DB2 SQL配对日期

Question

I am trying to pair up dates that I am getting from my SQL. The output at the moment looks something like this:

 start_date   end_date 
 2015-02-02  2015-02-02  
 2015-02-02  2015-02-03    
 2015-02-03  2015-02-03  
 2015-04-12  2015-02-12

I would like the ouput to be paired up so that the smallest and the biggest date of a date group is chosen, so that the output would look like this:

 start_date   end_date 
 2015-02-02  2015-02-03    
 2015-04-12  2015-02-12

Using the first response I get something like this, I believe I have formatted this wrong, I am getting the same date pairs as before, but it does run.

select min(date), max(date)
from (select date,
         sum(case when sum(inc) = 0 then 1 else 0 end) over (order by   date desc) as grp
  from (select t1.datev as date, 1 as inc
        from  table2 t1, 
              table3 c, 
              table4 cr

where t1.datev between date(c.e_start_date) and date(c.e_end_date)
and   t1.datev not in (select date(temp.datev) from mdmins11.temp temp where temp.number < 4000 and temp.organisation_id = 11111)
and   c.tp_cd in (1,6)
and   cr.from_id = c.id
and   cr.organisation_id = 11111

        union all
        select t.datev as date, -1 as inc
        from  table1 t,
              table3 c, 
              table4 cr

where t.datev between date(c.e_start_date) and date(c.e_end_date)
and   t.datev not in (select date(temp.datev) from mdmins11.temp temp where temp.number < 4000 and temp.organisation_id = 11111)
and   c.tp_cd in (1,6)
and   cr.from_id = c.id
and   cr.organisation_id = 11111
       ) t
  group by date
 ) t
group by grp;

Answer 1

One method is to determine where groups of non-overlapping dates start. For this, you can use not exists . Then count up this flag over all records. This uses window functions. However, this poses problems because you have multiple starts on the same date.

Another method is to keep track of starts and stops and note where the sum is zero. These represent boundaries between groups. The following should work on your data:

select min(date), max(date)
from (select date,
             sum(case when sum(inc) = 0 then 1 else 0 end) over (order by date desc) as grp
      from (select start_date as date, 1 as inc
            from table
            union all
            select end_date as date, -1 as inc
            from table
           ) t
      group by date
     ) t
group by grp;

This type of problem is made more complicated when duplicate values are allowed on a given date. Given only the dates, this is challenging. With a separate unique id for each row, then there are more robust solutions.

EDIT:

A more robust solution:

select min(start_date), max(end_date)
from (select t.*, sum(StartGroupFlag) over (order by start_date) as grp
      from (select t.*,
                   (case when not exists (select 1
                                          from table t2
                                          where t2.start_date < t.start_date and
                                                t2.end_date >= t.start_date
                                         )
                         then 1 else 0
                    end) as StartGroupFlag
            from table t
           ) t
      ) t
group by grp;