Question

我有一份工作，我必须在活跃期间选择记录

我的初始数据样本是：

 department_id,  employee_id,  start_dt,     end_dt
       1            11         2016-01-01    2016-01-03
       1            11         2016-01-04    2016-01-07
       1            11         2016-01-08    2016-01-11
       1            12         2016-01-12    2016-01-14
       1            12         2016-01-15    2016-01-17
       1            12         2016-01-18    2016-01-20
       1            11         2016-01-21    2016-01-24
       1            11         2016-01-25    2016-01-25
       1            14         2016-01-26    2016-01-27
       2            11         2016-04-01    2016-04-10

我的预期输出：

 department_id,  employee_id,  start_dt,     end_dt
       1            11         2016-01-01    2016-01-11
       1            12         2016-01-12    2016-01-20
       1            11         2016-01-21    2016-01-25
       1            14         2016-01-26    2016-01-27
       2            11         2016-04-01    2016-04-10

我尝试使用max/min and partition by，但employee_id可能会在不同时间在department_id内重复

Answer 1

以下是使用LAG解析函数的一种方法，该问题通常称为群组和岛屿问题

WITH cte 
     AS (SELECT department_id, 
                employee_id, 
                start_dt, 
                end_dt, 
                Sum(CASE WHEN pstart = employee_id THEN 0 ELSE 1 END) 
                  OVER(partition BY department_id ORDER BY start_dt) AS Counter 
         FROM   (SELECT department_id, 
                        employee_id, 
                        start_dt, 
                        end_dt, 
                        Lag(employee_id, 1, NULL) OVER( partition BY department_id 
                            ORDER BY start_dt) AS pstart 
                 FROM   #Table1 ) t) 
SELECT department_id, 
       employee_id, 
       min(start_dt) as start_dt, 
       max(end_dt) as  end_dt
FROM   (SELECT row_number() OVER(partition BY department_id, counter 
                   ORDER BY start_dt) AS rn, 
               department_id, employee_id, start_dt, end_dt ,counter
        FROM   cte) a 
group by department_id,
         employee_id,
         counter

Answer 2

您需要将活动时段链接在一起。一种方法是确定一个时期何时开始 - 并为此创建一个标志。累积总和然后标识一组活动。其余的只是聚合：

select department_id, employee_id, min(start_dt) as start_dt,
       max(end_dt) as end_t
from (select t.*,
             sum(IsGroupStart) over (partition by department_id, employee_id order by start_dt) as grp
      from (select t.*,
                   (case when exists (select 1
                                      from t t2
                                      where t2.department_id = t.department_id and  
                                            t2.employee_id = t.employee_id and
                                            t.start_dt between t2.start_dt and t2.end_dt + 1
                                     )
                         then 0 else 1
                    end) as IsGroupStart
employee_id
            from t
           ) t
     ) t
group by department_id, employee_id, grp;

选择历史表的有效期

2 个答案: