pg.ddx.io  pgsql-sql@postgresql.org mailing list archive  
help / color / mirror / Atom feed
From: Joseph Barillari <jbarilla@princeton.edu>
To: Josh Berkus <josh@agliodbs.com>
Cc: pgsql-sql@postgresql.org
Subject: Re: Casting dates
Date: Mon, 29 Apr 2002 12:19:21 -0400
Message-ID: <m3lmb6fpfa.fsf@washer.barillari.org> (raw)
In-Reply-To: <web-1387712@davinci.ethosmedia.com>
References: <web-1387712@davinci.ethosmedia.com>

>>>>> "JB" == Josh Berkus <josh@agliodbs.com> writes:

    JB> Joe,
    >> Does anyone know if it's possible to calculate the number of
    >> days (or weeks, seconds, minutes, or what-have-you) in a given
    >> interval without casting to EPOCH and performing manipulations
    >> there?
    >> 
    >> PostgreSQL usually returns intervals created by subtraction in
    >> days and smaller fragments, I would like to know if it's
    >> possible for it to return years. (Integer division of the
    >> number of days by 365 would produce an almost-correct result,
    >> but it's rather inelegant.)

    JB> Unfortunately, you've hit one of Postgres implementation
    JB> limits.  Multiplying and dividing INTERVAL values, while
    JB> covered by the SQL92 spec, has not been completely implemented
    JB> in Postgres.  That is, it should be possible for you to:

    JB> SELECT INTERVAL '3 days 5 hours' / INTERVAL '1 minute'

    JB> However, implementing this has been a challenge for all SQL-DB
    JB> designers.  I don't know of any DB that does handle INTERVAL
    JB> division.
 

    JB> It's a surmountable challenge, though, but requires somebody
    JB> to take the lead in a) working out the logic, and b) writing
    JB> the code to implement it.  A) is not a trivial task, either
    JB> ... while 60 minutes / 1 second is obvious, how about 4 months
    JB> / 4 days?  Months are not constant in length.

For the purposes of this application (because it deals in relatively
short intervals, it's a calendaring application), I just divide the
day count by 365 and throw away the remainder.

For example, 

select date_part('day', (timestamp 'jan 1, 1996' -
                        timestamp 'jan 1, 1992')); 

is 1461 days, or 1+365*4 (1992 is a leap year). Integer division of
1461/365 produces the correct year count. This doesn't work for longer
intervals:

cal=> select date_part('day', (timestamp 'jan 1, 2600' 
              - timestamp 'jan 1, 1000'))/365;
     ?column?     
------------------
 1601.06301369863  <-- an error of +1 year, due to accumulated leap days

Obviously, this naive method won't work for astronomical time
intervals, but for my purposes, few people have calendars that
stretch across millennia.

--Joe

view thread (3+ messages)

Message-ID: <m3lmb6fpfa.fsf@washer.barillari.org>
Permalink:  ../m3lmb6fpfa.fsf@washer.barillari.org/
Also on:    postgresql.org/message-id/m3lmb6fpfa.fsf@washer.barillari.org

 · 

reply

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Reply to all the recipients using the --to and --cc options:
  reply via email

  To: pgsql-sql@postgresql.org
  Cc: jbarilla@princeton.edu, josh@agliodbs.com
  Subject: Re: Casting dates
  In-Reply-To: <m3lmb6fpfa.fsf@washer.barillari.org>

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox