如何在 SQL Server 中进行增量计数

Question

我被分配了一个问题，我不确定如何在 SQL Server（版本 5/8）中解决。这是简单的查询及其输出：

Select location, date_time, Item_sold
From Product

Location        Date_time                   Item_sold
VA            12/10/2010 1:30:00 PM           Candy
VA            12/10/2010 3:30:00 PM           Chips
VA            12/13/2010 12:50:00 AM          Wine
DC            12/13/2010 8:00:00 AM           Gum
DC            12/13/2010 12:30:00 PM          Bags
DC            12/13/2010 1:16:00 PM           Cheese
DC            12/13/2010 12:00:00 AM          Hotdog
NJ            12/14/2010 12:00:00 AM          Coffee
NJ            12/14/2010 1:15:00 PM           Beers
NJ            12/14/2010 12:00:00 AM          Coffee
NJ            12/14/2010 1:45:00 PM           Water

这是我想要的输出，我猜一个 while/for 循环或一个 pivot 函数可以完成这项工作，但我的经验还没有。基本上，我需要从 Item_sold 列增量计算已售出的商品数量（基线日期从 12/8 到 12/9、12/8 到 12/10、12/8 到 12/11、12/8 到12 月 12 日...）

Location    12/10 to 12/11  12/10 to 12/12  12/10 to 12/13  12/10 to 12/14
VA               2             2                  3              3
DC               0             0                  3              3 
NJ                 0             0                  0              4

我该如何解决这个问题？

Answer 1

示例表和数据

create table Product(Location char(2), Date_time datetime, Item_sold varchar(20))
insert Product select 'VA', '20101210 1:30:00 PM' ,'Candy'
insert Product select 'VA', '20101210 3:30:00 PM' ,'Chips'
insert Product select 'VA', '20101213 12:50:00 AM' ,'Wine'
insert Product select 'DC', '20101213 8:00:00 AM' ,'Gum'
insert Product select 'DC', '20101213 12:30:00 PM' ,'Bags'
insert Product select 'DC', '20101213 1:16:00 PM' ,'Cheese'
insert Product select 'DC', '20101213 12:00:00 AM' ,'Hotdog'
insert Product select 'NJ', '20101215 12:00:00 AM' ,'Coffee'
insert Product select 'NJ', '20101215 1:15:00 PM' ,'Beers'
insert Product select 'NJ', '20101215 3:45:00 AM' ,'Cream'

生成所需结果的 T-SQL

declare @start datetime
declare @end datetime
select @start = '20101208', @end = '20110105'

declare @sql nvarchar(max);
-- generate the column names
select @sql = coalesce(@sql + ',', '') + QuoteName(Convert(char(5),@start,101)+' - '+Convert(char(5),DT,101))
from (
    select @start + number DT
    from master..spt_values
    where type='P' and number between 0 and DATEDIFF(D,@start,@end)) T;

-- replace the column names into the generic PIVOT form
set @sql = REPLACE('
;with COUNTS AS (
    select p.location, Convert(char(5),@start,101)+'' - ''+Convert(char(5),@start + v.number,101) DT, X.C
    from 
    (
        Select distinct location From Product
        where Date_time >= @start and Date_time < @end+1 -- * the date after, to handle the times
    ) p
    inner join master..spt_values v on v.type=''P'' and v.number between 0 and DATEDIFF(D,@start,@end)
    cross apply
    (
        select COUNT(*) C from product p2
        where p2.Location=p.Location
        and p2.date_time >= @start and p2.date_time < @start + v.number +1
    ) X
)
select location, :columns:
from COUNTS p
pivot (max(C) for DT in (:columns:)) pv',
':columns:', @sql)

-- execute for the results
exec sp_executesql @sql, N'@start datetime,@end datetime', @start, @end

Answer 2

以上来自cyberkiwi，太复杂了。你需要做一些叫做“桌子旋转”的事情。假设你有一个这样的产品表：

create table product
(
  location  char(2)     not null ,
  date_time datetime    not null ,
  item_sold varchar(32) not null ,

  primary key clustered ( location , date_time ) ,
)

你想要的sql是下面的SELECT语句：

select Location       = loc.location  ,
       "12/8"         =  t8.sales_cnt ,
       "12/8 - 12/09" =  t9.sales_cnt ,
       "12/8 - 12/10" = t10.sales_cnt ,
       "12/8 - 12/11" = t11.sales_cnt ,
       "12/8 - 12/12" = t12.sales_cnt ,
       "12/8 - 12/13" = t13.sales_cnt
from      ( select distinct location from #product ) loc
left join ( select location , sales_cnt = count(*) from #product where date_time between '2011-12-08 00:00:00.000' and '2011-12-08 23:59:59.997' group by location ) t8  on t8.location  = loc.location
left join ( select location , sales_cnt = count(*) from #product where date_time between '2011-12-08 00:00:00.000' and '2011-12-09 23:59:59.997' group by location ) t9  on t9.location  = loc.location
left join ( select location , sales_cnt = count(*) from #product where date_time between '2011-12-08 00:00:00.000' and '2011-12-10 23:59:59.997' group by location ) t10 on t10.location = loc.location
left join ( select location , sales_cnt = count(*) from #product where date_time between '2011-12-08 00:00:00.000' and '2011-12-11 23:59:59.997' group by location ) t11 on t11.location = loc.location
left join ( select location , sales_cnt = count(*) from #product where date_time between '2011-12-08 00:00:00.000' and '2011-12-12 23:59:59.997' group by location ) t12 on t12.location = loc.location
left join ( select location , sales_cnt = count(*) from #product where date_time between '2011-12-08 00:00:00.000' and '2011-12-13 23:59:59.997' group by location ) t13 on t13.location = loc.location
order by 1