3

我正在尝试将请求与随后的最早响应相匹配。不幸的是，UI 不会强制执行一个请求对一个响应。

要求：

ENC_ID  CONTACT EVENT_TIME
13991333    1   08/13/2012 22:00:36
13991333    1   08/13/2012 22:25:21
13991333    1   08/13/2012 22:26:14
13991333    1   08/13/2012 22:36:42
13991333    2   08/13/2012 22:29:40

回应：

ENC_ID  CONTACT EVENT_TIME
13991333    1   08/13/2012 22:25:40
13991333    1   08/13/2012 22:29:31
13991333    1   08/13/2012 22:30:47
13991333    1   08/13/2012 22:31:35
13991333    1   08/13/2012 22:31:49
13991333    2   08/13/2012 22:31:14

期望：

ENC_ID  CONTACT REQ_EVENT_TIME  RES_EVENT_TIME
13991333    1   08/13/2012 22:00:36 08/13/2012 22:25:40
13991333    1   08/13/2012 22:25:21 08/13/2012 22:29:31
13991333    1   08/13/2012 22:26:14 08/13/2012 22:30:47
13991333    1   08/13/2012 22:36:42 
13991333    2   08/13/2012 22:29:40 08/13/2012 22:31:14

规则：

  • 如果响应与请求相关联，则它不能与另一个相关联。
  • 请求没有响应是可以接受的，但反过来不行

我的查询创建了笛卡尔积：

SELECT  RQ.ENC_ID, RQ.CONTACT, RQ.EVENT_TIME REQ_EVENT_TIME,
        RS.EVENT_TIME RES_EVENT_TIME
FROM    RQ
LEFT OUTER JOIN RS ON RQ.ENC_ID=RS.ENC_ID
  AND RQ.CONTACT=RS.CONTACT
  AND RS.EVENT_TIME>RQ.EVENT_TIME

我一直在试验 LEAD() 函数，但无法让它按预期工作。

4

2 回答 2

1 
with t1(ENC_ID, CONTACT, EVENT_TIME) as(
  select 13991333,    1,   '08/13/2012 22:00:36' from dual union all
  select 13991333,    1,   '08/13/2012 22:25:21' from dual union all
  select 13991333,    1,   '08/13/2012 22:26:14' from dual union all
  select 13991333,    1,   '08/13/2012 22:36:42' from dual union all
  select 13991333,    2,   '08/13/2012 22:29:40' from dual
),
t2(ENC_ID,CONTACT,EVENT_TIME) as(
  select 13991333,    1,   '08/13/2012 22:25:40' from dual union all
  select 13991333,    1,   '08/13/2012 22:29:31' from dual union all
  select 13991333,    1,   '08/13/2012 22:30:47' from dual union all
  select 13991333,    1,   '08/13/2012 22:31:35' from dual union all
  select 13991333,    1,   '08/13/2012 22:31:49' from dual union all
  select 13991333,    2,   '08/13/2012 22:31:14' from dual
)
select enc_id
     , contact
     , req
     , min(res) resp
  from (select q.enc_id
             , q.contact
             , q.req
             , case  
                 when resp > req then nth_value(q.resp, q.gid) over(partition by q.enc_id, q.contact, gid order by q.req, q.gid)
               end res
         from (select t1.enc_id
                    , t1.contact
                    , t1.event_time req
                    , dense_rank() over (partition by t1.enc_id, t1.contact order by t1.event_time) gid
                    , t2.event_time resp
                 from t1
                 left outer join t2
                   on (t1.enc_id = t2.enc_id   and
                       t1.contact = t2.contact)
                order by 1,2,3,5
               ) q
         )
group by enc_id
       , contact
       ,  req
order by enc_id
       , contact
       ,  req

结果：

 ENC_ID    CONTACT     REQ                 RESP
---------- ----------  ------------------- -------------------
  13991333          1  08/13/2012 22:00:36 08/13/2012 22:25:40
  13991333          1  08/13/2012 22:25:21 08/13/2012 22:29:31
  13991333          1  08/13/2012 22:26:14 08/13/2012 22:30:47
  13991333          1  08/13/2012 22:36:42 
  13991333          2  08/13/2012 22:29:40 08/13/2012 22:31:14

我们在这里所做的是使用dense_rank分析函数索引请求组：

select t1.enc_id
     , t1.contact
     , to_date(t1.event_time, 'mm/dd/yyyy hh24:mi:ss') req
     , dense_rank() over (partition by t1.enc_id, t1.contact order by to_date(t1.event_time, 'mm/dd/yyyy hh24:mi:ss')) gid
     , to_date(t2.event_time, 'mm/dd/yyyy hh24:mi:ss') resp
  From t1
  left outer join t2
    on (t1.enc_id = t2.enc_id   and
        t1.contact = t2.contact)
 order by 1,2,3,5

