Rows data will go side by side

Question

Hi,
I have a table TEST in oracle 10g and under this there are rows like
account_code value
4142 300
4143 400
4144 500
Now I want data like this
4142 4143 4144
300 400 500
please help me to write the query.
Regards,
Jimut

Answer 1

Hi,
A query must have a fixed, known number of columns in its output. The number of columns and their names must be hard-coded when the qiuery is compiled.
You want a variable, unknown number of columns. You don't know the number or the names of the columns now, you want to get thos from the table itsel at run tme.
One way to get what you want is Dynamic SQL . This resolves the conflict of what must be known at compile time versus what is not known until run-time by delaying compile tme until run time. That is, the code you write today is not the query; it is code that will write and execute the query at some later date, using the data in the table at that date. There are dynamic SQL examples in the forum FAQ {message:id=9360005}
Another (simpler) work-around involves changing your requirements a little. Instead of getting a variable, unknown number of columns,you can get one big column, formatted to look like a variable number of columns. This is exactly what the <tt> SELECT ... PIVOT </tt> feature does when you tell it to generate XML output. The query actually produces 1 column of XML data, but, when it is displayed, it looks like a variable number of columns. Another way is String Aggregation , where we produce one big VARCHAR2 column that looks like a variable number of columns. The rest of this message shows an example of string aggregation.
Since you didn't post CREATE TABLE and INSERT statements for your own data (see the fm FAQ {message:id=9360002} ) I'll use the scott.dept table. The table actiually looks like this:
`   DEPTNO DNAME          LOC
        10 ACCOUNTING     NEW YORK
        20 RESEARCH       DALLAS
        30 SALES          CHICAGO
        40 OPERATIONS     BOSTONTo transpose this data so that it looks like this
OUTPUT
10             20             30             40
ACCOUNTING     RESEARCH       SALES          OPERATIONS
NEW YORK       DALLAS         CHICAGO        BOSTONwe can use a query like this:
WITH     cntr             AS
     SELECT     LEVEL     AS n
     FROM     dual
     CONNECT BY     LEVEL     <= 3     -- number of columns in orgiinal table
,     unpivoted_data     AS
     SELECT     d.deptno
     ,     DENSE_RANK () OVER (ORDER BY d.deptno)     AS deptno_rnk
     ,     c.n
     ,     CASE  c.n
              WHEN  1  THEN TO_CHAR (deptno)
              WHEN  2  THEN dname
              WHEN  3  THEN loc
          END     AS txt
     FROM        cntr     c
     CROSS JOIN  scott.dept  d
SELECT       REPLACE ( SYS_CONNECT_BY_PATH ( RPAD ( NVL (txt, ' ')
                                      , 14  -- Max column length
                               , '`~'          -- Never occurs in data
            , '`~'
            )     AS output
FROM       unpivoted_data
WHERE       CONNECT_BY_ISLEAF     = 1
START WITH     deptno_rnk     = 1
CONNECT BY     deptno_rnk     = PRIOR deptno_rnk + 1
     AND     n          = PRIOR n
ORDER BY  n
;This query will work in Oracle 10.1 (or higher).
It assumes we know how many columns in the original table we want to transpose (3 in this case, which will be the number of rows in the output), the maximum length we want to show for each item (14 characters in this case), and some sub-string that never occurs in any of the data ('`~' in this case). It does NOT assume we know the number of rows in the original table, which we be the number of "columns" in the output. In this eample, there are 4 rows that appear as 4 "columns", but the number 4 does not appear anywhere in the query.
Notice that there is only one real column in the output: a VARCHAR2 column called output. The column header that SQL*Plus supplies isn't helpful in this case, so we could suppress it using <tt> SET HEADING OFF </tt>. VARCHAR2 columns in Oracle SQL have a limit of 4000 chararacters , so if we display 14 characters for each "column", and have 1 character separating the "columns", then we can't have more than 266 "columns" in the output. Nobody would be able to read that many, anyway.
The query above can be modified so that the one output column looks more like a variable number of columns:
10             20             30             40
============== ============== ============== ==============
ACCOUNTING     RESEARCH       SALES          OPERATIONS
NEW YORK       DALLAS         CHICAGO        BOSTONYou could also add the original colum names as labels:
DEPTNO: 10             20             30             40
        ============== ============== ============== ==============
DNAME:  ACCOUNTING     RESEARCH       SALES          OPERATIONS
LOC:    NEW YORK       DALLAS         CHICAGO        BOSTONI'll leave these as execises.