Selecting different columns depending on condition

Question

Hi there,
I have a requirement where I run a query and depending on one of the conditions I need a different set of columns returned, and each of the columns is in a different row. i.e.
table1
id key val
2  2_1 val2_1
2  2_2 val2_2
5  5_1 val5_1
5  5_2 val5_2
5  5_3 val5_3
5  5_4 val5_4Then I'd like something like this:
select val from table1
where if id = 2 then return val where key in ('2_1', '2_2') else
if id = 5 then return val where key in ('5_1', '5_2', '5_4') else
if id = ....The output for instance I'd expect if id=5 is:
val5_1, val5_2, val5_4I need to be able to manually put which values to return for each id. I don't necessarely need all values of an id.. for instance I left out val5_3 where id = 5.
Is this possible? I have like 20 ids, and it's either doing something like this or create 20 specific queries. Please let me know if more clarification is needed.
Thanks,

Answer 1

malky wrote:
Thanks Alberto, I've used pivot before as functions of "max(decode(rn,1,nm)) ", I didn't notice your PIVOT in there, but I cannot use it, i should have mentioned i'm using 10g.
That is why the FAQ SQL and PL/SQL FAQ are saying to always post your Oracle version as result of this query:
SELECT * FROM v$version;If you are not using 10g then you have to go to a pivoting pre-11g solution that are also mentioned in the FAQ.
The substr(key,3) in my case the Id not always has the same keys, so what if i remove your substr function? would it still work?
Also, you're specifying 4 columns, there might be more, does that mean i'll have to hardcode the max number of columns an id would have and use them in all of them? or can that be replaced by what I posted earlier on?
CASE WHEN id = 2 and key='2_1' THEN val END
That is exactly what you can do.
my only problem with this is that i cannot specify the "," demiting the next columnSorry, I did not get this point. Could you explain it better?
In your case you can do in this way. You will have to hardcode the value in any case:
WITH mydata AS
   SELECT 2 id, '2_1' key, 'val2_1' val FROM DUAL UNION ALL
   SELECT 2 id, '2_2' key, 'val2_2' val FROM DUAL UNION ALL
   SELECT 5 id, '5_1' key, 'val5_1' val FROM DUAL UNION ALL
   SELECT 5 id, '5_2' key, 'val5_2' val FROM DUAL UNION ALL
   SELECT 5 id, '5_3' key, 'val5_3' val FROM DUAL UNION ALL
   SELECT 5 id, '5_4' key, 'val5_4' val FROM DUAL
SELECT MAX(CASE
              WHEN id = 2 and key = '2_1'
                OR id = 5 and key = '5_1'
              THEN
                 val
           END) AS val1
     , MAX(CASE
              WHEN id = 2 and key = '2_2'
                OR id = 5 and key = '5_2'
              THEN
                 val
           END) AS val2
     , MAX(CASE
              WHEN id = 5 and key = '5_3'
              THEN
                 val
           END) AS val3
     , MAX(CASE
              WHEN id = 5 and key = '5_4'
              THEN
                 val
           END) AS val4
  FROM mydata
GROUP BY id;
VAL1   VAL2   VAL3   VAL4 
val2_1 val2_2             
val5_1 val5_2 val5_3 val5_4If you think it can be done in a different way, please post some sample data reflecting as much as possible your real data.
Regards.
Al