在Oracle SQL中填充缺少的行

Question

I have following data:

date          product    amount
2013-01-31    a1           100
2013-02-28    a1           200
2013-01-31    b1           700
2013-04-30    b1           100
2013-06-30    b1          1300
2013-03-31    c1            10
2013-07-31    c1            70

Is it possible to create query in sql/plsql that would populate missing rows based on dates? I mean i would like to get following result:

date          product    amount
 2013-01-31    a1           100
 2013-02-28    a1           200
 2013-01-31    b1           700
*2013-02-28    b1           500*
*2013-03-31    b1           300*
 2013-04-30    b1           100
*2013-05-31    b1           700*
 2013-06-30    b1          1300
 2013-03-31    c1            10
*2013-04-30    c1            25*
*2013-05-31    c1            40*
*2013-06-30    c1            55*
 2013-07-31    c1            70

ie create missing dates, duplicate product code and calculate amount.

Answer 1

Using a combination of LEAD function and hierarchical queries, this can be achieved in a single query.

Demo here .

    SELECT DISTINCT
           ADD_MONTHS (product_date, LEVEL - 1), product, amount + ( (LEVEL - 1) * mul_factor)
      FROM (SELECT product_date, product, amount, next_date,
                   amount_diff / month_diff mul_factor
              FROM (SELECT product_date, product, amount,
                           LEAD (product_date, 1) OVER (PARTITION BY product ORDER BY product_date)
                              AS next_date,
                           MONTHS_BETWEEN (
                              (LEAD (product_date, 1) OVER (PARTITION BY product ORDER BY product_date)),
                              product_date)
                              AS month_diff,
                           LEAD (amount, 1) OVER (PARTITION BY product ORDER BY product_date)
                              AS next_amount,
                             LEAD (amount, 1) OVER (PARTITION BY product ORDER BY product_date)
                           - amount
                              AS amount_diff
                      FROM mytable)
             WHERE next_date IS NOT NULL)
CONNECT BY ADD_MONTHS (product_date, LEVEL - 1) <= next_date
  ORDER BY 2, 1

Here LEAD function is used to get the next date and next amount. Using this, the difference in the months, difference in amount and value required to evenly distribute the amount can be found. This is later used in hierarchal query to get all the months between the start and end dates. But this gives several duplicate rows, which i can't seem to eliminate. Hence have used DISTINCT keyword.

Output:

01/31/2013    a1    100
02/28/2013    a1    200
01/31/2013    b1    700
02/28/2013    b1    500
03/31/2013    b1    300
04/30/2013    b1    100
05/31/2013    b1    700
06/30/2013    b1    1300
03/31/2013    c1    10
04/30/2013    c1    25
05/31/2013    c1    40
06/30/2013    c1    55
07/31/2013    c1    70

Answer 2

I assume that product_amounts is the name of this table.

declare
  n integer;
  i integer;
  a integer;
  d date;
begin
  for x in
  (
    select *
    from   ( select product,
                    amount,
                    trunc(date, 'MONTH') mon,
                    lead(trunc(date, 'MONTH')) over(partition by product order by date) next_mon,
                    lead(amount) over(partition by product order by date) next_amount,
             from   product_amounts
           )
    where  months_between(next_mon, mon) > 1
  )
  loop
    n := months_between(x.next_mon, x.mon);
    for i in 1 .. n-1
    loop
      d := add_months(x.mon, i);
      a := x.amount + (x.next_amount - x.amount)/n;

      insert into product_amounts(date, product, amount)
      values (last_day(d), x.product, a);
    end loop;
  end loop;

  commit;
end;