使用正确的日期值更新表，而在SQL中没有间隙

Question

I have a table tab_assignment_xx

date_from   date_end     action  person_number
01-Apr-2014 31-Jul-2014   HIRE   050498
01-Aug-2014 31-Jan-2015   OTHERS    050498
01-Feb-2015 30-Jun-2015   OTHERS    050498
01-Jul-2015 15-Nov-2015   OTHERS    050498
16-Nov-2015 **31-Dec-4712** OTHERS  050498
01-Jan-2016 **31-Dec-4712** OTHERS  050498

now in this record for employee 050498 there are two dates with 31-dec-12 and a break in the dates. For example before 01-Jan-16 it should have been 31-dec-2015 and not 31-dec-12.

i want to find such breaks in the entire table.

The query :

select *
  from tab_assignment_xx a 
 where not exists (select 1
                     from tab_assignment_xx b
                    where a.date_end + 1  = b.date_from
                      and a.person_number = b.person_number)
   and a.date_from != (select max(date_from)
                         from tab_assignment_xx c
                        where a.person_number = c.person_number);

But this value i want to update in the table for example the output of the table of the above sample should be something like :

  date_from   date_end     action  person_number
    01-Apr-2014 31-Jul-2014   HIRE   050498
    01-Aug-2014 31-Jan-2015   OTHERS    050498
    01-Feb-2015 30-Jun-2015   OTHERS    050498
    01-Jul-2015 15-Nov-2015   OTHERS    050498
    16-Nov-2015 **31-Dec-2015** OTHERS  050498
    01-Jan-2016 **31-Dec-4712** OTHERS  050498


  update tab_assignment_xx

  set effective_end_date =(select max(date_from)-1
                           from tab_assignment_xx a 
                           where not exists (select 1
                           from tab_assignment_xx b
                           where a.date_end + 1  = b.date_from
                         and a.person_number = b.person_number)
                        and a.date_from != (select max(date_from)
                         from TAB_ASSIGNMENT_XX C
                        where a.person_number = c.person_number));

But this update is not working. can anyone give the logic ill try to update the query.

Answer 1

You can use the analytic lead() function to find the expected end date, as the following start date - 1:

select t.*,
  lead(date_from) over (partition by person_number order by date_from) - 1 as lead_from
from tab_assignment_xx t;

And you can then use that as the using clause of a merge , wrapped in another layer of select to only find those that are wrong ( lead_from != date_end ), and also to exclude the final record in the sequence, whose lead_from will be null:

merge into tab_assignment_xx t
using (
  select * from (
    select t.*,
      lead(date_from) over (partition by person_number order by date_from) - 1 as lead_from
    from tab_assignment_xx t
  )
  where lead_from is not null and date_end != lead_from
) tmp
on (t.date_from = tmp.date_from and t.person_number = tmp.person_number)
when matched then update set t.date_end = tmp.lead_from;

1 row merged.

select * from tab_assignment_xx;

DATE_FROM   DATE_END    ACTION PERSON
----------- ----------- ------ ------
01-Apr-2014 31-Jul-2014 HIRE   050498
01-Aug-2014 31-Jan-2015 OTHERS 050498
01-Feb-2015 30-Jun-2015 OTHERS 050498
01-Jul-2015 15-Nov-2015 OTHERS 050498
16-Nov-2015 31-Dec-2015 OTHERS 050498
01-Jan-2016 31-Dec-4712 OTHERS 050498

Notice, though, that this (and your original query) will close any legitimate gaps - if a person left and was rehired, for example. You could probably fix that by looking at the action, if necessary.