Function output in a select statement

Question

I have written a function(get_cols) which returns the following string (this string is created dynamically from the fuctions depending on the rows of the tables)
the output of the function is;
MAX (CASE WHEN field_code = 'test_pho' THEN VALUE END ) AS test_pho
,MAX (CASE WHEN field_code = 'ESN' THEN VALUE END ) AS ESN
,MAX (CASE WHEN field_code = 'IMSI' THEN VALUE END ) AS IMSI
,MAX (CASE WHEN field_code = 'PHONE_NO' THEN VALUE END ) AS PHONE_NO
What I need to do is to use this as it is in a another select statement like;
(1)
select
empno,
MAX (CASE WHEN field_code = 'test_pho' THEN VALUE END ) AS test_pho
,MAX (CASE WHEN field_code = 'ESN' THEN VALUE END ) AS ESN
,MAX (CASE WHEN field_code = 'IMSI' THEN VALUE END ) AS IMSI
,MAX (CASE WHEN field_code = 'PHONE_NO' THEN VALUE END ) AS PHONE_NO
from my_employee e, my_columns c
where e.emp_no = c.emp_no
and c.emp_no = '100003'
group by empno
function returns the correct output, but when i call the function in the select like below it get it as a whole string and doesn't give the correct output
(2)
select empno, get_cols('100003')
from my_employee e, my_columns c
where e.emp_no = c.emp_no
and c.emp_no = '100003'
how can i get the output of the function to the select as separate line as shown is above(1)
When I get the above output separately and give in the select as above (1) it gives the correct output I want ??
any help please

