CODEWITHSUNDEEP
sqlt-sqlwindow-functions
Lionell
Lionell

Reputation: 69

Date cycle in T-SQL

I need to write a query that displays time intervals that go on without interruption. Example:

Input:

create table calc(Id int, StartDate DATE, EndDate DATE);
insert into calc values(1, '2019-01-01', '2019-01-02');
insert into calc values(2, '2019-01-02', '2019-01-03');
insert into calc values(3, '2019-01-03', '2019-01-04');
insert into calc values(4, '2019-01-14', '2019-01-15');
insert into calc values(5, '2019-01-16', '2019-01-17');
insert into calc values(6, '2019-01-17', '2019-01-18');
insert into calc values(7, '2019-01-25', '2019-01-26');
insert into calc values(8, '2019-02-03', '2019-02-04');
insert into calc values(9, '2019-02-04', '2019-02-05');
insert into calc values(10, '2019-03-01', '2019-03-02');

Output:

    StartDate , EndDate 
 '2019-01-01', '2019-01-04'
 '2019-01-14', '2019-01-15'
 '2019-01-16', '2019-01-18'
 '2019-01-25', '2019-01-26'
 '2019-02-03', '2019-02-05'
 '2019-03-01', '2019-03-02'

I think we need to use the function DATEDIFF line by line. The problem is that I don’t know how to access indexes. Or can this problem be solved much easier?

Upvotes: 2

Views: 59

Answers (2)

Yogesh Sharma
Yogesh Sharma

Reputation: 50173

Other option uses dateadd() :

select min(startdate) as startdate, max(enddate) as enddate
from (select t.*, 
             sum(case when startdate <> prev_dt then 1 else 0 end) over (order by id) as grp
      from (select t.*,
                   dateadd(day, 1, lag(t.startdate) over (order by t.id)) as prev_dt
            from table t
           ) t
     ) t
group by grp
order by startdate;

Upvotes: 1

Gordon Linoff
Gordon Linoff

Reputation: 1270463

This is a type of gaps-and-islands problem. Use a lag() to see if adjacent rows overlap. Do a cumulative sum of gaps, and then aggregate:

select min(startdate), max(enddate)
from (select c.*,
             sum(case when prev_ed = startdate then 0 else 1 end) over
                 (order by startdate) as grp
      from (select c.*,
                   lag(enddate) over (order by startdate) as prev_ed
            from calc c
           ) c
     ) c
group by grp
order by min(startdate);

Here is a db<>fiddle.

Upvotes: 4

Related Questions