SQL/Postgres datetime division / normalizing

Question

I have this activity table

+--------------+------------------+
| Field        | Type             |
+--------------+------------------+
| id           | int(11) unsigned |
| start_date   | timestamp        |
| end_date     | timestamp        |
| ...          |                  |
+--------------+------------------+

I need a view which groups these activities by start_date by DAY, but in such a way that, if the end_date is not in the same day as start_date, the view contain the entry again but with the start_date set to 00:00 of the next day.. (and so on, repeated as many times as needed until the start_date is in the same day as the end_date)

As an example:

if the activity table contains:

+--------------+----------------------------+----------------------------+
| id           | start_date                 | end_date                   |
+--------------+----------------------------+----------------------------+
|  1           | 2014-12-02 14:12:00+00     | 2014-12-03 06:45:00+00     |   
|  2           | 2014-12-05 15:25:00+00     | 2014-12-05 07:29:00+00     |                                
+--------------+----------------------------+----------------------------+

The view should contain:

+--------------+----------------------------+----------------------------+
| activity_id  | start_date                 | end_date                   |
+--------------+----------------------------+----------------------------+
|  1           | 2014-12-02 14:12:00+00     | 2014-12-02 23:59:59+00     |   
|  1           | 2014-12-03 00:00:00+00     | 2014-12-03 06:45:00+00     |
|  2           | 2014-12-05 15:25:00+00     | 2014-12-05 07:29:00+00     |                                  
+--------------+----------------------------+----------------------------+

Any help would be greatly appreciated!

PS: I'm using postgresql

Denis de Bernardy · Accepted Answer · 2014-12-10 11:16:35Z

To get the needed rows, start by using a set returning function along with a lateral join. From there, use CASE statements and date arithmetics to pull out the relevant values.

Here's an example to get you started:

with data as (
  select id, start_date, end_date
  from (values
    (1, '2014-12-02 14:12:00+00'::timestamptz, '2014-12-03 06:45:00+00'::timestamptz),
    (2, '2014-12-05 15:25:00+00'::timestamptz, '2014-12-05 07:29:00+00'::timestamptz)
  ) as rows (id, start_date, end_date)
)
select data.id,
      case days.d = date_trunc('day', data.start_date)
        when true then data.start_date
        else days.d
      end as start_date,
      case days.d = date_trunc('day', data.end_date)
        when true then data.end_date
        else days.d + interval '1 day' - interval '1 sec'
      end as end_date
from data
join generate_series(
      date_trunc('day', data.start_date),
      date_trunc('day', data.end_date),
      '1 day'
      ) as days (d)
      on days.d >= date_trunc('day', data.start_date)
      and days.d <= date_trunc('day', data.end_date)

 id |       start_date       |        end_date        
----+------------------------+------------------------
  1 | 2014-12-02 15:12:00+01 | 2014-12-02 23:59:59+01
  1 | 2014-12-03 00:00:00+01 | 2014-12-03 07:45:00+01
  2 | 2014-12-05 16:25:00+01 | 2014-12-05 08:29:00+01
(3 rows)

As an aside, depending on what you're doing, it might make more sense for you to use a date range:

with data as (
  select id, start_date, end_date
  from (values
    (1, '2014-12-02 14:12:00+00'::timestamptz, '2014-12-03 06:45:00+00'::timestamptz),
    (2, '2014-12-05 07:25:00+00'::timestamptz, '2014-12-05 15:29:00+00'::timestamptz)
  ) as rows (id, start_date, end_date)
)
select data.id,
      tstzrange(data.start_date, data.end_date)
from data;

 id |                      tstzrange                      
----+-----------------------------------------------------
  1 | ["2014-12-02 15:12:00+01","2014-12-03 07:45:00+01")
  2 | ["2014-12-05 08:25:00+01","2014-12-05 16:29:00+01")
(2 rows)

Collectives™ on Stack Overflow

SQL/Postgres datetime division / normalizing

1 Answer 1

Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related