Question

I have a month and day; I’m trying to work out the next date that this month and day will exist after a given date.

For instance if I have a table as follows, where the reference_date is of the format 'MMDD'. Please don’t blame me for this; it is the official format for filing in the UK.

create table tmp_ben_dates ( filing_date date, reference_date varchar2(4));
insert all
into tmp_ben_dates values ( to_date('31/12/2011','dd/mm/yyyy'), '1231')
into tmp_ben_dates values ( to_date('31/12/2011','dd/mm/yyyy'), '1130')
into tmp_ben_dates values ( to_date('31/12/2011','dd/mm/yyyy'), '0101')
into tmp_ben_dates values ( to_date('31/07/2011','dd/mm/yyyy'), '0601')
into tmp_ben_dates values ( to_date('31/07/2011','dd/mm/yyyy'), '0801')
select * from dual;

I would like to return the first date that each reference_date occurs after each filing_date. For instance in the first example this would be 31/12/2012 and in the last this would be 01/08/2011.

My best attempt, so far, is as follows

with new_date as ( 
  select reference_date
       , filing_date
       , add_months( trunc(filing_date,'y')
                   , to_number(substr(reference_date,1,2)) - 1) 
           + to_number(substr(reference_date,3)) - 1 as the_date
    from tmp_ben_dates
         )
select filing_date
     , reference_date
     , case when filing_date < the_date then next_date 
            else add_months(the_date,12) end
  from new_date

Which returns the correct results:

FILING_DATE         REFE NEXT_DATE
------------------- ---- -------------------
31/12/2011 00:00:00 1231 31/12/2012 00:00:00
31/12/2011 00:00:00 1130 30/11/2012 00:00:00
31/12/2011 00:00:00 0101 01/01/2012 00:00:00
31/07/2011 00:00:00 0601 01/06/2012 00:00:00
31/07/2011 00:00:00 0801 01/08/2011 00:00:00

However, it’s also completely ridiculous; difficult to read and understand.

It’s also possible using interval but I don’t really see how it makes it any less confusing to determine what’s going on.

with new_date as ( 
  select reference_date
       , filing_date
       , trunc(filing_date,'y') 
          + to_yminterval( 'P' 
                           || to_char(to_number(substr(reference_date,1,2)) - 1) 
                           || 'M') 
          + to_dsinterval( 'P' 
                           || to_char(to_number(substr(reference_date,3)) - 1) 
                           || 'D') as the_date
    from tmp_ben_dates
         )
select filing_date
       , reference_date
       , case when filing_date < the_date then the_date 
              else add_months(the_date,12) end as next_date
  from new_date

Is there something really obvious that I’m missing here? Is there an easy method of doing this in SQL?

Answer 1

You might find this simpler. It simply puts the date in YYYYMMDD format and uses that as the second argument to to_date():

with new_date as (
    select reference_date, filing_date,
           to_date(to_char(filing_date, 'yyyy')||reference_date), 'yyyymmdd'
                  ) - 1 as the_date
    from tmp_ben_dates
  )
select filing_date, reference_date,
      (case when filing_date < the_date then next_date
            else add_months(the_date,12)
       end)
from new_date