Summarise null rows in Oracle - sql

Question

I have a dataset like this:
+---------------+-------+
| SAMPLE_NUMBER | SCORE |
+---------------+-------+
| 1 | 100 |
| 2 | 97 |
| 3 | 124 |
| 4 | 762 |
| 5 | 999 |
| 6 | 1200 |
| 7 | NULL |
| 8 | NULL |
| 9 | NULL |
| 10 | NULL |
+---------------+-------+
I want to be able to summarise the NULL rows instead of displaying them all. So ideally, I would want the above to look like this:
+---------------+-------+
| SAMPLE_NUMBER | SCORE |
+---------------+-------+
| 1 | 100 |
| 2 | 97 |
| 3 | 124 |
| 4 | 762 |
| 5 | 999 |
| 6 | 1200 |
| 7-10 | NULL |
+---------------+-------+
Is there any way in Oracle to do this? Or is it something I will have to do post-query?

Answer 1

Yes. For your sample data:
select (case when score is null then min(sample_number) || '-' || max(sample_number)
else min(sample_number)
end) as sample_number,
score
from table t
group by score
order by min(id)
In other words, group by score and then fiddle with the sample number. Note: this assumes that you do not have duplicate scores. If you do, you can do so with a more complicated version:
select (case when score is null then min(sample_number) || '-' || max(sample_number)
else min(sample_number)
end) as sample_number,
score
from (select t.*,
row_number() over (partition by score order by sample_number) as seqnum
from table t
) t
group by score, (case when score is not null then seqnum end);

Answer 2

My guess is that this should be part of your presentation layer since you will have to cast sample_number to a string (assuming it is a numeric type. An alternative to your requirements is to return the min and max consecutive sample_number:
with t (SAMPLE_NUMBER, SCORE) as (
values (1, 100)
, (2, 97)
, (3, 124)
, (4, 762)
, (5, 999)
, (6, 1200)
, (7, NULL)
, (8, NULL)
, (9, NULL)
, (10, NULL)
)
select min(sample_number), max(sample_number), grp, score
from (
select SAMPLE_NUMBER, SCORE
, row_number() over (order by SAMPLE_NUMBER)
- row_number() over (partition by SCORE
order by SAMPLE_NUMBER) as grp
from t
) group by grp, score
order by grp;
1 2 GRP SCORE
----------- ----------- -------------------- -----------
1 1 0 100
2 2 1 97
3 3 2 124
4 4 3 762
5 5 4 999
6 6 5 1200
7 10 6 -
Tried against db2, so you may have to adjust it slightly.
Edit: treat rows as individuals when score is not null
with t (SAMPLE_NUMBER, SCORE) as (
values (1, 100)
, (2, 97)
, (3, 97)
, (4, 762)
, (5, 999)
, (6, 1200)
, (7, NULL)
, (8, NULL)
, (9, NULL)
, (10, NULL)
)
select min(sample_number), max(sample_number), grp, score
from (
select SAMPLE_NUMBER, SCORE
, row_number() over (order by SAMPLE_NUMBER)
- row_number() over (partition by SCORE
order by SAMPLE_NUMBER) as grp
from t
) group by grp, score
, case when score is not null then sample_number end
order by grp;
1 2 GRP SCORE
----------- ----------- -------------------- -----------
1 1 0 100
2 2 1 97
3 3 1 97
4 4 3 762
5 5 4 999
6 6 5 1200
7 10 6 -
You may want to map max to null in case it is the same as min:
[...]
select min(sample_number)
, nullif(max(sample_number), min(sample_number))
, grp
, score
from ...
1 2 GRP SCORE
----------- ----------- -------------------- -----------
1 - 0 100
2 - 1 97
3 - 1 97
4 - 3 762
5 - 4 999
6 - 5 1200
7 10 6 -

Answer 3

SELECT DISTINCT
DECODE(SCORE
,NULL
,(SELECT COUNT()+1
FROM TAB_NAME
WHERE SCORE IS NOT NULL)
|| '-'
|| (SELECT COUNT()
FROM TAB_NAME)
,SAMPLE_NUMBER) NUM
, NVL(TO_CHAR(SCORE),'NULL') SCRE
FROM TAB_NAME
ORDER BY 1 ASC;