SQL Server - 多个 WHERE 查询的矩阵

Question

如果我有如下表：

Source  Event           Date        Qty
Site A  Create Account  5/05/2018   6
Site B  Create Account  4/05/2018   12
Site A  Update Account  6/05/2018   1
Site A  Update Notes    7/05/2018   2
Site B  Add Dependant   5/05/2018   1
Site C  Create Account  5/05/2018   14

还有一个这样的：

Date        OrdersRec
4/05/2018   162
5/05/2018   123
6/05/2018   45
7/05/2018   143

并希望构建一种矩阵表，以便能够一次又一次地用于各种日期范围的查询，如下所示：

Date        Create  Update  UpdateNotes AddDependant    OrdersRec
4/05/2018   12      0       0           0               162
5/05/2018   20      0       0           1               123
6/05/2018   0       1       0           0               45
7/05/2018   0       0       2           0               143

最好的方法是什么？

我从日期和 CreateAccount 数字的 INSERT 开始，但我不确定如何返回并更新特定日期的其他计数，或者即使那很愚蠢而且我遗漏了一些明显的东西。

谢谢！

Answer 1

只需使用union all 或join。没有理由为此创建一个新表：

select date, sum(v_create) as v_create, sum(v_update) as v_update,
       sum(v_updatenotes) as v_updatenotes, sum(v_adddependent) as v_adddependent,
       sum(orderrec) as orderec
from ((select t1.date, v.*, 0 as OrdersRec
       from table1 t1 cross apply
            (values ( (case when event = 'Create Account' then quantity else 0 end),
                      (case when event = 'Update Account' then quantity else 0 end),
                      (case when event = 'Update Notes' then quantity else 0 end),
                      (case when event = 'Add Dependent' then quantity else 0 end)
                    )
             ) v(v_create, v_update, v_updatenotes, v_adddependent)
      ) union all
      (select date, 0, 0, 0, 0, orderrec
       from table2
      )
     ) t
group by date
order by date;

这可以处理很多边缘情况，例如当一个或另一个表中缺少日期时。

Answer 2

为什么不创建一个包含所有日期的变量表，并将其用作主表。

那么你会是这样的

select m.date,
sum(case when event = 'Create Account' then 1 else 0) as create,
sum(case when event = 'Update Account' then 1 else 0) as updates,
sum(o.ordersrec) as orders
from @main as m
       inner join @orders as o on o.date = m.date
       inner join @events as e on e.date = m.date
group by m.date