Re: Timestamp to time_t

From: "David E(dot) Wheeler" <david(at)kineticode(dot)com>
To: Andrew Gierth <andrew(at)tao11(dot)riddles(dot)org(dot)uk>
Cc: "Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov>, <pgsql-hackers(at)postgresql(dot)org>, "Scott Mohekey" <scott(dot)mohekey(at)telogis(dot)com>
Subject: Re: Timestamp to time_t
Date: 2009-09-15 20:16:38
Message-ID: 57856C2B-4524-4FB2-9D1B-B06A4E8A0666@kineticode.com
Views: Raw Message | Whole Thread | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

On Sep 15, 2009, at 11:01 AM, Andrew Gierth wrote:

> If you want to store both a timestamp and an associated timezone you
> can do
> it right now, using a composite type or two columns, with the
> advantage that
> you get semantics that you can rely on.

How would a composite work in practice? Can you index it on the
timestamp? Or would you have to use two columns for that?

I could see a real advantage to a type that stored the TZ with which
it was created, with the ability to fetch it back out. Internally the
data could be stored just like it is with timestamptz, and by default,
perhaps, it would display in $PGTZ, but if $PGTZ was set to a value
like "original" or something, it should display the originals. Now
*that* would be really useful IMHO.

Best,

David

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Kevin Grittner 2009-09-15 20:20:51 Re: [BUGS] BUG #5053: domain constraints still leak
Previous Message David Fetter 2009-09-15 19:49:27 Re: WIP: generalized index constraints