Pivoting using sys_connect_by_path

Question

Can you please let me know how to get this output using sys_connect_by_path hierarchical function ?
JOB DEPT_10 DEPT_20 DEPT_30 DEPT_40
CLERK 1 2 1 (null)
SALESMAN (null) (null) 4 (null)
PRESIDENT 1 (null) (null) (null)
MANAGER 1 1 1 (null)
ANALYST (null) 2 (null) (null)
I know how to acheive this by decode function.Even I have the query ready .
select job,
max( decode( deptno, 10, cnt, null ) ) dept_10,
max( decode( deptno, 20, cnt, null ) ) dept_20,
max( decode( deptno, 30, cnt, null ) ) dept_30,
max( decode( deptno, 40, cnt, null ) ) dept_40
from ( select job, deptno, count(*) cnt from emp group by job, deptno )
group by job
But my requirement is to get the same output by using SYS_CONNECT_BY_PATH.
Please let me know the query if its possible.Else let me know the reason why we cant do .

Answer 1

Hi,
Welcome to thre forum!
If you're using Oracle 11, then SELECT ... PIVOT can help do what you want. In any version, you can do something like what you're already doing, though you don't need a sub-query:
SELECT       job
,       SUM (DECODE (deptno, 10, 1))     AS dept_10
,       SUM (DECODE (deptno, 20, 1))     AS dept_20
,       SUM (DECODE (deptno, 30, 1))     AS dept_30
,       SUM (DECODE (deptno, 40, 1))     AS dept_40
FROM       scott.emp
GROUP BY  job
ORDER BY  job     -- If wanted
;Either way, SYS_CONNECT_BY_PATH doesn't help any.
The one reason why you might want to use SYS_CONNECT_BY_PATH is if you didn't know how many different departments there might be. In that case, you could do something like this:
WITH     got_cnt          AS
     SELECT       e.job
     ,       d.deptno
     ,       NULLIF (COUNT (e.deptno), 0)               AS cnt
     ,       DENSE_RANK () OVER (ORDER BY d.deptno)     AS d_num
     FROM             scott.dept  d
     LEFT OUTER JOIN  scott.emp   e   PARTITION BY (e.job)
                                        ON  d.deptno = e.deptno
     GROUP BY  e.job
     ,            d.deptno
SELECT       job
,       REPLACE ( SYS_CONNECT_BY_PATH ( NVL2 ( cnt
                                      , TO_CHAR (cnt, '9999')
            )          AS all_cnts
FROM       got_cnt
WHERE       CONNECT_BY_ISLEAF     = 1
START WITH     d_num     = 1
CONNECT BY     d_num     = PRIOR d_num + 1
     AND     job     = PRIOR job
ORDER BY  job
;As you can see, it's quite a bit more complicated than what you already have, and it's less efficient.
Also, it doesn't produce separate columns for each depararment. All the departments are concatenated together, formatted so they look like separate columns. Also, the query above doesn't have a header identifiyuing which department corresponds to each "column" , but you could fix that by doing a UNION with another CONNECT BY query to generate a header row.
To get truly separate columns, you could use various string manipulation techniques to split the results of SYS_CONNECT_BY_PATH into separate columns, but that requires knowing exactly how many departments to expect, which defeats the purpose of using SYS_CONNECT_BY_PATH.