如果在同一天可以找到多个记录，请选择一个值

Question

有人建议我改进此子查询，以便能够根据以下计算获取单个记录，本帖子下面的子查询会带来以下结果：

PERCENT ¦ LOG_DATE ¦ APP ¦ REQ_ID 
55 ¦ 2017-02-07 15:44:22 ¦ HUO ¦ 253333 
63 ¦ 2017-02-08 10:42:18 ¦ CQS ¦ 265265 
75 ¦ 2017-02-08 06:55:12 ¦ CQS ¦ 265265 
84 ¦ 2017-02-09 08:35:42 ¦ CQS ¦ 265265 
40 ¦ 2017-02-09 09:45:14 ¦ PLK ¦ 277777 

我想看到的只是结果，它具有每日期的最新日期（al.AU_TIME）内的记录。我们的目标是努力寻找具有'63'百分比值的记录。查询应该带来：

55 ¦ 2017-02-07 15:44:22 ¦ HUO ¦ 253333 
63 ¦ 2017-02-08 10:42:18 ¦ CQS ¦ 265265 
84 ¦ 2017-02-09 08:35:42 ¦ CQS ¦ 265265 
40 ¦ 2017-02-09 09:45:14 ¦ PLK ¦ 277777 

那么我应该怎么做，如果多个记录在同一天有相同的REQ_ID。

SELECT TO_NUMBER(RTRIM(ap.AP_NEW_VALUE,'%')) as PERCENT, 
    al.AU_TIME as LOG_DATE, 
    req.RQ_USER_03 as APP, 
    req.RQ_REQ_ID as REQ_ID 
FROM AUDIT_PROPERTIES ap, 
    AUDIT_LOG al, 
    REQ 
WHERE al.AU_ACTION_ID = ap.AP_ACTION_ID and 
     req.RQ_REQ_ID = al.AU_ENTITY_ID and 
     req.rq_req_date BETWEEN TO_DATE('05/02/2017','dd/mm/yyyy') AND TO_DATE('20/02/2017','dd/mm/yyyy') and 
     ap.AP_FIELD_NAME = 'RQ_USER_58' 

在此先感谢。

Answer 1

这可以在单个聚合查询中完成（无子查询）。按req_id和trunc(log_date)分组，然后为每个组选择一条记录。 req_id和max(log_date)（每个组内的，即每一天内）是显而易见的。然后使用first/last功能（keep (dense_rank...)。你需要照顾的只是一点点，如果有可能有完全相同的日期和时间（精确到秒），对于一个给定req_id两行。

下面是做这件事，我把你的测试和演示目的的WITH条款提供的测试数据，但是这是查询的不一部分。

with 
    test_data (percent, log_date, app, req_id) as (
     select 55, to_date('2017-02-07 15:44:22', 'yyyy-mm-dd hh24:mi:ss'), 'HUO', 253333 
     from dual union all 
     select 63, to_date('2017-02-08 10:42:18', 'yyyy-mm-dd hh24:mi:ss'), 'CQS', 265265 
     from dual union all 
     select 75, to_date('2017-02-08 06:55:12', 'yyyy-mm-dd hh24:mi:ss'), 'CQS', 265265 
     from dual union all 
     select 84, to_date('2017-02-09 08:35:42', 'yyyy-mm-dd hh24:mi:ss'), 'CQS', 265265 
     from dual union all 
     select 40, to_date('2017-02-09 09:45:14', 'yyyy-mm-dd hh24:mi:ss'), 'PLK', 277777 
     from dual 
    ) 
-- End of test data (not part of the solution). SQL query begins BELOW THIS LINE 
select max(percent) keep (dense_rank last order by log_date)   as percent, 
     max(log_date)             as log_date, 
     max(app)  keep (dense_rank last order by log_date, percent) as app, 
     req_id 
from test_data 
group by req_id, trunc(log_date) 
order by log_date, req_id 
; 

PERCENT LOG_DATE APP REQ_ID 
------- ---------- --- ------- 
    55 2017-02-07 HUO 253333 
    63 2017-02-08 CQS 265265 
    84 2017-02-09 CQS 265265 
    40 2017-02-09 PLK 277777 

4 rows selected. 

Answer 2

这样做的典型方法是使用row_number()。最好的，我可以在你的情况下猜测：

select a.* 
from (select a.*, 
      row_number() over (partition by req_id order by log_date desc) as seqnum 
     from atable a 
    ) a 
where seqnum = 1; 

Answer 3

您可以通过编写内部查询或创建一个中间表，通过只是时间组的数据做计算每个日期 内部查询会看的MAX_TIME像这样：

SELECT cast(al.AU_TIME as date) as LOG_DATE, 
max(al.AU_TIME) as max_LOG_TIME 
FROM AUDIT_PROPERTIES ap, 
    AUDIT_LOG al, 
    REQ 
WHERE al.AU_ACTION_ID = ap.AP_ACTION_ID and 
     req.RQ_REQ_ID = al.AU_ENTITY_ID and 
     req.rq_req_date BETWEEN TO_DATE('05/02/2017','dd/mm/yyyy') AND TO_DATE('20/02/2017','dd/mm/yyyy') and 
     ap.AP_FIELD_NAME = 'RQ_USER_58' 
GROUP BY LOG_DATE 

再查询外层的需要与时间戳值这个内部查询结合，以获得只是这些记录

Select TO_NUMBER(RTRIM(ap.AP_NEW_VALUE,'%')) as PERCENT, 
     al.AU_TIME, 
     req.RQ_USER_03 as APP, 
    req.RQ_REQ_ID as REQ_ID 
    FROM AUDIT_PROPERTIES ap 
    inner join (
SELECT cast(al.AU_TIME as date) as LOG_DATE, 
max(al.AU_TIME) as max_LOG_TIME 
FROM AUDIT_PROPERTIES ap, 
    AUDIT_LOG al, 
    REQ 
WHERE al.AU_ACTION_ID = ap.AP_ACTION_ID and 
     req.RQ_REQ_ID = al.AU_ENTITY_ID and 
     req.rq_req_date BETWEEN TO_DATE('05/02/2017','dd/mm/yyyy') AND TO_DATE('20/02/2017','dd/mm/yyyy') and 
     ap.AP_FIELD_NAME = 'RQ_USER_58' 
GROUP BY LOG_DATE) 
AS X 
ON ap.AU_TIME= X.max_LOG_TIME 
INNER JOIN AUDIT_LOG al 
on al.AU_ACTION_ID = ap.AP_ACTION_ID 
INNER JOIN REQ 
ON req.RQ_REQ_ID = al.AU_ENTITY_ID 
WHERE req.rq_req_date BETWEEN TO_DATE('05/02/2017','dd/mm/yyyy') AND TO_DATE('20/02/2017','dd/mm/yyyy') and 
     ap.AP_FIELD_NAME = 'RQ_USER_58'