On 2013-01-25, Paul Jones <pbj@cmicdo.com> wrote:
> Is it possible for LC_TIME locale to affect the format with which
> timestamps are input?
> I have DB2 CSV dumps with timestamps like '2003-10-21-22.59.44.000000'
All the non-digit symbols between "21" and the "44" look unusual
> that I want to load into Postgres with \copy. I would like to eliminate
> the sed scripts that convert the timestamps in order to speed up the load.
> (I know I could stage it through a temp table and use to_timestamp()
> but that requires a temp table for each real table, which is not viable
> w.r.t. our project goals).
could you use a text column and then do alter table?
> I created a special locale with the DB2 timestamp format defined and did
>
> set lc_time='en_DB.UTF-8';
what's "DB" ?
> It didn't affect anything, in or out with Postgres. I know the locale
> works because date(1) displays the DB2 format correctly.
That feels like a bug somewhere
AIUI date uses strftime("%c",...
and strftime() doesn't support fractional seconds, so postgres can't
use it.
.
--
⚂⚃ 100% natural