFoots的简单性做一个标记。真棒@BlueFoots谢谢。我想我把注意力集中在一个复杂的问题上，真的需要退一步，从一个简单的方法来看待它。简单易行@兰斯：是的，当您最初发布它时，它似乎是一个简单的轴心，这就是为什么我建议使用行编号，但一旦您提供了其他详细信息，显然使用联接的查询将更容易

;with cte as
(
    select departmentID, 
        departmentParentID,  
        departmentName as Department1,
        cast(null as varchar(100)) Department2,
        cast(null as varchar(100)) Department3,
        1 as Level
    from dbo.department
    where departmentParentID is null
    union all
    select d.departmentID, 
        d.departmentParentID,
        Department1,
        case when Level + 1 = 2 then d.departmentName else Department2 end,
        case when Level + 1 = 3 then d.departmentName else Department3 end,
        Level + 1
    from dbo.department d
    inner join cte h
        on d.departmentParentID = h.departmentID
)
select *
from cte;