Re: [PATCH] indexability of << operator for inet/cidr

Attached is a patch that explains to optimizer that its possible to use
index when performing a << b where a is an inet/cidr value and b is a
constant.

Indexpath generated for such an expression is this:
(a > network(b)) and (a <= set_masklen(broadcast(b, 32)))

Since this is my first time delving in the guts of postgres, someone
definitely should review it :)

I mostly based my code on prefix_quals function for string types.

Thanks

Alex Pilosov <alex(at)pilosoft(dot)com> writes:
> Indexpath generated for such an expression is this:
> (a > network(b)) and (a <= set_masklen(broadcast(b, 32)))

What happens to that set_masklen thing for IPv6?

If the network.c code were exporting a function that made this value,
I'd not worry; but I don't like wiring an IPv4 assumption into code far
away in the planner. Can't we do better here? Perhaps move the
generation of the indexscan bound values into a subroutine in network.c?

regards, tom lane

Your patch has been added to the PostgreSQL unapplied patches list at:

http://candle.pha.pa.us/cgi-bin/pgpatches

I will try to apply it within the next 48 hours.

> Attached is a patch that explains to optimizer that its possible to use
> index when performing a << b where a is an inet/cidr value and b is a
> constant.
>
> Indexpath generated for such an expression is this:
> (a > network(b)) and (a <= set_masklen(broadcast(b, 32)))
>
> Since this is my first time delving in the guts of postgres, someone
> definitely should review it :)
>
> I mostly based my code on prefix_quals function for string types.
>
> Thanks
>

Content-Description:

[ Attachment, skipping... ]

>
> ---------------------------(end of broadcast)---------------------------
> TIP 5: Have you checked our extensive FAQ?
>
> http://www.postgresql.org/users-lounge/docs/faq.html

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026

On Thu, 14 Jun 2001, Tom Lane wrote:

> Alex Pilosov <alex(at)pilosoft(dot)com> writes:
> > Indexpath generated for such an expression is this:
> > (a > network(b)) and (a <= set_masklen(broadcast(b, 32)))
>
> What happens to that set_masklen thing for IPv6?
>
> If the network.c code were exporting a function that made this value,
> I'd not worry; but I don't like wiring an IPv4 assumption into code far
> away in the planner. Can't we do better here? Perhaps move the
> generation of the indexscan bound values into a subroutine in network.c?

Good point. I already rewrote it, but I am going to send it in tomorrow, I
want to resync to HEAD, since some of network.c was taken in and I want to
have a clean patch for you guys :)

Thanks
-alex

Patch removed from queue, awaiting updated version.

> Attached is a patch that explains to optimizer that its possible to use
> index when performing a << b where a is an inet/cidr value and b is a
> constant.
>
> Indexpath generated for such an expression is this:
> (a > network(b)) and (a <= set_masklen(broadcast(b, 32)))
>
> Since this is my first time delving in the guts of postgres, someone
> definitely should review it :)
>
> I mostly based my code on prefix_quals function for string types.
>
> Thanks
>

Content-Description:

[ Attachment, skipping... ]

>
> ---------------------------(end of broadcast)---------------------------
> TIP 5: Have you checked our extensive FAQ?
>
> http://www.postgresql.org/users-lounge/docs/faq.html

--
Bruce Momjian | http://candle.pha.pa.us
pgman(at)candle(dot)pha(dot)pa(dot)us | (610) 853-3000
+ If your life is a hard drive, | 830 Blythe Avenue
+ Christ can be your backup. | Drexel Hill, Pennsylvania 19026