问题
I have a table in the ORACLE database, details below:
--------------------------------------------
| FRUITS |
--------------------------------------------
| FRUIT_NAME | GROWTH_TIME | GROWTH_PLACE |
--------------------------------------------
| melon | 0600 | shelf1 |
| melon | 0630 | shelf1 |
| melon | 0700 | shelf1 |
| melon | 0730 | shelf1 |
| melon | 0800 | shelf1 |
| orange | 0600 | shelf5 |
| orange | 0630 | shelf5 |
| orange | 0700 | shelf5 |
| orange | 0730 | shelf5 |
| orange | 0800 | shelf5 |
| orange | 0830 | shelf5 |
| orange | 0900 | shelf5 |
| orange | 0930 | shelf5 |
| orange | 1000 | shelf5 |
| orange | 1200 | shelf5 |
| orange | 1230 | shelf5 |
| orange | 1300 | shelf5 |
| orange | 1330 | shelf5 |
| orange | 1400 | shelf5 |
| apple | 0600 | shelf3 |
| apple | 0630 | shelf3 |
| apple | 0700 | shelf3 |
| apple | 0730 | shelf3 |
| apple | 0800 | shelf3 |
--------------------------------------------and I would like to get results like these below:
--------------------------------------------
| FRUIT_NAME | GROWTH_TIME | GROWTH_PLACE |
--------------------------------------------
| melon | 0600-0800 | shelf1 |
| orange | 0600-1000 | shelf5 |
| orange | 1200-1400 | shelf5 |
| apple | 0600-0800 | shelf3 |
or like these:
-------------------------------------------------------------------
| FRUIT_NAME | GROWTH_START_TIME | GROWTH_END_TIME | GROWTH_PLACE |
-------------------------------------------------------------------
| melon | 0600 | 0800 | shelf1 |
| orange | 0600 | 1000 | shelf5 |
| orange | 1200 | 1400 | shelf5 |
| apple | 0600 | 0800 | shelf3 |
There is a small gap in the ORANGE case (between 1000 and 1400) and this is still the same shelf but with a small gap in time. It happens but I don't know how to solve this problem.
回答1:
you can solve this by analytics:
SQL> select fruit_name, min(growth_time) || '-' || max(growth_time) growth_time, growth_place
2 from (select fruit_name, growth_place, growth_time,
3 max(grp) over(partition by fruit_name, growth_place order by growth_time) grp
4 from (select fruit_name, growth_time, growth_place,
5 case
6 when to_date(lag(growth_time, 1)
7 over(partition by fruit_name, growth_place order by growth_time), 'hh24mi')
8 < to_date(growth_time, 'hh24mi') - (30/1440)
9 then
10 row_number() over(partition by fruit_name, growth_place order by growth_time)
11 when row_number() over(partition by fruit_name, growth_place order by growth_time) = 1
12 then
13 1
14 end grp
15 from fruits))
16 group by fruit_name, growth_place, grp
17 order by fruit_name, growth_time
18 /
FRUIT_ GROWTH_TIME GROWTH
------ ---------------------------------------- ------
apple 0600-0800 shelf3
melon 0600-0800 shelf1
orange 0600-1000 shelf5
orange 1200-1400 shelf5
i.e. first we break the result set into groups where a group is defined as contigious dates for a given fruut/shelf.
We do this by checking the prior date and seeing if its < the current rows date - 30 minutes with
lag(growth_time, 1) over (partition by fruit_name, growth_place order by growth_time)
from this we can derive groups where the prior row was over 30 minutes older that this row:
SQL> select fruit_name, growth_time, growth_place,
2 case
3 when to_date(lag(growth_time, 1)
4 over(partition by fruit_name, growth_place order by growth_time), 'hh24mi')
5 < to_date(growth_time, 'hh24mi') - (30/1440)
6 then
7 row_number() over(partition by fruit_name, growth_place order by growth_time)
8 when row_number() over(partition by fruit_name, growth_place order by growth_time) = 1
9 then
10 1
11 end grp
12 from fruits;
FRUIT_ GROW GROWTH GRP
------ ---- ------ ----------
apple 0600 shelf3 1
apple 0630 shelf3
apple 0700 shelf3
apple 0730 shelf3
apple 0800 shelf3
melon 0600 shelf1 1
melon 0630 shelf1
melon 0700 shelf1
melon 0730 shelf1
melon 0800 shelf1
orange 0600 shelf5 1
orange 0630 shelf5
orange 0700 shelf5
orange 0730 shelf5
orange 0800 shelf5
orange 0830 shelf5
orange 0900 shelf5
orange 0930 shelf5
orange 1000 shelf5
orange 1200 shelf5 10
orange 1230 shelf5
orange 1300 shelf5
orange 1330 shelf5
orange 1400 shelf5
now we just assign the group to each row with a max() analytic:
SQL> select fruit_name, growth_place, growth_time,
2 max(grp) over(partition by fruit_name, growth_place order by growth_time) grp
3 from (select fruit_name, growth_time, growth_place,
4 case
5 when to_date(lag(growth_time, 1)
6 over(partition by fruit_name, growth_place order by growth_time), 'hh24mi')
7 < to_date(growth_time, 'hh24mi') - (30/1440)
8 then
9 row_number() over(partition by fruit_name, growth_place order by growth_time)
10 when row_number() over(partition by fruit_name, growth_place order by growth_time) = 1
11 then
12 1
13 end grp
14 from fruits);
FRUIT_ GROWTH GROW GRP
------ ------ ---- ----------
apple shelf3 0600 1
apple shelf3 0630 1
apple shelf3 0700 1
apple shelf3 0730 1
apple shelf3 0800 1
melon shelf1 0600 1
melon shelf1 0630 1
melon shelf1 0700 1
melon shelf1 0730 1
melon shelf1 0800 1
orange shelf5 0600 1
orange shelf5 0630 1
orange shelf5 0700 1
orange shelf5 0730 1
orange shelf5 0800 1
orange shelf5 0830 1
orange shelf5 0900 1
orange shelf5 0930 1
orange shelf5 1000 1
orange shelf5 1200 10
orange shelf5 1230 10
orange shelf5 1300 10
orange shelf5 1330 10
orange shelf5 1400 10
now all that's left was the final group by on the GRP to get the final answer.
回答2:
You can use the LEAD function, using the partition by FRUIT_NAME ordered by GROWTH_TIME to get the GROWTH_TIME of the next record, comparing to the current one you can know if there is a gap.
SELECT FRUIT_NAME , MIN(GROWTH_TIME ) || '-' || MAX(GROWTH_TIME ), GROWTH FROM (
SELECT FRUIT_NAME , GROWTH_TIME ,
NVL(lead (GROWTH_TIME ) over (partition by FRUIT_NAME order by GROWTH_TIME ) - GROWTH_TIME , 0) as gap
FROM FRUITS
)
GROUP BY FRUIT_NAME, gap
HAVING (gap <= 70)
来源：https://stackoverflow.com/questions/14521261/oracle-select-group-by-something-that-i-dont-know