Date: Wed, 26 Mar 2003 09:50:24 +0000 (GMT)
From: Peter Childs <blue.dragon@blueyonder.co.uk>
Cc: pgsql-sql@postgresql.org
Subject: Re: Does anyone use TO_CHAR(INTERVAL)?
In-Reply-To: <3E81620E.20604@klaster.net>
Message-ID: <Pine.LNX.4.44.0303260929090.5091-100000@RedDragon.Childs>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII

On Wed, 26 Mar 2003, Tomasz Myrta wrote:

> 
> Thanks for noticing this, these results are really strange for me.
> I should remember NOT to use months when working with intervals. It 
> looks like intervals contain 2 totally independent values inside.
> 
> I have one more question to this thread - Does anyone use intervals 
> month value?

	Since all months are not equal in length how do else do you come 
up with an interval that is 1 month since us humans like to use months.
	As you note Postgres trys to ovide using Months if at all possible 
because 1 month <> 1 month hense.

select (cast('2-2-2003' as date) + cast('1 month' as interval)) - 
cast('2-2-2003' as date);
 ?column?
----------
 28 days
(1 row)

but if all you say is 1 month from now how many days you month you mean 
will mean a different number of days. Oh one more good one is....

select cast('1 year' as interval) - cast('365 days' as interval);
     ?column?
------------------
 1 year -365 days
(1 row)

	It would seam that a year is 12 months but not 365 days which 
while this is true (due to leap years) most people would never think like 
that. Its seams that postgres will not tell you how many days there are in 
a year but this seams even more stupid 

select cast('1 month' as interval)/2 * 2;
 ?column?
----------
 360 days
(1 row)

oh and just when you thought somthing was really straght forward you 
get...

select cast('1 year' as interval)/24 * 24;
 ?column?
----------
 360 days
(1 row)

	This reminds me of a calculator that comes with MS Windows....
	How ever the implementation in postgres is the best I've sean so 
long as you don't do anything stupid!

Peter Childs