按连续外键​​值分组的 SQL 查询

Question

我有这样的数据:

+----+-------------------------+----------+----------+
| ID |      DateReceived       | Quantity | VendorID |
+----+-------------------------+----------+----------+
|  1 | 2010-08-09 06:53:44.783 |        2 |        1 |
|  2 | 2010-08-01 13:31:26.893 |        1 |        1 |
|  3 | 2010-07-26 07:52:29.403 |        2 |        1 |
|  4 | 2011-03-22 13:31:11.000 |        1 |        2 |
|  5 | 2011-03-22 13:31:11.000 |        1 |        2 |
|  6 | 2011-03-22 11:27:01.000 |        1 |        2 |
|  7 | 2011-03-18 09:04:58.000 |        1 |        1 |
|  8 | 2011-12-17 08:21:29.000 |        1 |        3 |
|  9 | 2012-08-10 10:55:20.000 |        9 |        3 |
| 10 | 2012-08-02 20:18:10.000 |        5 |        1 |
| 11 | 2012-07-12 20:44:36.000 |        3 |        1 |
| 12 | 2012-07-05 20:45:29.000 |        1 |        1 |
| 13 | 2013-03-22 13:31:11.000 |        1 |        2 |
| 14 | 2013-03-22 13:31:11.000 |        1 |        2 |
+----+-------------------------+----------+----------+

我想按 DateReceived 对数据进行排序并对 Quantity 求和。但是，我想对按 VendorID 分组的 Quantity 求和，只要它们像下面的示例输出那样相邻即可。

+----------+----------+
| VendorID | Quantity |
+----------+----------+
|        1 |        5 |
|        2 |        3 |
|        1 |        1 |
|        3 |       10 |
|        1 |        9 |
|        2 |        2 |
+----------+----------+

我目前正在通过加载所有行并在我的应用程序代码中遍历它们来执行此操作。这是目前我想要消除的软件瓶颈。

生成所需输出的 ​​MS Sql Server 查询是什么？

附言。对更好的标题有什么建议吗？

Answer 1

在 SQL Server 2005+ 中，您可以这样做:

with cte as (
    select
        VendorID, Quantity,
        row_number() over(partition by VendorID order by DateReceived) as rn1,
        row_number() over(order by DateReceived) as rn2
    from Table1
)
select
    VendorID, sum(Quantity) as Quantity
from cte 
group by VendorID, rn2 - rn1
order by min(rn2)

sql fiddle demo

在 SQL Server 2012 中，您可以使用 lag() 函数:

with cte as (
    select
        VendorID, Quantity, DateReceived,
        case when lag(VendorID) over(order by DateReceived) <> VendorID then 1 else 0 end as rn
    from Table1
), cte2 as (
    select
        VendorID, Quantity, sum(rn) over(order by DateReceived) as s
    from cte 
)
select
    VendorID, sum(Quantity)
from cte2
group by VendorID, s
order by s asc

sql fiddle demo

顺便说一句，看起来你的输出不正确。正确的是:

+----------+----------+
| VendorID | Quantity |
+----------+----------+
|        1 |        6 |
|        2 |        3 |
|        3 |        1 |
|        1 |        9 |
|        3 |        9 |
|        2 |        2 |
+----------+----------+