Convert rows to columns like pivot - oracle

Question

I have a data like:
formid formownerid approverid
1 100 102
1 100 103
1 100 104
2 200 107
2 200 103
2 200 109
2 200 105
3 400 201
3 400 210
I want to convert it to:
formid formownerid approverid approverid approverid approverid
1 100 102 103 104 NULL
2 200 107 103 109 105
3 400 201 202 NULL NULL
Wherever I looked at I saw pivot/unpivot but it looks unrelated since we don't need aggregation.

Answer 1

The aggregate is a necessary part of the pivot, but it's simple to apply here; you don't want sum but a max aggregate works fine:
select *
from (
select t.*,
row_number() over (partition by formid, formownerid
order by approverid) as rn
from t42 t
)
pivot (max(approverid) as approverid for (rn) in (1, 2, 3, 4));
FORMID FORMOWNERID 1_APPROVERID 2_APPROVERID 3_APPROVERID 4_APPROVERID
---------- ----------- ------------ ------------ ------------ ------------
1 100 102 103 104
2 200 103 105 107 109
3 400 201 210
Or you can specify the column name prefix explicitly to make them valid identifiers:
pivot (max(approverid) as approverid
for (rn) in (1 as a, 2 as b, 3 as c, 4 as d));
The inner query is adding a pseudocolumn rn to the table results to give you a fixed value to pivot against, since the actual approver IDs aren't going to be known in advance.
The manual approach might make this a bit clearer:
select formid, formownerid,
max(case when rn = 1 then approverid end) as approverid_1,
max(case when rn = 2 then approverid end) as approverid_2,
max(case when rn = 3 then approverid end) as approverid_3,
max(case when rn = 4 then approverid end) as approverid_4
from (
select t.*,
row_number() over (partition by formid, formownerid
order by approverid) as rn
from t42 t
)
group by formid, formownerid
order by formid, formownerid;
FORMID FORMOWNERID APPROVERID_1 APPROVERID_2 APPROVERID_3 APPROVERID_4
---------- ----------- ------------ ------------ ------------ ------------
1 100 102 103 104
2 200 103 105 107 109
3 400 201 210
The inner query is the same. The case statement produces each column as above, but without the max and grouping you get multiple rows with lots of extra blanks:
select formid, formownerid,
case when rn = 1 then approverid end as approverid_1,
case when rn = 2 then approverid end as approverid_2,
case when rn = 3 then approverid end as approverid_3,
case when rn = 4 then approverid end as approverid_4
from (
select t.*,
row_number() over (partition by formid, formownerid
order by approverid) as rn
from t42 t
);
FORMID FORMOWNERID APPROVERID_1 APPROVERID_2 APPROVERID_3 APPROVERID_4
---------- ----------- ------------ ------------ ------------ ------------
1 100 102
1 100 103
1 100 104
2 200 103
2 200 105
2 200 107
2 200 109
3 400 201
3 400 210
Notice that there's only a value in (at most) one column for each formid/formownerid combination, but that they appear in different rows. The max suppresses those multiple rows; and the pivot version does something similar under the hood.
SQL Fiddle showing the manual approach with the intermediate step, and the pivot version.

Answer 2

One possible Approch:
SELECT FROMID, FROMOWNERID, APPROVERID, NULL APPROVERID, NULL APPROVERID, NULL APPROVERID
FROM yourtable
WHERE FROMID = 100
AND APPROVERID = 102
UNION ALL
SELECT FROMID, FROMOWNERID, NULL APPROVERID, APPROVERID APPROVERID, NULL APPROVERID, NULL APPROVERID
FROM yourtable
WHERE FROMID = 100
AND APPROVERID = 103
UNION ALL
SELECT FROMID, FROMOWNERID, NULL APPROVERID, NULL APPROVERID, APPROVERID APPROVERID, NULL APPROVERID
FROM yourtable
WHERE FROMID = 100
AND APPROVERID = 104
----
-------
And So On