generate_series over daylight savings change - varied results depending on server timezone

Question

When my postgres server is in the America/New_York timezone, or I use SET SESSION TIME ZONE 'America/New_York' , generate_series respects the daylight savings change and I can get the proper epoch or point in time that I want:

postgres=# SET SESSION TIME ZONE 'America/New_York';
SET
postgres=#
postgres=# with seq(ts) as (select * from generate_series('2018-11-03 00:00:00.000 -04:00', '2018-11-06 13:40:39.067 -05:00', '1d'::interval))
postgres-# select ts, extract(epoch from ts) as epoch, ts at time zone 'America/New_York' as eastern
postgres-# from seq;
           ts           |   epoch    |       eastern
------------------------+------------+---------------------
 2018-11-03 00:00:00-04 | 1541217600 | 2018-11-03 00:00:00
 2018-11-04 00:00:00-04 | 1541304000 | 2018-11-04 00:00:00
 2018-11-05 00:00:00-05 | 1541394000 | 2018-11-05 00:00:00
 2018-11-06 00:00:00-05 | 1541480400 | 2018-11-06 00:00:00
(4 rows)

But when my server is in UTC as it is in production, generate_series does not respect the daylight savings change:

postgres=# SET SESSION TIME ZONE 'UTC';
SET
postgres=# with seq(ts) as (select * from generate_series('2018-11-03 00:00:00.000 -04:00', '2018-11-06 13:40:39.067 -05:00', '1d'::interval))
postgres-# select ts, extract(epoch from ts) as epoch, ts at time zone 'America/New_York' as eastern
postgres-# from seq;
           ts           |   epoch    |       eastern
------------------------+------------+---------------------
 2018-11-03 04:00:00+00 | 1541217600 | 2018-11-03 00:00:00
 2018-11-04 04:00:00+00 | 1541304000 | 2018-11-04 00:00:00
 2018-11-05 04:00:00+00 | 1541390400 | 2018-11-04 23:00:00
 2018-11-06 04:00:00+00 | 1541476800 | 2018-11-05 23:00:00
(4 rows)

Notice 11/4 and 11/5 have not been adjusted for the DST change.

Is there any way around this without setting the session time zone when I query?

Using postgres 9.6...

Answer 1

The session time zone is what controls how an interval of 1 day is interpreted, so I think the answer to your question is simply, no.

The Postgres docs explain this as follows:

When adding an interval value to (or subtracting an interval value from) a timestamp with time zone value, the days component advances or decrements the date of the timestamp with time zone by the indicated number of days. Across daylight saving time changes (when the session time zone is set to a time zone that recognizes DST), this means interval '1 day' does not necessarily equal interval '24 hours' . For example, with the session time zone set to CST7CDT , timestamp with time zone '2005-04-02 12:00-07' + interval '1 day' will produce timestamp with time zone '2005-04-03 12:00-06' , while adding interval '24 hours' to the same initial timestamp with time zone produces timestamp with time zone '2005-04-03 13:00-06' , as there is a change in daylight saving time at 2005-04-03 02:00 in time zone CST7CDT .

As far as I am aware, there's no other mechanism of how to tell an interval to be interpreted in a specific time zone, other than setting that as the session time zone. In other words, I think you'd be looking for something like '1 day tz America/New_York'::interval and I don't believe such a syntax exists.

Answer 2

You can set the timezone automatically upon connection for specific users (ALTER USER), or for a specific database (ALTER DATABASE).

The only other way I can think of is with a function.

BEGIN;

CREATE FUNCTION f1() RETURNS SETOF timestamptz
AS $$
SELECT generate_series('2018-11-03 00:00:00-04', '2018-11-06 00:00:00-05', interval '1 day');
$$
LANGUAGE sql
SET timezone = 'America/New_York';                                                                                                               

SET timezone = 'UTC';
SELECT generate_series('2018-11-03 00:00:00-04', '2018-11-06 00:00:00-05', interval '1 day');
SELECT * FROM f1();

ROLLBACK;

Produces:

    generate_series
------------------------
 2018-11-03 04:00:00+00
 2018-11-04 04:00:00+00
 2018-11-05 04:00:00+00
 2018-11-06 04:00:00+00
(4 rows)

           f1
------------------------
 2018-11-03 04:00:00+00
 2018-11-04 04:00:00+00
 2018-11-05 05:00:00+00
 2018-11-06 05:00:00+00
(4 rows)

Answer 3

You can try generating the series in the desired timezone, then converting to UTC afterwards:

select generate_series('2018-11-03 00:00:00.000', '2018-11-06 13:40:39.067', '1d'::interval)::timestamp at time zone 'America/New_York'

Output:

2018-11-03 04:00:00+00
2018-11-04 04:00:00+00
2018-11-05 05:00:00+00
2018-11-06 05:00:00+00