数据结构类似
SQL> select * from t;
B E N
----------------- ----------------- --------------------
20150106 01:00:02 20150106 01:00:42 A
20150106 01:00:02 20150106 01:00:12 B
20150106 02:01:02 20150106 01:10:03 C
20150106 02:01:07 20150106 02:01:08 D
20150106 02:00:01 20150106 02:00:02 E
20150106 01:00:03 20150106 01:00:07 F
6 rows selected.
数据表定义
SQL> @desc t
Name Null? Type
------------------------------- -------- ----------------------------
1 B DATE
2 E DATE
3 N VARCHAR2(10)
需求抽象
以分钟为单位统计指定天每分钟内N的个数。
思路：
1、每天1440分钟，构造全天的分钟集合
SELECT to_date(‘20150106 00:00:00‘ , ‘yyyy-mm-dd hh24:mi:ss‘) +
numtodsinterval (level , ‘minute‘) dt
FROM dual
CONNECT BY level <= 24 * 60
2、左外连接，条件 dt >= begin and dt < end + 1
3、需要截取t.begin 和 t.end到分钟
对于begin , trunc(begin,‘mi‘)
对于end , trunc(end+ numtodsinterval(1, ‘minute‘),‘mi‘)) // trunc( end + 1 ,[format])
4、按分钟聚合统计
结果
WITH each_minute AS
( SELECT to_date( ‘20150106 00:00:00‘ , ‘yyyy-mm-dd hh24:mi:ss‘ ) +
numtodsinterval (level , ‘minute‘) dt
FROM dual
CONNECT BY level <= 24 * 60 )
SELECT trunc (em.dt, ‘mi‘) d, count (t.n) cnt
FROM each_minute em
LEFT OUTER JOIN t
ON (em.dt >= trunc (t.b, ‘mi‘) AND em.dt < trunc (t.e+ numtodsinterval (1 , ‘minute‘), ‘mi‘ ))
group by trunc(em.dt, ‘mi‘ )
order by 1;
简单测试
SQL> WITH each_minute AS
2 (SELECT to_date(‘20150106 00:00:00‘, ‘yyyy-mm-dd hh24:mi:ss‘) +
3 numtodsinterval(level, ‘minute‘) dt
4 FROM dual
5 CONNECT BY level <= 24 * 60)
6 SELECT trunc(em.dt, ‘mi‘) d, count(t.n) cnt
7 FROM each_minute em
8 LEFT OUTER JOIN t
9 ON (em.dt >= trunc(t.b, ‘mi‘) AND
10 em.dt < trunc(t.e + numtodsinterval(1, ‘minute‘), ‘mi‘))
11 where t.n is not null -- no need for 1440,just for test.
12 group by trunc(em.dt, ‘mi‘)
13 order by 1;
D CNT
----------------- ----------
20150106 01:00:00 3
20150106 02:00:00 1
20150106 02:01:00 1