Help on query : How to produce columns based on rows?

Question

Hi All,
I have a table table1 with below structure and data
Item     State          Value
A1      MS          100
A1      GOA          3
A1      UP          4
B2      MS          34
B2      GOA          55
B2      PB          3
B2      DL          4
B3      MS          99
B3      TN          1I have an requirement to query this table and generate below output
Item      MS     GOA     UP     PB     DL     TN
A1     100      3     4     0     0     0
B2      34     55     0     3     4     0
B3      99      0     0     0     0     1The issue here is that we can have more State codes hence accordingly the query should dynamically return those number of columns
Is there any way to produce above output? Any help will be appreciated.
Oracle DB Version - Oracle 8i
Thanks in advance!

Answer 1

Hi,
That's called a pivot .
The basic tecnique is something like:
SELECT     item
,     SUM (CASE WHEN state = 'MS'  THEN value ELSE 0)     AS ms
,     SUM (CASE WHEN state = 'GOA' THEN value ELSE 0)     AS goa
,     SUM (CASE WHEN state = 'UP'  THEN value ELSE 0)     AS up
,     SUM (CASE WHEN state = 'PB'  THEN value ELSE 0)     AS pb
,     SUM (CASE WHEN state = 'DL'  THEN value ELSE 0)     AS dl
,     SUM (CASE WHEN state = 'TN'  THEN value ELSE 0)     AS tn
FROM     table_x
GROUP BY  item;but this requires you to know how many states there are, and what they are, when you write the query.
Since you don't, you need to run a preliminary query that will generate a SELECT statement like the one above, or at least the SUM expressions for the variable columns. That's called dynamic SQL , and a technique for doing that is included below. (The example uses COUNT, but any aggregate function, including SUM, will work.)
Another possibility is String Aggregation, where instead of individual columns for each state, you have one huge VARCHAR2 column, formatted to look like separate columns. The tools avaiable for doing that in Oracle 8.1 are pretty poor, however, so I think you're better off with dynamic SQL.
How to Pivot a Table with a Dynamic Number of Columns
This works in any version of Oracle
The "SELECT ... PIVOT" feature introduced in Oracle 11
is much better for producing XML output.
Say you want to make a cross-tab output of
the scott.emp table.
Each row will represent a department.
There will be a separate column for each job.
Each cell will contain the number of employees in
     a specific department having a specific job.
The exact same solution must work with any number
of departments and columns.
(Within reason: there's no guarantee this will work if you
want 2000 columns.)
Case 0 "Basic Pivot" shows how you might hard-code three
job types, which is exactly what you DON'T want to do.
Case 1 "Dynamic Pivot" shows how get the right results
dynamically, using SQL*Plus. 
(This can be easily adapted to PL/SQL or other tools.)
NOTE: Using SQL*Plus file I/O, as in this example, is just
one of many ways to do dynamic SQL.
PROMPT     ==========  0. Basic Pivot  ==========
SELECT     deptno
,     COUNT (CASE WHEN job = 'ANALYST' THEN 1 END)     AS analyst_cnt
,     COUNT (CASE WHEN job = 'CLERK'   THEN 1 END)     AS clerk_cnt
,     COUNT (CASE WHEN job = 'MANAGER' THEN 1 END)     AS manager_cnt
FROM     scott.emp
WHERE     job     IN ('ANALYST', 'CLERK', 'MANAGER')
GROUP BY     deptno
ORDER BY     deptno
PROMPT     ==========  1. Dynamic Pivot  ==========
--     *****  Start of dynamic_pivot.sql  *****
-- Suppress SQL*Plus features that interfere with raw output
SET     FEEDBACK     OFF
SET     PAGESIZE     0
SPOOL     p:\sql\cookbook\dynamic_pivot_subscript.sql
SELECT     DISTINCT
     ',     COUNT (CASE WHEN job = '''
||     job
||     ''' '     AS txt1
,     'THEN 1 END)     AS '
||     job
||     '_CNT'     AS txt2
FROM     scott.emp
ORDER BY     txt1;
SPOOL     OFF
-- Restore SQL*Plus features suppressed earlier
SET     FEEDBACK     ON
SET     PAGESIZE     50
SPOOL     p:\sql\cookbook\dynamic_pivot.lst
SELECT     deptno
@@dynamic_pivot_subscript
FROM     scott.emp
GROUP BY     deptno
ORDER BY     deptno
SPOOL     OFF
--     *****  End of dynamic_pivot.sql  *****
EXPLANATION:
The basic pivot assumes you know the number of distinct jobs,
and the name of each one.  If you do, then writing a pivot query
is simply a matter of writing the correct number of ", COUNT ... AS ..."\
lines, with the name entered in two places on each one.  That is easily
done by a preliminary query, which uses SPOOL to write a sub-script
(called dynamic_pivot_subscript.sql in this example).
The main script invokes this sub-script at the proper point.
In practice, .SQL scripts usually contain one or more complete
statements, but there's nothing that says they have to.
This one contains just a fragment from the middle of a SELECT statement.
Before creating the sub-script, turn off SQL*Plus features that are
designed to help humans read the output (such as headings and
feedback messages like "7 rows selected.", since we do not want these
to appear in the sub-script.
Turn these features on again before running the main query.
*/