Question

需要根据字段中文本的出现对行进行分组。说，当你看到 “ Started_Process1”应该启动其他聚合组。

需要的输出

Prcs    Count   Min_Process_DT
Process1    2   1/15/2019 1:15
Process5    5   1/15/2019 1:17
Process1    3   1/15/2019 1:21
Process5    3   1/15/2019 1:30
Process1    4   1/15/2019 1:25

样本数据集

S_ID    Msg                 Process_DT  Stack_Trace
1   Started_Process1    1/15/2019 1:15  Something happened1
2   Ended_Process1      1/15/2019 1:16  Something happened2
3   Started_Process5    1/15/2019 1:17  Something happened3
4   InProgress_Process5 1/15/2019 1:18  Something happened4
5   InProgress_Process5 1/15/2019 1:19  Something happened5
6   InProgress_Process5 1/15/2019 1:20  Something happened6
7   Started_Process1    1/15/2019 1:21  Something happened7
8   Ended_Process5      1/15/2019 1:22  Something happened8
9   InProgress_Process1 1/15/2019 1:23  Something happened9
10  Ended_Process1      1/15/2019 1:24  Something happened10
11  Started_Process1    1/15/2019 1:25  Something happened11
12  InProgress_Process1 1/15/2019 1:26  Something happened12
13  InProgress_Process1 1/15/2019 1:27  Something happened13
14  InProgress_Process1 1/15/2019 1:28  Something happened14
16  Started_Process5    1/15/2019 1:30  Something happened16
17  InProgress_Process5 1/15/2019 1:31  Something happened17
18  Ended_Process5      1/15/2019 1:32  Something happened18

感谢您的帮助

Answer 1

这是解决问题的一种方法：

创建示例数据：

Declare @table table (S_ID int, Msg nvarchar(100), Process_DT datetime, Stack_Trace nvarchar(100))
    Insert into @table Select 1,   'Started_Process1'       ,'1/15/2019 1:15',  'Something happened1'
    Insert into @table Select 2 ,  'Ended_Process1'         ,'1/15/2019 1:16', 'Something happened2'
    Insert into @table Select 3 ,  'Started_Process5'       ,'1/15/2019 1:17', 'Something happened3'
    Insert into @table Select 4 ,  'InProgress_Process5'    ,'1/15/2019 1:18', 'Something happened4'
    Insert into @table Select 5 ,  'InProgress_Process5'    ,'1/15/2019 1:19', 'Something happened5'
    Insert into @table Select 6 ,  'InProgress_Process5'    ,'1/15/2019 1:20', 'Something happened6'
    Insert into @table Select 7 ,  'Started_Process1'       ,'1/15/2019 1:21', 'Something happened7'
    Insert into @table Select 8 ,  'Ended_Process5'         ,'1/15/2019 1:22', 'Something happened8'
    Insert into @table Select 9 ,  'InProgress_Process1'    ,'1/15/2019 1:23', 'Something happened9'
    Insert into @table Select 10,  'Ended_Process1'         ,'1/15/2019 1:24', 'Something happened10'
    Insert into @table Select 11,  'Started_Process1'       ,'1/15/2019 1:25', 'Something happened11'
    Insert into @table Select 12,  'InProgress_Process1'    ,'1/15/2019 1:26', 'Something happened12'
    Insert into @table Select 13,  'InProgress_Process1'    ,'1/15/2019 1:27', 'Something happened13'
    Insert into @table Select 14,  'InProgress_Process1'    ,'1/15/2019 1:28', 'Something happened14'
    Insert into @table Select 16,  'Started_Process5'       ,'1/15/2019 1:30', 'Something happened16'
    Insert into @table Select 17,  'InProgress_Process5'    ,'1/15/2019 1:31', 'Something happened17'
    Insert into @table Select 18,  'Ended_Process5'         ,'1/15/2019 1:32', 'Something happened18'

对于查询

Select    substring(t1.Msg, Charindex('_', t1.Msg, 1) + 1, len(t1.Msg) - Charindex('_', t1.Msg, 1)) [Prcs]
        , t2.count                                                                                  [Count]
        , t1.Process_DT                                                                             [Min_Process_DT]
        from    @table t1
               ,(select   ta.grp               [grp]
                         ,count(ta.Process_DT) [count]
                         ,min(ta.S_ID)         [S_ID]
                 from   (
                         select   sum(case when msg like 'Started_Process%' then 1 end) 
                                  over (order by substring(Msg, Charindex('_', Msg, 1) + 1, len(Msg) - Charindex('_', Msg, 1)), Process_DT) grp
                                 ,Process_DT
                                 ,S_ID
                                 ,substring(msg, charindex('_', Msg, 1) + 1, len(msg) - charindex('_', Msg, 1)) process
                         from    @table
                        ) ta
                group by ta.grp) t2
        where  Msg like ('Started%')
        and    t1.S_ID = t2.S_ID
        order by t1.Process_DT

Answer 2

您的数据存在的问题是流程可能重叠，但是我认为这是正确的逻辑：

select p, min(dt) dt, count(1) cnt
  from (
    select id, p, dt, sum(g) over (partition by p order by id) grp 
      from (
        select s_id id, substr(msg, instr(msg, 'Process')) p, msg, process_dt dt, 
               case when msg like 'Started%' then 1 end g 
          from data))
  group by p, grp
  order by dt

^{dbfiddle demo}

在不同组中聚合的SQL行每次文本均出现在日志中

2 个答案: