基于上一个值计算运行计数器的SQL查询

我有一个表，有vDateTime，Category列。 我想创建一个列，根据以前的类别值递增计数。i、 e如果当前类别与前一个计数器相同，则计数器将递增，否则将重置为1。 我使用的是密集排列，但如果前一个值与当前值不同，如何重置该值

SELECT vDateTime, Category,
DENSE_RANK()over(Partition by Category  order by vDateTime, Category) as Rank
FROM tblA 
ORDER BY vDateTime, Category

---

这是一个缺口和孤岛问题。我建议行数的差异：

select a.*,
       row_number() over (partition by category, (seqnum_2 - seqnum) order by vDateTime) as expected_output
from (select a.*,
             row_number() over (order by vDateTime) as seqnum,
             row_number() over (partition by category order by vDateTime) as seqnum_2,
      from tbla 
     ) a
order by vDateTime, Category;

---

您可以尝试使用窗口函数LAG和SUM的另一种可能方法：

表:

CREATE TABLE Data (
   vDateTime datetime, 
   Category varchar(10)
)
INSERT INTO Data 
   (vDateTime, Category)
VALUES   
   ('2010-04-27T16:12:00', 'KCLK'), 
   ('2010-06-15T17:40:00', 'KCLE'), 
   ('2010-07-12T10:29:00', 'KCLK'), 
   ('2010-08-13T10:41:00', 'KCLK'), 
   ('2010-08-13T11:33:00', 'KCLE'), 
   ('2010-08-17T15:37:00', 'KCLE'), 
   ('2010-09-01T11:10:00', 'KCLE'), 
   ('2010-09-17T10:37:00', 'KCLE'), 
   ('2010-09-21T12:22:00', 'KCLE'),
   ('2010-09-27T15:38:00', 'KCLE'), 
   ('2010-09-28T14:11:00', 'KXAMC'), 
   ('2010-10-08T11:18:00', 'KCLK'), 
   ('2010-10-08T15:45:00', 'KCLE')

声明：

;WITH ChangesCTE AS (   
   SELECT 
      vDateTime, category, 
      CASE WHEN LAG(Category) OVER (ORDER BY vDateTime) = Category THEN 0 ELSE 1 END AS [Change]
   FROM Data   
), GroupsCTE AS (
   SELECT 
      vDateTime, category, 
      SUM([Change]) OVER (ORDER BY vDateTime) AS [Group]
   FROM ChangesCTE
)   
SELECT 
   vDateTime, category, 
   ROW_NUMBER() OVER (PARTITION BY [Group] ORDER BY vDateTime) AS rn
FROM GroupsCTE

一种选择是使用滞后分析函数一次，然后对分析函数求和两次

with tblB as
(
select Category, vDateTime, 
       lag(Category) over (order by vDateTime) as lg
  from tblA
), tblC as
(
select Category, vDateTime,
       case 
       when category != lg then 
         1
       else
         sum(case when category = lg then 0 else 1 end) over ( order by vDateTime )
       end as Rank   
  from tblB
)
select Category, vDateTime, 
       sum(case when rank = 1 then 0 else 1 end) over 
          ( partition by rank order by vDateTime ) + 1 as Rank 
  from tblC
 order by vDateTime 

---

请以text.4/27/2010 16:12 KCLK 1 6/15/2010 17:40 KCL1 7/12/2010 10:29 KCLK 2 8/13/2010 10:41 KCLK 3 8/13/2010 11:33 KCL2 8/17/2010 15:37 KCL3 9/1/2010 11:10 KCL4 9/17/2010 10:37 KCL5 9/21/2010 12:22 KCL6 9/27/2010 15:38 KCL7 9/28/2010 14:11 KXAMC 1 10/8/2010 11:18 KCLK 4 10/8:45的形式发布样本数据