Back-referencing data using MODEL

Question

Not sure how possible this is...
with t as (select 1 as id, 'blah blah' as txt from dual union all
           select 2,       '# header 1' from dual union all
           select 3,       '41 text line 1' from dual union all
           select 4,       '22 text line 2' from dual union all
           select 5,       '17 text line 3' from dual union all
           select 6,       'garbage' from dual union all
           select 7,       '# header 2' from dual union all
           select 8,       '77 text line 4' from dual union all
           select 9,       '80 text line 5' from dual union all
           select 10,      '#80 old text line 5' from dual union all
           select 11,      '14 text line 6' from dual union all
           select 12,      'more garbage' from dual)In an ordered sequence of data, "Header" details need determining for all subsequent rows that begin with a <numeric> or #<numeric> value,
such that the output of the above data would appear as below:
ID TXT                  HEADER
1  blah blah      
2  # header 1     
3  41 text line 1       # header 1
4  22 text line 2       # header 1
5  17 text line 3       # header 1
6  garbage
7  # header 2
8  77 text line 4       # header 2
9  80 text line 5       # header 2
10 #80 old text line 5  # header 2
11 14 text line 6       # header 2
12 more garbageIs there a way to do this purely in SQL, with the model clause, perhaps using iterations?
select id, txt, header
from   t
model  dimension by (id)
       measures (txt txt, CAST(NULL AS VARCHAR2(30)) header)
       rules
          header[id is any] = ?
         )I'm still trying to get my head around the usage of the model clause, especially trying to reference backwards with iterations and the until clause.
Cheers
Blu
edit: just in case it's not obvious... "header" details are lines beginning with #<non numeric>
Message was edited by:
BluShadow

Answer 1

Just for fun, a model clause solution:
SQL> with t as (select 1 as id, 'blah blah' as txt from dual union all
  2             select 2,       '# header 1' from dual union all
  3             select 3,       '41 text line 1' from dual union all
  4             select 4,       '22 text line 2' from dual union all
  5             select 5,       '17 text line 3' from dual union all
  6             select 6,       'garbage' from dual union all
  7             select 7,       '# header 2' from dual union all
  8             select 8,       '77 text line 4' from dual union all
  9             select 9,       '80 text line 5' from dual union all
10             select 10,      '#80 old text line 5' from dual union all
11             select 11,      '14 text line 6' from dual union all
12             select 12,      'more garbage' from dual)
13  -- end of test data
14  select id
15       , txt
16       , header
17    from t
18   model
19         dimension by (id)
20         measures (txt, cast(null as varchar2(30)) header, cast(null as varchar2(30)) tmp)
21         rules
22         ( tmp[any] order by id
23           = case
24             when txt[cv()] like '# header%' then txt[cv()]
25             else tmp[cv()-1]
26             end
27         , header[any] order by id
28           = case
29             when  txt[cv()] not like '# header%'
30               and translate(ltrim(substr(txt[cv()],1,instr(txt[cv()]||' ',' ')-1),'#'),'_0123456789','_') is null
31             then tmp[cv()]
32             end
33         )
34  /
                                    ID TXT                 HEADER
                                     1 blah blah
                                     2 # header 1
                                     3 41 text line 1      # header 1
                                     4 22 text line 2      # header 1
                                     5 17 text line 3      # header 1
                                     6 garbage
                                     7 # header 2
                                     8 77 text line 4      # header 2
                                     9 80 text line 5      # header 2
                                    10 #80 old text line 5 # header 2
                                    11 14 text line 6      # header 2
                                    12 more garbage
12 rijen zijn geselecteerd.Regards,
Rob.