Answer 1

josleen wrote:
Hi BluShadow,
Your solution seem interesting. Can you explain how can this be used to produce the required output ? Do we need to maintain a separate my_columns table ?Not quite sure what you are asking?
As with any query, the output columns have to be defined at design time, you cannot have the number of columns dynamically generated based on the data. So, if you said you wanted to pivot data from rows to columns and the number of possible values in those rows could change, you cannot pivot those to columns unless you are expecting those values or have allowed for maximum number of values.
Basic example... Let say we have the following data:
SQL> select * from dept;
    DEPTNO DNAME          LOC
        10 ACCOUNTING     NEW YORK
        20 RESEARCH       DALLAS
        30 SALES          CHICAGO
        40 OPERATIONS     BOSTONAnd we want to pivot that data to put the departments as columns rather than rows. We can simply do something like:
SQL> ed
Wrote file afiedt.buf
  1  select max(decode(deptno,10,dname)) as dname_10
  2        ,max(decode(deptno,10,loc)) as   loc_10
  3        ,max(decode(deptno,20,dname)) as dname_20
  4        ,max(decode(deptno,20,loc)) as   loc_20
  5        ,max(decode(deptno,30,dname)) as dname_30
  6        ,max(decode(deptno,30,loc)) as   loc_30
  7        ,max(decode(deptno,40,dname)) as dname_40
  8        ,max(decode(deptno,40,loc)) as   loc_40
  9* from dept
SQL> /
DNAME_10       LOC_10        DNAME_20       LOC_20        DNAME_30       LOC_30        DNAME_40       LOC_40
ACCOUNTING     NEW YORK      RESEARCH       DALLAS        SALES          CHICAGO       OPERATIONS     BOSTON
SQL>However if a further department is added:
SQL> insert into dept values (50, 'IT SUPPORT', 'LONDON');
1 row created.
SQL> select max(decode(deptno,10,dname)) as dname_10
  2        ,max(decode(deptno,10,loc)) as   loc_10
  3        ,max(decode(deptno,20,dname)) as dname_20
  4        ,max(decode(deptno,20,loc)) as   loc_20
  5        ,max(decode(deptno,30,dname)) as dname_30
  6        ,max(decode(deptno,30,loc)) as   loc_30
  7        ,max(decode(deptno,40,dname)) as dname_40
  8        ,max(decode(deptno,40,loc)) as   loc_40
  9  from dept
10  /
DNAME_10       LOC_10        DNAME_20       LOC_20        DNAME_30       LOC_30        DNAME_40       LOC_40
ACCOUNTING     NEW YORK      RESEARCH       DALLAS        SALES          CHICAGO       OPERATIONS     BOSTON
SQL>we obviously don't get to see the new data, unless we change our query to add this expected additional column(s) in...
SQL> ed
Wrote file afiedt.buf
  1  select max(decode(deptno,10,dname)) as dname_10
  2        ,max(decode(deptno,10,loc)) as   loc_10
  3        ,max(decode(deptno,20,dname)) as dname_20
  4        ,max(decode(deptno,20,loc)) as   loc_20
  5        ,max(decode(deptno,30,dname)) as dname_30
  6        ,max(decode(deptno,30,loc)) as   loc_30
  7        ,max(decode(deptno,40,dname)) as dname_40
  8        ,max(decode(deptno,40,loc)) as   loc_40
  9        ,max(decode(deptno,50,dname)) as dname_50
10        ,max(decode(deptno,50,loc)) as   loc_50
11* from dept
SQL> /
DNAME_10       LOC_10        DNAME_20       LOC_20        DNAME_30       LOC_30        DNAME_40       LOC_40        DNAME_50       LOC_50
ACCOUNTING     NEW YORK      RESEARCH       DALLAS        SALES          CHICAGO       OPERATIONS     BOSTON        IT SUPPORT     LONDON
SQL>Now, rather than having a messy SQL statment with lots of max(decode... statements we can provide a pipelined function to return the same thing...
SQL> ed
Wrote file afiedt.buf
  1  CREATE OR REPLACE TYPE mydepts AS OBJECT
  2  ( dname_10   VARCHAR2(14),
  3    loc_10     VARCHAR2(13),
  4    dname_20   VARCHAR2(14),
  5    loc_20     VARCHAR2(13),
  6    dname_30   VARCHAR2(14),
  7    loc_30     VARCHAR2(13),
  8    dname_40   VARCHAR2(14),
  9    loc_40     VARCHAR2(13),
10    dname_50   VARCHAR2(14),
11    loc_50     VARCHAR2(13)
12* )
13  /
Type created.
SQL> ed
Wrote file afiedt.buf
  1* CREATE OR REPLACE TYPE mydepttable AS TABLE OF mydepts
SQL> /
Type created.
SQL> ed
Wrote file afiedt.buf
  1  CREATE OR REPLACE FUNCTION alldepts RETURN mydepttable PIPELINED IS
  2    v_obj mydepts := mydepts(NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL,NULL);
  3    CURSOR cur_depts IS
  4      select deptno, dname, loc from dept;
  5  BEGIN
  6    FOR i IN cur_depts
  7    LOOP
  8      CASE i.deptno
  9       WHEN 10 THEN v_obj.dname_10 := i.dname; v_obj.loc_10 := i.loc;
10       WHEN 20 THEN v_obj.dname_20 := i.dname; v_obj.loc_20 := i.loc;
11       WHEN 30 THEN v_obj.dname_30 := i.dname; v_obj.loc_30 := i.loc;
12       WHEN 40 THEN v_obj.dname_40 := i.dname; v_obj.loc_40 := i.loc;
13       WHEN 50 THEN v_obj.dname_50 := i.dname; v_obj.loc_50 := i.loc;
14      ELSE NULL;
15      END CASE;
16    END LOOP;
17    PIPE ROW (v_obj);
18    RETURN;
19* END;
SQL> /
Function created.
SQL> select *
  2  from table(alldepts());
DNAME_10       LOC_10        DNAME_20       LOC_20        DNAME_30       LOC_30        DNAME_40       LOC_40        DNAME_50       LOC_50
ACCOUNTING     NEW YORK      RESEARCH       DALLAS        SALES          CHICAGO       OPERATIONS     BOSTON        IT SUPPORT     LONDON
SQL>So, now there is a single function that acts like a table and does the functionality of pivoting the data. What the function actually does to generate the data is entirely up to you whether that is obtaining data from different tables or some PL/SQL code processing to perform some complex algorithmic type thing on the data from a passed in parameter etc. The point of the pipeline function is that it can return multiple columns of data and act as if it is a table that can be queried against, however it still remains that the output columns must be known at design time. This is actually a requirement of the SQL engine, as you cannot make a dynamic function that returns X number of columns based on data.
There is however a technique that can allow you to dynamically generate a number of columns based on data, but it involves getting a little more under the hood of Oracle and interfacing with the ODCI interface, actually using a pipelined funcion technique in conjunction with, in laymans terms, stepping in at the point the query is executing and telling oracle what columns this pipelined function is going to return, but at the same time as defining the returned columns.
An example of doing this is given by AScheffer on this thread: How to pipeline a function with a dynamic number of columns?
... but you really are getting into a complex world if you try and write your queries this way, just to try and make generic queries with dynamic columns.
In essence, the actual business need to actually return a dynamic number of columns is very slim as most applications, whether that is a user facing interface or report generator etc. will be expecting certain data columns to be returned to expected columns or fields within them. If you really want to dynamically pivot data with an unknown number of columns it is usually best to let things like reporting tools deal with this area as that is what they are best at, and don't try and do it within SQL.