Re: Absolute value of intervals

Joshua Berry wrote:
> On Mon, Nov 2, 2009 at 8:35 AM, Jasen Betts <jasen(at)xnet(dot)co(dot)nz> wrote:
> |
> | for this case: convert to seconds and then do abs.
> |
> | ?select * from enviados e, recibidos r where abs(extract ( epoch from
> | ?(e.fecha - r.fecha) )) < 1
>
> Cheers for that. The query cost is pretty heavy, but the same as the
> following query which involves two WHERE sections:
>
> | select * from enviados e, recibidos r where (e.fecha - r.fecha) <
> | interval '1 second' AND (r.fecha - e.fecha) < interval '1 second'
>
> But both have a higher cost than the following query with "OVERLAPS"
>
> | select * from enviados e, recibidos r where (r.fecha + interval '1
> | seconds', r.fecha - interval '1 seconds') OVERLAPS (e.fecha, e.fecha);
>
> We're only talking about a 4% difference on estimated costs in the
> billions, but it made me wonder if I should be using OVERLAPS more
> regularly as it may be more highly optimized.
>
> | thanks for sparking an interesting discussion.
>
> Thanks. I feel a bit humbled by thinking that I could help add an
> operator that didn't exist simply because no one had gotten around too
> it! I realize now that it's a lot more complex of an issue than I
> thought it was.

I have added a C comment the references this discussion so we remember
why the function does not exist:

Add C comment about why there is no interval_abs(): it is unclear what
value to return:

http://archives.postgresql.org/pgsql-general/2009-10/msg01031.php
http://archives.postgresql.org/pgsql-general/2009-11/msg00041.php

--
Bruce Momjian <bruce(at)momjian(dot)us> http://momjian.us
EnterpriseDB http://enterprisedb.com

+ If your life is a hard drive, Christ can be your backup. +