查询以显示事务表中的已用信用

Question

我正在使用一个包含信用交易的表格,我想要显示销售时花费的信用额度.

在表中:

• Credits由实体使用唯一实体代码添加(记录在列中GivenByUserCode)
• 信用额度总是有这样的代码.
• 花费的积分总是具有负值.
• 花费的信用将不具有实体代码(值为GivenByUserCodeis null).

如果用户2018-01-02在报表上进行购买,则使用上述数据作为示例,应显示源自的所有这些信用BM01.增加的复杂性在于购买可以分为多个添加,请参阅购买2018-02-03分为3个添加.

我认为解决方案将与使用cte和over有关,但我没有使用这些的经验.我确实在SqlServerCentral上发现了类似(不相同)的问题.

任何帮助/方向将是最受欢迎的.

---

输入和DDL

DECLARE @CreditLogs TABLE(CreditLogId int not null identity(1,1), Credits INT NOT NULL, OccurredOn DATETIME2(7) NOT NULL, GivenByUserCode VARCHAR(100) NULL)

INSERT INTO @CreditLogs (Credits, OccurredOn, GivenByUserCode) VALUES
  (10,  '2018-01-01', 'BM01')
, (10,  '2018-01-01', 'BM01')
, (-10, '2018-01-02', NULL)
, (-5,  '2018-01-04', NULL)
, (5,   '2018-02-01', 'SP99')
, (40,  '2018-02-02', 'BM02')
, (-40, '2018-02-03', NULL)
, (-4,  '2018-03-05', NULL)

以表格形式输入

CreditLogId | Credits | OccurredOn | GivenByUserCode
------------+---------+------------+----------------
          1 |      10 | 2018-01-01 |            BM01
          2 |      10 | 2018-01-01 |            BM01
          3 |     -10 | 2018-01-02 |            NULL
          4 |      -5 | 2018-01-04 |            NULL
          5 |       5 | 2018-02-01 |            SP99
          6 |      40 | 2018-02-02 |            BM02
          7 |     -40 | 2018-02-03 |            NULL
          8 |      -4 | 2018-03-05 |            NULL

预期产出

SELECT *
FROM (VALUES
     (3, '2018-01-02', 10, 'BM01')
    ,(4, '2018-01-04', 5, 'BM01')
    ,(7, '2018-02-03', 5, 'BM01')
    ,(7, '2018-02-03', 5, 'SP99')
    ,(7, '2018-02-03', 30, 'BM02')
    ,(8, '2018-03-05', 4, 'BM02')
) expectedOut (CreditLogId, OccurredOn, Credits, GivenByUserCode)

产生输出

CreditLogId | Occurred on | Credits | GivenByUserCode
------------+-------------+---------+----------------
          3 |  2018-01-02 |      10 |            BM01
          4 |  2018-01-04 |       5 |            BM01
          7 |  2018-02-03 |       5 |            BM01
          7 |  2018-02-03 |       5 |            SP99
          7 |  2018-02-03 |      30 |            BM02
          8 |  2018-03-05 |       4 |            BM02

代码到目前为止

它并不多,我不知道从哪里开始.

WITH totals AS (
    SELECT CreditLogId, OccurredOn, credits, sum(credits) OVER(ORDER BY OccurredOn) AS TotalSpent
    FROM @CreditLogs
    WHERE Credits < 0
)
SELECT *
FROM totals

---

补充说明

预期产出是用于这些信用来自的每个用过的信用额度.积分以先进先出(FIFO)为基础.这里解释了样本输出中的每个值,希望这可以澄清所需的输出.

• 对于10个信用点(信用日志ID 3)的消费可以追溯到信用日志ID 1的添加
• 对于5个信用点(信用日志ID 4)的支出可以追溯到信用日志ID 2的添加(因为信用日志ID 1被"用完")
• 对于信用记录中的40个信用额度,ID 7可以追溯到
  • 来自信用记录ID 2,5学分的额外剩余
  • 信用记录ID 5(添加5)
  • 信用记录ID 6(另外40个剩余10个)
• 对于信用记录8中4个信用的消费,使用信用记录id 6的余额

请注意,剩余6个学分的总余额,余额不必为零,但永远不会为负,因为用户只能花费他们所拥有的.

Answer 1

试试这个:

WITH Credits_added AS (
    SELECT CreditLogId, OccurredOn, credits
    , SUM(credits) OVER (ORDER BY CreditLogId) - credits AS b --before
    , SUM(credits) OVER (ORDER BY CreditLogId) AS a --after
    , GivenByUserCode
    FROM @CreditLogs
    WHERE Credits > 0)
, Credits_spent AS (
    SELECT CreditLogId, OccurredOn, credits
    , SUM(credits) OVER (ORDER BY CreditLogId) * -1 + credits AS b
    , SUM(credits) OVER (ORDER BY CreditLogId) * -1 AS a
    FROM @CreditLogs
    WHERE Credits < 0)
SELECT s.CreditLogId, s.OccurredOn
, CASE WHEN a.a > s.a THEN s.a ELSE a.a END - CASE WHEN a.b > s.b THEN a.b ELSE s.b END AS Credits 
, a.GivenByUserCode
FROM Credits_added AS a
INNER JOIN Credits_spent AS s ON a.a > s.b AND s.a > a.b

Answer 2

既然你提到每年会有400到500万条记录,即使这可以通过查询来完成,也会是一个缓慢的查询.

我建议有另一个creditSpent包含的表(PurchaseCreditLogId , additionCreditLogId, Amount)

在插入购买时,查找所有记录,计算应从每个记录中减少的金额,并将该信息存储在该表中

然后,当您运行报表时,可以对此表执行简单查询