Horton

Re: generate syscache info automatically

John Naylor <john.naylor@enterprisedb.com>

View thread View on postgr.es
From: John Naylor <john.naylor@enterprisedb.com>
To: Peter Eisentraut <peter@eisentraut.org>
Cc: pgsql-hackers <pgsql-hackers@postgresql.org>
Date: 2023-09-02T10:37:50Z
Lists: pgsql-hackers

Attachments

I wrote:

> + # XXX This one neither, but if I add it to @skip, PerfectHash will
fail. (???)
> + #FIXME: AttributeRelationId
>
> I took a quick look at this, and attached is the least invasive way to
get it working for now, which is to bump the table size slightly. The
comment says doing this isn't reliable, but it happens to work in this
case. Playing around with the functions is hit-or-miss, and when that
fails, somehow the larger table saves the day.

To while away a rainy day, I poked at this a bit more and found the input
is pathological with our current methods. Even with a large-ish exhaustive
search, the two success are strange in that they only succeeded by
accidentally bumping the table size up to where I got it to work before
(77):

With multipliers (5, 19), it recognizes that the initial table size (75) is
a multiple of 5, so increases the table size to 76, which is a multiple of
19, so it increases it again to 77 and succeeds.

Same with (3, 76): 75 is a multiple of 3, so up to 76, which of course
divides 76, so bumps it to 77 likewise.

Turning the loop into

a = (a ^ c) * 257;
b = (b ^ c) * X;

...seems to work very well.

In fact, now trying some powers-of-two for X before the primes works most
of the time with our inputs, even for some unicode normalization functions,
on the first seed iteration. That likely won't make any difference in
practice, but it's an interesting demo. I've attached these two draft ideas
as text.

--
John Naylor
EDB: http://www.enterprisedb.com

Commits

  1. Generate syscache info from catalog files

  2. Fill in more of ObjectProperty

  3. Correct ObjectProperty entry for transforms

  4. genbki.pl: Factor out boilerplate generation

  5. Restructure DECLARE_INDEX arguments

  6. Update DECLARE_INDEX documentation

  7. Replace hardcoded switch in object_exists() with a lookup table.