UNPIVOT the results of a delimited string on several rows

Question

I am struggling with some complex hierarchical data. I have successfully used a CONNECT BY query to limit the rows down to the subset that i want - and i have used SYS_CONNECT_BY_PATH to return the full tree up to the nodes of interest.

this gives me essentially some rows like this (delimited by '|'):

id  path
-------------------
1, '|10|11|12|13'
2, '|10|14|15'
3, '|16|11|12|13'
4, '|16|17'

now - my challenge is to unwrap or UNPIVOT these values back into a structure like this:

id  ord node
-------------
1, 1, 10
1, 2, 11
1, 3, 12
1, 4, 13
2, 1, 10
2, 2, 14
2, 3, 15
3, 1, 16
3, 2, 11
3, 3, 12
3, 4, 13
4, 1, 16
4, 2, 17

I think i am unable to use UNPIVOT directly as that is working on a fixed set of columns - which this is not.

I am playing with a PIPELINE function to unwrap this, but frankly - passing all these rows to the function is an issue since they come from another query. I am wondering if anyone has a way to UNPIVOT the values from a SYS_CONNECT_BY_PATH result set back into rows that is maybe a pure sql solution - probably with REGEX parsing...

help always appreciated - thanks

Answer 1

Yes, UNPIVOT operator wont do much here to help you produce the desired output.

As one of the approaches you could user regexp_count()(11g R1 version and up) regular

expression function to count all occurrences of numbers and then use regexp_substr() regular

expression function to extract the numbers as follows:

-- sample of data 
SQL> with t1(id1, path1) as(
  2    select 1, '|10|11|12|13' from dual union all
  3    select 2, '|10|14|15' from dual union all
  4    select 3, '|16|11|12|13' from dual union all
  5    select 4, '|16|17' from dual
  6  ),
  7  occurrences(ocr) as( -- occurrences 
  8    select level
  9     from ( select max(regexp_count(path1, '[^|]+')) as mx_ocr
 10              from t1
 11          ) t
 12   connect by level <= t.mx_ocr
 13  )
 14  select id1
 15       , row_number() over(partition by id1 order by id1) as ord
 16       , node
 17    from ( select q.id1
 18                , regexp_substr(q.path1, '[^|]+', 1, o.ocr)        as node
 19             from t1 q
 20            cross join occurrences o
 21          )
 22  where  node is not null
 23  order by id1, 2, node
 24  ;

Result:

       ID1        ORD NODE
---------- ---------- ------------------------------------------------
         1          1 10
         1          2 11
         1          3 12
         1          4 13
         2          1 10
         2          2 14
         2          3 15
         3          1 11
         3          2 12
         3          3 13
         3          4 16
         4          1 16
         4          2 17

13 rows selected

As another approach, starting from 10g version and up, you could use model clause:

 SQL> with t1(id1, path1) as(
  2    select 1, '|10|11|12|13' from dual union all
  3    select 2, '|10|14|15' from dual union all
  4    select 3, '|16|11|12|13' from dual union all
  5    select 4, '|16|17' from dual
  6  )
  7  select id1
  8       , ord
  9       , node
 10    from t1
 11   model
 12   partition by ( rownum as id1)
 13   dimension by ( 1 as ord)
 14   measures( path1
 15           , cast(null as varchar2(11)) as node
 16           , nvl(regexp_count(path1, '[^|]+'), 0) as ocr )
 17   rules(
 18      node[for ord from 1 to ocr[1] increment 1] = 
 19          regexp_substr(path1[1], '[^|]+', 1, cv(ord))
 20  )
 21  order by id1, ord, node
 22  ;

Result:

       ID1        ORD NODE
---------- ---------- -----------
         1          1 10
         1          2 11
         1          3 12
         1          4 13
         2          1 10
         2          2 14
         2          3 15
         3          1 16
         3          2 11
         3          3 12
         3          4 13
         4          1 16
         4          2 17

13 rows selected

SQLFiddle Demo