SQL 交叉表数据透视查询答案

【问题标题】：SQL cross tab pivot querySQL 交叉表数据透视查询
【发布时间】：2012-04-07 02:02:56
【问题描述】：

我有一张如下表。

CaseID   StatusID    StageID     CaseRegisterTime   City
1        1            5            datetime         XYZ
2        1            5            datetime         ABC

现在我想要它的 Citywise 计数，并且仅适用于特定日期，并且在 statusid = 1 和 stageid = 5 的条件下。

Cities     CurrentDate-1    CurrentDate-2   January2012-CurrentDate-3
XYZ           5                  51                  5008
JUS           0                   0                   125
ABC           1                   0                   48

我希望我的标题为 CaseRegisterTime 分组案例，如上所示。

请帮忙。

【问题讨论】：

尝试谷歌搜索数据透视表查询 sql 你会发现很多你正在尝试做的例子。
我尝试过旋转，但主要问题是我只限制了日期列，如上面的示例输出所示，有限日期不想查找所有其他日期。
你从哪里得到日期？一般来说，当我们问“你试过什么？”我们想看代码
@Keyur，请提供更有意义的数据（即使它是假的）......很难看出你想要的结果来自哪里。那些 CurrentDate 列应该是计数、总和还是什么？
@Brian Just Google 这没有帮助.. 因为当其他人在进行 Google 搜索时，他们会不断得到结果，上面写着“只需 Google it”。导致讨厌的递归。

标签： sql sql-server sql-server-2008 tsql crosstab

【解决方案1】：

使用case when 将您感兴趣的日期转换为'CurrentDate-1' 和'CurrentDate-2'，然后使用此字符串作为新列旋转结果。

或者，您可以执行以下操作：

select City, sum(Date1) as Date1, sum(Date2) as Date2
from(
select City, 
    case when CaseRegisterTime='2012-01-01' then 1 else 0 end as Date1,
    case when  CaseRegisterTime='2012-15-01' then 1 else 0 end as Date2
from sample
) as T
group by City

您还必须过滤掉没有所需日期的寄存器。

【讨论】：

【解决方案2】：

这是在 SQL Server 2008 中执行此操作的多种方法之一（使用 Date 数据类型）：

select distinct a.City as Cities
    , (select count(*)
        from MyTable
        where CaseRegisterTime >= cast(getdate() - 1 as date)
            and CaseRegisterTime < cast(getdate() - 0 as date)
            and StatusID = a.StatusID
            and StageID = a.StageID
            and City = a.City
    ) as [CurrentDate-1]
    , (select count(*)
        from MyTable
        where CaseRegisterTime >= cast(getdate() - 2 as date)
            and CaseRegisterTime < cast(getdate() - 1 as date)
            and StatusID = a.StatusID
            and StageID = a.StageID
            and City = a.City
    ) as [CurrentDate-2]
    , (select count(*)
        from MyTable
        where CaseRegisterTime >= cast('20120101' as date)
            and CaseRegisterTime < cast(getdate() - 2 as date)
            and StatusID = a.StatusID
            and StageID = a.StageID
            and City = a.City
    ) as [January2012-CurrentDate-3]
from MyTable a
where a.StatusID = 1
    and a.StageID = 5

更新

@JotaBe 使用的 case 和 sum 方法在我的机器上的速度大约是两倍（扫描和读取次数更少），所以它可能看起来像这样：

select a.City as Cities
    , sum(a.[CurrentDate-1]) as [CurrentDate-1]
    , sum(a.[CurrentDate-2]) as [CurrentDate-2]
    , sum(a.[January2012-CurrentDate-3]) as [January2012-CurrentDate-3]
from (
    select City
        , case when CaseRegisterTime >= cast(getdate() - 1 as date)
                and CaseRegisterTime < cast(getdate() - 0 as date)
                then 1 else 0 end [CurrentDate-1]
        , case when CaseRegisterTime >= cast(getdate() - 2 as date)
                and CaseRegisterTime < cast(getdate() - 1 as date)
                then 1 else 0 end [CurrentDate-2]
        , case when CaseRegisterTime >= cast('20120101' as date)
                and CaseRegisterTime < cast(getdate() - 2 as date)
                then 1 else 0 end [January2012-CurrentDate-3]
    from MyTable 
    where StatusID = 1
        and StageID = 5
) as a
group by a.City

【讨论】：

【解决方案3】：

这样就可以了：

begin tran;
go
create table #t1(
    ID int identity,
    City varchar,
    RegisterDate  datetime
);
declare  @firstDate datetime, @secondDate datetime;
 set @firstDate = '2012-1-1';
 set @secondDate = '2012-1-2';
insert into #t1 values
    ('A', @firstDate),
    ('A', @firstDate),
    ('B', @firstDate),
    ('B', @firstDate),
    ('B', @firstDate),
    ('A', @secondDate),
    ('A', @secondDate),
    ('A', @secondDate),
    ('B', @secondDate),
    ('B', @secondDate);

select * from #t1;

select pvt.*
from(
    select ID, City, RegisterDate
    from #t1
) a
pivot(
    count(a.ID)
    for a.RegisterDate in ([2012-1-1], [2012-1-2])
)   as pvt;

drop table #t1;
go
rollback tran;

【讨论】：