Dynamic column names in Oracle

Question

HI SOS!! calling for help...
i am trying to query the following. I want to remove the duplicates in the ID, and combine them into one... While doing so, I want to keep the data as it is by creating dynamic columns reason_1, reason_2, reason_3 with their corresponding TIME. In case of same reasons, I want to add the TIME. The database is as follows:
ID     REASON     TIME
A     41A..........     27
A     93K..........16
B     89C...........3
B     93K...........7
B     48C..........     4
C     93K..........     24
C     93K..........     7
C     48C.........10
Expected Result is
ID     REASON_1......TIME_1.... REASON_2...TIME_2....REASON_3.........TIME_3
A........ 41A............ 27........... 93K............16          
B........ 89C............ 3............ 93K............7...........48C.................. 4
C........ 93K........... 31.......... 48C..............10          
Would be grateful if someone could try helping me out of this as early as possible.
Edited by: 968125 on Oct 28, 2012 11:01 PM

Answer 1

You still haven't made it clear what the requirement is.
Is there a maximum number of reasons that can occur per ID? If so, you can code the pivoting of your data with something like:
SQL> ed
Wrote file afiedt.buf
  1  with t as (select 'A' as ID, '41A' as reason, 27 as time from dual union all
  2             select 'A', '93K', 16 from dual union all
  3             select 'B', '89C', 3 from dual union all
  4             select 'B', '93K', 7 from dual union all
  5             select 'B', '48C', 4 from dual union all
  6             select 'C', '93K', 24 from dual union all
  7             select 'C', '93K', 7 from dual union all
  8             select 'C', '48C', 10 from dual)
  9  --
10  -- END OF TEST DATA TABLE SIMULATED USING 'WITH' CLAUSE
11  -- SIMPLY USE THE SELECT STATEMENT BELOW ON YOUR OWN TABLE
12  --
13  select id
14        ,max(decode(rn,1,reason)) as reason_1
15        ,max(decode(rn,1,time)) as time_1
16        ,max(decode(rn,2,reason)) as reason_2
17        ,max(decode(rn,2,time)) as time_2
18        ,max(decode(rn,3,reason)) as reason_3
19        ,max(decode(rn,3,time)) as time_3
20        ,max(decode(rn,4,reason)) as reason_4
21        ,max(decode(rn,4,time)) as time_4
22        ,max(decode(rn,5,reason)) as reason_5
23        ,max(decode(rn,5,time)) as time_5
24  from (/* assign some row numbers within each ID, ordering by 'reason' then 'time' */
25        select id, reason, time
26              ,row_number() over (partition by id order by reason, time) as rn
27        from   t
28       )
29* group by id
SQL> /
I REA     TIME_1 REA     TIME_2 REA     TIME_3 REA     TIME_4 REA     TIME_5
A 41A         27 93K         16
B 48C          4 89C          3 93K          7
C 48C         10 93K          7 93K         24In this example, it will cater for up to 5 reasons/times per ID... and you can code for more if you need very easily.
Also, as you're using 11g, you can use the new PIVOT query, for which you can find plenty of examples if you google or search these forums, or follow the FAQ link.
If you don't have any idea how many reasons there could be, then you really are looking at generating columns dynamically.
The problem with that is, because Oracle needs to know the SQL projection (see the FAQ link already provided), it has to know how many columns are going to be returned from a query before any data is fetched... but you are saying that the columns returned have to be based on the data itself. To do that you have two main options:
1) You query the data once to determine how many columns you are going to need and then dynamically build up a query in PL/SQL code with the right number of columns (or use one of the other dynamic methods shown in the FAQ)
or
2) You step back and ask yourself if SQL is really the best place to be trying to produce this output. Reporting tools are often better suited to this requirement as they are made to query the data back from a database and then pivot and layout the data based on the data returned.