结果：

  ENC_ID    CONTACT REQ                        GID RESP
---------- ---------- ------------------- ---------- -------------------
  13991333          1 08/13/2012 22:00:36          1 08/13/2012 22:25:40
  13991333          1 08/13/2012 22:00:36          1 08/13/2012 22:29:31
  13991333          1 08/13/2012 22:00:36          1 08/13/2012 22:30:47
  13991333          1 08/13/2012 22:00:36          1 08/13/2012 22:31:35
  13991333          1 08/13/2012 22:00:36          1 08/13/2012 22:31:49
  13991333          1 08/13/2012 22:25:21          2 08/13/2012 22:25:40
  13991333          1 08/13/2012 22:25:21          2 08/13/2012 22:29:31
  13991333          1 08/13/2012 22:25:21          2 08/13/2012 22:30:47
  13991333          1 08/13/2012 22:25:21          2 08/13/2012 22:31:35
  13991333          1 08/13/2012 22:25:21          2 08/13/2012 22:31:49
  13991333          1 08/13/2012 22:26:14          3 08/13/2012 22:25:40
  13991333          1 08/13/2012 22:26:14          3 08/13/2012 22:29:31
  13991333          1 08/13/2012 22:26:14          3 08/13/2012 22:30:47
  13991333          1 08/13/2012 22:26:14          3 08/13/2012 22:31:35
  13991333          1 08/13/2012 22:26:14          3 08/13/2012 22:31:49
  13991333          1 08/13/2012 22:36:42          4 08/13/2012 22:25:40
  13991333          1 08/13/2012 22:36:42          4 08/13/2012 22:29:31
  13991333          1 08/13/2012 22:36:42          4 08/13/2012 22:30:47
  13991333          1 08/13/2012 22:36:42          4 08/13/2012 22:31:35
  13991333          1 08/13/2012 22:36:42          4 08/13/2012 22:31:49
  13991333          2 08/13/2012 22:29:40          1 08/13/2012 22:31:14

然后根据请求组的索引选择第n个响应值。因此，对于第一组请求，它将是组中响应时间的第一个值，对于第二组请求，它将是组中响应时间的第二个值，依此类推。

更新（模拟 10g 的 nth_value 分析函数的行为）

SQL> with t1(ENC_ID, CONTACT, EVENT_TIME) as(
  2    select 13991333,    1,   '08/13/2012 22:00:36' from dual union all
  3    select 13991333,    1,   '08/13/2012 22:25:21' from dual union all
  4    select 13991333,    1,   '08/13/2012 22:26:14' from dual union all
  5    select 13991333,    1,   '08/13/2012 22:36:42' from dual union all
  6    select 13991333,    2,   '08/13/2012 22:29:40' from dual
  7  ),
  8  t2(ENC_ID,CONTACT,EVENT_TIME) as(
  9    select 13991333,    1,   '08/13/2012 22:25:40' from dual union all
 10    select 13991333,    1,   '08/13/2012 22:29:31' from dual union all
 11    select 13991333,    1,   '08/13/2012 22:30:47' from dual union all
 12    select 13991333,    1,   '08/13/2012 22:31:35' from dual union all
 13    select 13991333,    1,   '08/13/2012 22:31:49' from dual union all
 14    select 13991333,    2,   '08/13/2012 22:31:14' from dual
 15  ),
 16  t3 as
 17  (
 18  select z.*
 19       , row_number() over (partition by enc_id, contact, req order by req) rn
 20    from (
 21    select t1.enc_id
 22         , t1.contact
 23         , t1.event_time req
 24         , dense_rank() over (partition by t1.enc_id, t1.contact order by t1.event_time) gid
 25         , t2.event_time resp
 26      from t1
 27      left outer join t2
 28        on (t1.enc_id = t2.enc_id   and
 29            t1.contact = t2.contact)
 30      order by 1,2,3,5
 31    ) z
 32  )
 33  select enc_id , contact, req, min(resp) resp
 34    from (
 35           select q.enc_id
 36                , q.contact
 37                , q.req
 38                , case
 39                    when min(case rn when gid then resp end) over(partition by enc_id, contact, req order by req) > req
 40                    then min(case rn when gid then resp end) over(partition by enc_id, contact, req order by req)
 41                 end resp
 42            from t3 q
 43         )
 44  group by enc_id
 45         , contact
 46         , req
 47  order by enc_id
 48         , contact
 49         , req
 50  ;

Result:


    ENC_ID    CONTACT REQ                 RESP
---------- ---------- ------------------- -------------------
  13991333          1 08/13/2012 22:00:36 08/13/2012 22:25:40
  13991333          1 08/13/2012 22:25:21 08/13/2012 22:29:31
  13991333          1 08/13/2012 22:26:14 08/13/2012 22:30:47
  13991333          1 08/13/2012 22:36:42 
  13991333          2 08/13/2012 22:29:40 08/13/2012 22:31:14
于 2012-10-20T05:58:22.510 回答
0

你非常接近......试试这个：

SELECT  RQ.ENC_ID, RQ.CONTACT, RQ.EVENT_TIME REQ_EVENT_TIME,
        min(RS.EVENT_TIME) RES_EVENT_TIME
FROM    RQ
LEFT OUTER JOIN RS ON RQ.ENC_ID=RS.ENC_ID
  AND RQ.CONTACT=RS.CONTACT
  AND RS.EVENT_TIME>RQ.EVENT_TIME
GROUP BY RQ.ENC_ID, RQ.CONTACT, RQ.EVENT_TIME
ORDER BY 1,2,3,4
于 2012-10-19T20:13:23.503 回答