Sql server SQL Server按值分组=不带光标循环的上一个值

我必须编写一个T-SQL来获取超速行（Max/Min/Avg），但条件是，如果任何行 例如，这是我的数据库表，它有标志，如果这一行是超速或没有

rownum  Time                overspeed   overspeedrowid
1   2015-12-06 06:18:46.283 0           0
2   2015-12-06 06:19:46.283 1           0
3   2015-12-06 06:20:46.283 1           0
4   2015-12-06 06:21:46.283 0           0
5   2015-12-06 06:22:46.283 0           0
6   2015-12-06 06:23:46.283 1           0
7   2015-12-06 06:24:46.283 1           0
8   2015-12-06 06:25:46.283 1           0
9   2015-12-06 06:26:46.283 1           0
10  2015-12-06 06:27:46.283 0           0
11  2015-12-06 06:28:46.283 0           0
12  2015-12-06 06:29:46.283 0           0
13  2015-12-06 06:30:46.283 1           0
14  2015-12-06 06:31:46.283 1           0
15  2015-12-06 06:32:46.283 1           0
16  2015-12-06 06:33:46.283 0           0
17  2015-12-06 06:34:46.283 0           0
18  2015-12-06 06:35:46.283 1           0
19  2015-12-06 06:36:46.283 0           0
20  2015-12-06 06:37:46.283 0           0
21  2015-12-06 06:38:46.283 0           0
22  2015-12-06 06:39:46.283 1           0
23  2015-12-06 06:40:46.283 1           0

我的计划是使用以下工具

在行上循环，如果当前！=prev和prev=0，然后是新的过速宽值

如果电流=prev且prev=1，则电流过速宽=prev过速宽

通过超速从计算表最小/最大组中选择

所以计算表必须是这样的

rownum  Time                overspeed   overspeedrowid
1   2015-12-06 06:18:46.283 0           0
2   2015-12-06 06:19:46.283 1           1
3   2015-12-06 06:20:46.283 1           1
4   2015-12-06 06:21:46.283 0           0
5   2015-12-06 06:22:46.283 0           0
6   2015-12-06 06:23:46.283 1           2
7   2015-12-06 06:24:46.283 1           2
8   2015-12-06 06:25:46.283 1           2
9   2015-12-06 06:26:46.283 1           2
10  2015-12-06 06:27:46.283 0           0
11  2015-12-06 06:28:46.283 0           0
12  2015-12-06 06:29:46.283 0           0
13  2015-12-06 06:30:46.283 1           3
14  2015-12-06 06:31:46.283 1           3
15  2015-12-06 06:32:46.283 1           3
16  2015-12-06 06:33:46.283 0           0
17  2015-12-06 06:34:46.283 0           0
18  2015-12-06 06:35:46.283 1           4
19  2015-12-06 06:36:46.283 0           0
20  2015-12-06 06:37:46.283 0           0
21  2015-12-06 06:38:46.283 0           0
22  2015-12-06 06:39:46.283 1           5
23  2015-12-06 06:40:46.283 1           5

• 我尝试了一个游标，但它需要很多时间，因为生产数据大约是2000000行

• 我尝试用join row-1更新，并将row-1与当前的fast进行比较，因为它只更新了第二行

这是我所做的，但它只更新了第二行

update #temp1 
set #temp1.overspeedrowid = @a, @a = @a + 1 
from #temp1 
left join #temp1 prev on prev.rownum = #temp1.rownum - 1 
where 
     (#temp1.overspeed != prev.overspeed and #temp1.overspeed = 1)

update #temp1 
set #temp1.overspeedrowid = prev. overspeedrowid 
from #temp1 
left join #temp1 prev on prev.rownum = #temp1.rownum - 1 
where 
     (#temp1.overspeed = prev.overspeed and #temp1.overspeed = 1)

---

在SQL Server 2012+中，可以使用

lag（）

后跟“条件”累积和：

select t.*,
       (case when overspeed = 0 then 0
             else sum(OverSpeedFlag) over (order by time)
        end) as overspeedrowid
from (select t.*,
             (case when lag(overspeed) over (order by time) = 0 and overspeed = 1
                   then 1 else 0 
              end) as OverSpeedFlag
      from #temp1 t
     ) t;

---

提示：正如Gordon Linoff的回答所示，使用适当的软件（MySQL、Oracle、DB2等）和版本（例如，

sql-server-2014

）标记数据库问题是很有帮助的。语法和特征的差异通常会影响答案。