Lead analytic function

Question

hi all, i am trying to find out how lead works and i have a good idea of it. however i am running into problem.
i have a table that contain this data
1 2.2.2010
2 2.2.2010
3 2.2.2010
10 2.15.2010
11 2.15.2010
12 2.15.2010
20 2.20.2010
for every row with date 2.2.2010 i want to lead to 2.15.2010. i have this query that doesnt work as expected
SELECT     id, dt,
     LEAD (dt, 3, dt+1) )
          OVER (ORDER BY dt)          AS next
     FROM     t1
so my output should be
id dt next
1 2.2.2010 2.15.2010
2 2.2.2010 2.15.2010
3 2.2.210 2.15.2010
10 2.15.2010 2.16.2010
11 2.15.2010 2.16.2010
12 2.15.2010 2.16.2010
20 2.20.2010
as you can see in my query, i am hardcoding the value 3 which will lead 3 rows for row1 but for row 2 it will lead 3 more rows ahead too and so on.
what i want is that all rows with 2.2.10 lead to the first row that start with 2.15.2010. i dont know the count of row from the first row to the first row with 2.15.2010. there could be many rows witn 2.2.2010 so i cannot hardcode the value of row to lead.
can somebody tell me how to write a query that display the output above

Answer 1

If I understand you correctly, then I think this is what you're after:
with your_data as (select 1 id, to_date('02/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 2 id, to_date('02/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 3 id, to_date('02/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 4 id, to_date('02/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 10 id, to_date('15/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 11 id, to_date('15/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 12 id, to_date('15/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 13 id, to_date('15/02/2010', 'dd/mm/yyyy') dt from dual union all
                   select 20 id, to_date('20/02/2010', 'dd/mm/yyyy') dt from dual),
--- end of mimicking a table called "your_data"
       results as (select id,
                          dt,
                          case when lead(dt) over (order by id) = dt then NULL
                               else lead(dt, 1, dt+1) over (order by id)
                          end next_date
                   from   your_data)
select id,
       dt,
       first_value(next_date ignore nulls) over (order by id
                                                 ROWS BETWEEN current row
                                                          AND unbounded following) next_date
from   results
        ID DT         NEXT_DATE
         1 02/02/2010 15/02/2010
         2 02/02/2010 15/02/2010
         3 02/02/2010 15/02/2010
         4 02/02/2010 15/02/2010
        10 15/02/2010 20/02/2010
        11 15/02/2010 20/02/2010
        12 15/02/2010 20/02/2010
        13 15/02/2010 20/02/2010
        20 20/02/2010 21/02/2010You have to do it in two stages - first, identify the "next" date where it's different to the current date - this will only populate one row and leave the rest blank, then backfill the values.