A function instead of UNBOUNDED PRECEDING (like "Last field with value=0")

Question

Hello,
I have a table with many rows. The attributes of the table are code, month and value.
For each code there are 12 months and 12 values.
No I want to add the gaps between the months...
Is it possible to count the following gaps between the different rows...?
For example:
Original table1:_
code, month, value
1,1,20
1,2,0
1,3,30
1,4,0
1,5,40
1,6,0
1,7,0
1,8,20
1,9,0
1,10,10
1,11,0
1,12,0
5,1,0
5,2,20
5,3,10
description:*
january value = 20
february value = 0 (=>count this gap => new value 1 )
march value = 30
april value = 0 (=>count this gap => new value 1 )
may value = 40
june value = 0
july value = 0 (=>count this two following gaps => new value 2 )
agust value = 20
september value = 0 (=>count this gap => new value 1 )
october value = 10
november value = 0
december value = 0 (=>count this two following gaps => new value 2 )
New target table:_
code, month, value
1,1,20
1,2,*1*
1,3,30
1,4,*1*
1,5,40
1,6,0
1,7,*2*
1,8,20
1,9,*1*
1,10,10
1,11,0
1,12,*2*
5,1,*1*
5,2,20
5,3,10
I tried this code:
select code, month
sum(value) over (
order by month
rows between unbounded preceding and current row
*) as final_value*
from table1 order by month;
This adds all following fields cumulative from beginning to current_row. But I need this adding only for the following gaps... then start with countin by 0.
I need only the following like in the example on top. Maybe is there an other function like decode to count only the following gaps...!?
A function instead of unbounded preceding....like "*Last field with value=0*" or something... ?
Best regards,
Tim

Answer 1

TimB83 wrote:
I have a table with many rows. The attributes of the table are code, month and value.
For each code there are 12 months and 12 values.
No I want to add the gaps between the months...
Is it possible to count the following gaps between the different rows...?
For example:
Original table1:_
code, month, value
1,1,20
1,2,0
1,3,30
1,4,0
1,5,40
1,6,0
1,7,0
1,8,20
1,9,0
1,10,10
1,11,0
1,12,0
5,1,0
5,2,20
5,3,10
description:*
january value = 20
february value = 0 (=>count this gap => new value 1 )
march value = 30
april value = 0 (=>count this gap => new value 1 )
may value = 40
june value = 0
july value = 0 (=>count this two following gaps => new value 2 )
agust value = 20
september value = 0 (=>count this gap => new value 1 )
october value = 10
november value = 0
december value = 0 (=>count this two following gaps => new value 2 )
New target table:_
code, month, value
1,1,20
1,2,*1*
1,3,30
1,4,*1*
1,5,40
1,6,0
1,7,*2*
1,8,20
1,9,*1*
1,10,10
1,11,0
1,12,*2*
5,1,*1*
5,2,20
5,3,10
...Tim,
you should post this question on the "SQL and PL/SQL" forum since it's a typical SQL question.
There are probably much better ways to accomplish this and the guys over there will tell you, but here are two examples that might get you started:
1. Pre-10g without MODEL clause
with t as (
select 1 as code, 1 as month, 20 as value from dual union all
select 1 as code, 2 as month, 0 as value from dual union all
select 1 as code, 3 as month, 30 as value from dual union all
select 1 as code, 4 as month, 0 as value from dual union all
select 1 as code, 5 as month, 40 as value from dual union all
select 1 as code, 6 as month, 0 as value from dual union all
select 1 as code, 7 as month, 0 as value from dual union all
select 1 as code, 8 as month, 20 as value from dual union all
select 1 as code, 9 as month, 0 as value from dual union all
select 1 as code, 10 as month, 10 as value from dual union all
select 1 as code, 11 as month, 0 as value from dual union all
select 1 as code, 12 as month, 0 as value from dual union all
select 5 as code, 1 as month, 0 as value from dual union all
select 5 as code, 2 as month, 20 as value from dual union all
select 5 as code, 3 as month, 10 as value from dual
r1 as (
select
        case
        when value = 0
        then 1
        else 0
        end as is_gap
      , case
        when value != 0
        then rownum
        else null
        end as grp_info
      , code
      , month
      , value
from
        t
r2 as (
select
        last_value(grp_info ignore nulls) over (partition by code order by month) as grp
      , is_gap
      , code
      , month
      , value
from
        r1
select
        code
      , month
      , case
        when value = 0
        and (lead(value) over (partition by code order by month) != 0 or
             lead(value) over (partition by code order by month) is null)
        then sum(is_gap) over (partition by code, grp)
        else value
        end as value
from r2;2. 10g and later with MODEL clause
with t as (
select 1 as code, 1 as month, 20 as value from dual union all
select 1 as code, 2 as month, 0 as value from dual union all
select 1 as code, 3 as month, 30 as value from dual union all
select 1 as code, 4 as month, 0 as value from dual union all
select 1 as code, 5 as month, 40 as value from dual union all
select 1 as code, 6 as month, 0 as value from dual union all
select 1 as code, 7 as month, 0 as value from dual union all
select 1 as code, 8 as month, 20 as value from dual union all
select 1 as code, 9 as month, 0 as value from dual union all
select 1 as code, 10 as month, 10 as value from dual union all
select 1 as code, 11 as month, 0 as value from dual union all
select 1 as code, 12 as month, 0 as value from dual union all
select 5 as code, 1 as month, 0 as value from dual union all
select 5 as code, 2 as month, 20 as value from dual union all
select 5 as code, 3 as month, 10 as value from dual
select
         code
       , month
       , value
from
         t
model
partition by (code)
dimension by (month)
measures (value, 0 as gap_cnt)
rules (
  gap_cnt[any] order by month =
  case
  when value[cv() - 1] = 0
  then gap_cnt[cv() - 1] + 1
  else 0
  end,
  value[any] order by month =
  case
  when value[cv()] = 0 and presentv(value[cv() + 1], value[cv() + 1], 1) != 0
  then presentv(gap_cnt[cv() + 1], gap_cnt[cv() + 1], gap_cnt[cv()] + 1)
  else value[cv()]
  end
);Regards,
Randolf
Oracle related stuff blog:
http://oracle-randolf.blogspot.com/
SQLTools++ for Oracle (Open source Oracle GUI for Windows):
http://www.sqltools-plusplus.org:7676/
http://sourceforge.net/projects/sqlt-pp/