Re: [HACKERS] cidr'

From: Bruce Momjian <maillist(at)candle(dot)pha(dot)pa(dot)us>
To: tgl(at)sss(dot)pgh(dot)pa(dot)us (Tom Lane)
Cc: pgsql-hackers(at)hub(dot)org
Subject: Re: [HACKERS] cidr'
Date: 1998-07-21 15:45:18
Views: Raw Message | Whole Thread | Download mbox
Lists: pgsql-hackers

> Bruce Momjian <maillist(at)candle(dot)pha(dot)pa(dot)us> writes:
> >> Would it make sense to use atttypmod to distinguish several different
> >> subtypes of CIDR? "4 bytes", "4 bytes + mask", "6 bytes", "6 bytes
> >> + mask" seem like interesting possibilities.
> > Yes, that is the proper way to go, though atttypmod is something on
> > column, not on each data row. It is specified when the column is
> > created.
> Right, that's what I had in mind. If you *know* that every entry in
> your table only needs IPv4, you can specify that when making the table
> and save a couple of bytes per entry.
> The alternative solution is to make CIDR a variable-length type, but
> I think the overhead of that would be as much or more than the possible
> savings, no?
> I don't know whether having multiple top-level types would be better
> or worse than one type with a subtype code.

The byte size is really not an issue to me. You can do ip6 and still
put it in eight bytes. If you make it a variable-lengh type, you have
the length on each field, and that is four bytes right there, so you are
better doing eight bytes from the start.

ip4 5 btyes(4 + precision)
ip6 7 bytes(6 + precision)

If you want ip6 now, just take eight bytes and make it a fixed length.
The backend it going to round the disk storage of 5 bytes up to eight
anyway, unless the next field is int2 or char1.

Bruce Momjian | 830 Blythe Avenue
maillist(at)candle(dot)pha(dot)pa(dot)us | Drexel Hill, Pennsylvania 19026
+ If your life is a hard drive, | (610) 353-9879(w)
+ Christ can be your backup. | (610) 853-3000(h)

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Bruce Momjian 1998-07-21 15:47:40 Re: Complexity of contrib types
Previous Message Hannu Krosing 1998-07-21 15:31:00 Re: Complexity of contrib types