Sql 从组记录中选择前N名

当价格改变时，我需要列出第一条记录。使用下面的例子

发票表记录按客户分组并按价格和日期排序

Index  Customer Date         Price
01     A        May-01-2016  $12.00
02     A        May-11-2016  $12.00

03     A        May-21-2016  $13.00
04     A        May-22-2016  $13.00
05     A        May-23-2016  $13.00
06     A        May-24-2016  $13.00

07     A        Jun-01-2016  $14.00
08     A        Jun-11-2016  $14.00
09     A        Jun-21-2016  $14.00
10     A        Jun-25-2016  $14.00

11     B        May-02-2016  $12.50
12     B        May-12-2016  $12.50

13     B        May-22-2016  $13.50

14     B        May-24-2016  $13.80
15     B        May-26-2016  $13.80
16     B        May-28-2016  $13.80

17     B        Jun-02-2016  $14.60
18     B        Jun-12-2016  $14.60
19     B        Jun-22-2016  $14.60
20     B        Jun-26-2016  $14.60

当价格改变时，我需要得到第一张记录。结果将是：

01     A        May-01-2016  $12.00

03     A        May-21-2016  $13.00

07     A        Jun-01-2016  $14.00

11     B        May-02-2016  $12.50

13     B        May-22-2016  $13.50

14     B        May-24-2016  $13.80

17     B        Jun-02-2016  $14.60

03     A        May-21-2016  $13.00

07     A        Jun-01-2016  $14.00

14     B        May-24-2016  $13.80

17     B        Jun-02-2016  $14.60

除此之外，我是否可以将结果限制为每个客户的最新2条记录？结果将是：

01     A        May-01-2016  $12.00

03     A        May-21-2016  $13.00

07     A        Jun-01-2016  $14.00

11     B        May-02-2016  $12.50

13     B        May-22-2016  $13.50

14     B        May-24-2016  $13.80

17     B        Jun-02-2016  $14.60

03     A        May-21-2016  $13.00

07     A        Jun-01-2016  $14.00

14     B        May-24-2016  $13.80

17     B        Jun-02-2016  $14.60

谢谢。

第一部分：

SELECT id, customer, dte, price
  FROM (SELECT id, customer, dte, price
              ,COALESCE(LAG(price) OVER(PARTITION BY customer ORDER BY id), price - 1) AS prev_price
          FROM invoice) x
  WHERE x.price <> x.prev_price
  ORDER BY x.id

第二部分：

SELECT id, customer, dte, price
  FROM (SELECT id, customer, dte, price
              ,ROW_NUMBER() OVER (PARTITION BY customer ORDER BY id DESC) AS row_num
          FROM (SELECT id, customer, dte, price
                      ,COALESCE(LAG(price) OVER (PARTITION BY customer ORDER BY id), price - 1) AS prev_price
                  FROM invoice) x
          WHERE x.price <> x.prev_price) y
  WHERE row_num <= 2
  ORDER BY id

结果集

---

假设date列采用某种有效的dateformat，否则您应该对其进行转换，下面是另一种方法：

;with cte as (
    select 
       Invoice.[index], Invoice.Customer, Invoice.Date, Invoice.Price, 
       rn = row_number() over (partition by Invoice.customer order by date desc) 
    from Invoice
    join (
       select Customer, price, min(date) min_date 
       from Invoice 
       group by Customer, price
    ) t1 on Invoice.Customer = t1.Customer and Invoice.Date = t1.min_date
)
select [index], Customer, date, price 
from cte where rn <= 2 order by [index]

如果您使用的是SQL Server 2012+，则可以使用窗口聚合函数并执行此查询，该查询可能更有效，也可能更无效：

select [index], Customer, date, price 
from (
    select *, rn = dense_rank() over (partition by customer order by min_date desc) 
    from (select *, min(date) over (partition by customer, price) min_date from Invoice) a
) b where rn <= 2 and Date = min_date order by [index]

---

对我来说，lag似乎是最好的解决方案。如果lag函数可用，确实是一个很好的解决方案。如果它是Access sql，它将是什么？@YellowLarry Access没有任何窗口函数，因此您必须创造性地使用多个嵌套查询，或者在VBA中执行。您使用的sql Server的具体版本是什么？某些窗口功能需要2012+

select [index], Customer, date, price 
from (
    select *, rn = dense_rank() over (partition by customer order by min_date desc) 
    from (select *, min(date) over (partition by customer, price) min_date from Invoice) a
) b where rn <= 2 and Date = min_date order by [index]