SQL 按一定数量分组

Question

我有一个订单表，其中包含客户 ID 和订单金额。我想加入这些订单，但加入的订单不能超过一定数量。下面是一个例子:

假设最大数量是 33 个托盘，我有一个这样的表:

Order ID    Client ID   Amount  
1           100001        10    
2           100001        22    
3           100001        13    
4           100001        33    
5           100001        1    
6           100001        5     
7           100001        6    

结果应该是:

Order ID    Client ID   Amount  Joined ID   Joined Amount
1            100001       10    100001A         32
2            100001       22    100001A         32
3            100001       13    100001B         13
4            100001       33    100001C         33
5            100001       1     100001D         12
6            100001       5     100001D         12
7            100001       6     100001D         12

这里，如果我们也能想出一种方法，将编号为5、6、7的订单加入到订单10001B中，那就太好了。但即使是这个解决方案也足够了。

我对如何解决这个问题有一些想法，但我无法真正想出一个可行的解决方案。我将像这样处理大约 2000 个订单 ID，所以我也不希望这是一个缓慢的操作。我正在使用 SQL Server 2014

Answer 1

您可以在这里借助递归 CTE 找到建议的解决方案(sql 定义):http://sqlfiddle.com/#!6/285c16/45

基本上 CTE 迭代有序列表(按 clientID、orderID)并评估总金额是否不超过 33。

我已将下一个 clientID 添加到模拟数据，以测试正确的子计数标准评估。

这里是获取结果的查询:

-- prepare numbering for iteration
with orders_nr
as 
(
  select row_number() over(order by clientID, id) as [nr],
     o.*
  from orders o

)

,
-- prepare sum totals
re
as
(
  select  id, amount, amount as amount_total ,o.[nr] as nr,
    clientID
  from orders_nr o
  where o.[nr]=1
  UNION ALL
  select o.id, o.amount, 
    CASE WHEN o.clientID <> r.clientID then o.amount
    ELSE o.amount+ r.amount_total END,
    o.[nr] as nr, o.clientID
  from orders_nr o  join re r
    on (o.[nr]=r.[nr]+1)

) 
,

 -- iterate total - evaluate current criteria (<=33)
 re2 as
 (
   select re.id, re.amount, re.amount_total, 
     re.[nr] as [group], re.[nr], re.clientID
   from re
   where re.[nr]=1

   UNION ALL

   select r.id, r.amount, 
     CASE WHEN r.amount+re2.amount_total >33
       OR r.clientID<>re2.clientID
         then r.amount ELSE re2.amount_total+r.amount END
       as amount_total,
   CASE WHEN r.amount+re2.amount_total >33 
     OR r.clientID<>re2.clientID THEN 
     r.[nr] ELSE re2.[group] END as [group], r.[nr], r.clientID
   from re r join re2
     on (r.[nr]=re2.[nr]+1 )

  )
  , group_total
  AS
  (
    select [group], clientID, max(amount_total) as total
    FROM re2
    group by [group], clientID
  ),


  result
  as
  (
      select 
   r.id,  r.clientID, r.amount, 
    cast(r.clientid as varchar(20))
      +'-'+char(64+cast(
        dense_rank() 
          over( partition by r.clientID 
               order by r.[clientID], r.[group])
       as varchar(3))) as joinedID
    ,     gt.total as joinedAmount
 from re2 as r join group_total gt
  on (r.clientID=gt.clientID AND r.[group]=gt.[group])
    )
    select * from result