How to correctly handle daylight saving times with timestamps - sql

Question

I’m trying to create a table in Postgres that stores events which occur once every full hour every day for the next couple of years. So I populated a coloumn using the following expression:
INSERT INTO tablename(time)
SELECT CAST('2013-01-01' AS DATE) + (n || ' hour')::INTERVAL
FROM generate_series(0, 100000) n;
As a datatype for this column I chose timestamp with time zone and hoped in this way daylight saving time would be automatically taken into account. (Btw, my default time zone is CET, so it's UTC+1 or UTC+2 when DST applies). As a result of the above query I get this:
2013-03-31 00:00:00 +01
2013-03-31 01:00:00 +01
2013-03-31 03:00:00 +02
2013-03-31 03:00:00 +02
2013-03-31 04:00:00 +02
...
2013-10-27 00:00:00 +02
2013-10-27 01:00:00 +02
2013-10-27 02:00:00 +01
2013-10-27 03:00:00 +01
2013-10-27 04:00:00 +01
...
The offset to UTC changes and I expected that 02:00 is left out on March 31st as this day only has 23 hours, but I don’t know why 03:00 is there twice, whereas on October 27th 02:00 is only there once instead of twice as this day has 25 hours. What I would like to achieve is that for all years on the specific day in March 2 o'clock is not skipped (I would rather put in 'n. a.' or something for the corresponding value) and that there are two entries for 3'oclock on the specific day in October (but not in March), so that I'll get a column of the following form (where 1 stands for the hour from 00:00-1:00, 2 for 1:00-2:00, etc.):
2013-03-31 1 +01
2013-03-31 2 +01
2013-03-31 3 +02
2013-03-31 4 +02
2013-03-31 5 +02
...
2013-10-27 1 +02
2013-10-27 2 +02
2013-10-27 3A +02
2013-10-27 3B +01
2013-10-27 4 +01
2013-10-27 5 +01
...
Has anybody an idea how to go about it? Am I doing something basically wrong? Is it just a matter of formatting? Do I have to write a function? Any help would be appreciated. Thank you.

Answer 1

Date and time in Postgres are stored in UTC and are converted into local time according to the zone specified by the timezone configuration.
This means that you only need to solve the representation problem. Try using AT TIME ZONE 'UTC+2' to convert UTC time to your timezone and see the result. Here's query:
SELECT (CAST('2013-03-30' AS DATE) + (n || ' hour')::INTERVAL) AT TIME ZONE 'UTC+2'
FROM generate_series(0, 1000) n;

Answer 2

The timestamp is always stored as UTC regardless of time zone settings. From the manual
For timestamp with time zone, the internally stored value is always in UTC (Universal Coordinated Time, traditionally known as Greenwich Mean Time, GMT). An input value that has an explicit time zone specified is converted to UTC using the appropriate offset for that time zone. If no time zone is stated in the input string, then it is assumed to be in the time zone indicated by the system's TimeZone parameter, and is converted to UTC using the offset for the timezone zone.
set time zone 'CET';
drop table if exists events;
create table events (
tstz timestamp with time zone
);
insert into events (tstz)
select generate_series('2013-01-01', '2013-10-28', interval '1 hour') s(tstz)
;
Notice the use of the generate_series function.
select
tstz at time zone 'UTC' as "UTC",
tstz at time zone 'CET' as "CET",
tstz at time zone 'CEST' as "CEST",
tstz as "LOCAL"
from events
where date_trunc('day', tstz) in ('2013-03-31', '2013-10-27')
order by tstz
;
UTC | CET | CEST | LOCAL
---------------------+---------------------+---------------------+------------------------
2013-03-30 23:00:00 | 2013-03-31 00:00:00 | 2013-03-31 01:00:00 | 2013-03-31 00:00:00+01
2013-03-31 00:00:00 | 2013-03-31 01:00:00 | 2013-03-31 02:00:00 | 2013-03-31 01:00:00+01
2013-03-31 01:00:00 | 2013-03-31 02:00:00 | 2013-03-31 03:00:00 | 2013-03-31 03:00:00+02
2013-03-31 02:00:00 | 2013-03-31 03:00:00 | 2013-03-31 04:00:00 | 2013-03-31 04:00:00+02
2013-03-31 03:00:00 | 2013-03-31 04:00:00 | 2013-03-31 05:00:00 | 2013-03-31 05:00:00+02
...
2013-10-26 22:00:00 | 2013-10-26 23:00:00 | 2013-10-27 00:00:00 | 2013-10-27 00:00:00+02
2013-10-26 23:00:00 | 2013-10-27 00:00:00 | 2013-10-27 01:00:00 | 2013-10-27 01:00:00+02
2013-10-27 00:00:00 | 2013-10-27 01:00:00 | 2013-10-27 02:00:00 | 2013-10-27 02:00:00+02
2013-10-27 01:00:00 | 2013-10-27 02:00:00 | 2013-10-27 03:00:00 | 2013-10-27 02:00:00+01
2013-10-27 02:00:00 | 2013-10-27 03:00:00 | 2013-10-27 04:00:00 | 2013-10-27 03:00:00+01
2013-10-27 03:00:00 | 2013-10-27 04:00:00 | 2013-10-27 05:00:00 | 2013-10-27 04:00:00+01
2013-10-27 04:00:00 | 2013-10-27 05:00:00 | 2013-10-27 06:00:00 | 2013-10-27 05:00:00+01
If a timestamp with timestamp column is selectec without using at time zone as in the LOCAL column above, it will be outputed at the server time zone at that timestamp. That is why there are missing and duplicated hours.
I think your desired output is wrong. But it is achievable with some query fu
I can't reproduce your actual output. What is the server time zone?
show time zone;
TimeZone
----------
CET