On Tue, 18 Jan 2005 16:15:57 +1100, Neil Conway <[EMAIL PROTECTED]>
wrote:
>900 line functions are almost
>universally bad
Amen. So you might be interested in reviewing
http://archives.postgresql.org/pgsql-hackers/2004-06/msg00398.php ;-)
Servus
Manfred
---(end of broa
On Tue, 08 Jun 2004 13:13:01 -0400, Tom Lane <[EMAIL PROTECTED]> wrote:
>Possibly the relsize axis ought to be measured on a log scale, or
>something like that, but that didn't seem to work nicely when relsize
>approaches zero.
In my experiments I used log(relsize) on the x axis, and I don't think
On Tue, 08 Jun 2004 11:13:43 -0400, Tom Lane <[EMAIL PROTECTED]> wrote:
>This patch seems to do considerably more violence to the equations than
>is needed to cover that oversight, though. The old behavior was
>intentionally nonlinear in relsize; this is not.
The comment says "entirely ad-hoc" an
The comment describing cost_nonsequential_access() says that the two
functions "meet in the middle". They meet at random_page_cost/2,
however, not in the middle between 1 and random_page_cost. For
random_page_cost < 2 the result can be less than 1 for relpages near
effective_cache_size. I don't
. rename variables
. cur_buffer -> dst_buffer
. ToPage -> dst_page
. cur_page -> dst_vacpage
. move variable declarations into block where variable is used
. various Asserts instead of elog(ERROR, ...)
. extract functionality from repair_frag() into new routines
. mo
Basically replaces (*a).b with a->b as it is everywhere else in
Postgres.
Servus
Manfred
diff -Ncr ../base/src/backend/storage/page/bufpage.c src/backend/storage/page/bufpage.c
*** ../base/src/backend/storage/page/bufpage.c Sat Nov 29 20:51:57 2003
--- src/backend/storage/page/bufpage.c Wed Jun
On Sun, 23 May 2004 17:32:36 -0400, Tom Lane <[EMAIL PROTECTED]> wrote:
>I took out the TupleCount typedef and went
>back to using doubles for the tuple counts; this is more consistent with
>the coding style used elsewhere, and I really doubt that it's any
>slower.
Performance was not the primary
On Sun, 23 May 2004 13:16:49 -0400, Tom Lane <[EMAIL PROTECTED]> wrote:
>> $ gcc --version
>> 2.7.2.1
>
>Man, that's ancient.
Don't say that before you have seen my bike :-)
>How do you feel about upgrading?
This is on my notebook. I'm a little afraid of changing major parts of
the system. Onc
This patch implements the new tuple sampling method as discussed on
-hackers and -performance a few weeks ago.
"Large DB" -hackers 2004-04-02
"query slows down with more accurate stats" -perform 2004-04-13
"Tuple sampling" -hackers 2004-04-19
Servus
Manfred
diff -rcN ../base/src/backend/commands
Yesterday's snapshot couldn't be compiled from a separate build
directory, and after configure --without-docdir ... make install failed,
because it tried to mkdir /postgresql/html.
The enclosed patch fixes these problems, at least for me.
Servus
Manfred
diff -Ncr ../base/src/Makefile.global.in .
On Wed, 28 Apr 2004 12:02:44 -0400, Alvaro Herrera
<[EMAIL PROTECTED]> wrote:
>In fact, I think we should mark ERROR as aborting the whole transaction
>tree, and create a new level which would abort the innermost
>subtransaction. We would then change whatever is appropiate to the new
>elevel. Doi
On Mon, 05 Apr 2004 18:30:29 -0400, Tom Lane <[EMAIL PROTECTED]> wrote:
>> noise-contributing factors.
>
>I think it would have to be visibility-bit updates. Can you try it with
>a pre-vacuumed relation, so that there is no slowdown for updates?
I'd like to avoid VACUUM to keep the dead tuples.
On Mon, 05 Apr 2004 15:37:07 -0400, Tom Lane <[EMAIL PROTECTED]> wrote:
>I wouldn't bother with a GUC variable for the production patch.
Among other things the GUC variable will be thrown out for the final
version.
>> Once a block is selected for inspection, all tuples of this
>> block are access
Here is the first preview of the two-stage tuple sampling as discussed
on -hackers last week. It is for review only, please do not apply.
As I didn't have a current CVS checkout when I worked on the patch, it
is based on 7.4.2 sources. I hope this is ok for the purpose of
reviewing.
The old sam
On Mon, 01 Dec 2003 13:32:10 -0500, Tom Lane <[EMAIL PROTECTED]> wrote:
>Manfred Koizar <[EMAIL PROTECTED]> writes:
>> comparetup_index() compares two IndexTuples. The structure
>> IndexTupleData consists basically of not much more than an ItemPointer,
>> an
On Mon, 1 Dec 2003 00:02:54 -0500 (EST), Bruce Momjian
<[EMAIL PROTECTED]> wrote:
>Tom Lane wrote:
>> Bruce Momjian <[EMAIL PROTECTED]> writes:
>> >> And if it doesn't help index
>> >> creation speed, at least the resulting index has better correlation.
... which has been shown by the example in
On Fri, 21 Nov 2003 03:09:02 -0500, Neil Conway <[EMAIL PROTECTED]> wrote:
>attachment; filename=weird_regression.diffs
This was caused by a small oversight in ALTER TABLE ... ADD COLUMN:
diff -ruN ../base/src/backend/commands/tablecmds.c src/backend/commands/tablecmds.c
--- ../base/src/backend/c
On Thu, 20 Nov 2003 21:31:20 -0500, Neil Conway <[EMAIL PROTECTED]>
wrote:
>I searched a couple different archives of -hackers, but I didn't find
>anything that looked relevant.
It's on -patches:
http://archives.postgresql.org/pgsql-patches/2002-08/msg00583.php
Servus
Manfred
--
On Tue, 30 Sep 2003 20:00:22 -0400, Neil Conway <[EMAIL PROTECTED]>
wrote:
>This patch adds a new GUC var, "default_use_oids"
Shouldn't it be honoured by CREATE TABLE AS SELECT ... ?
This checkin might contain hints to the places that have to be
modified:
2003-01-23 tgl
* src/backend/executor/
On Wed, 8 Oct 2003 11:33:24 -0400 (EDT), Bruce Momjian
<[EMAIL PROTECTED]> wrote:
>Can someone explain what was broken?
COPY FROM removed backslashes before comparing the input to the
external null representation. (It had a hard-wired special code path
that allowed \N to be recognized.) The text
On Wed, 08 Oct 2003 11:31:30 +0200, I wrote:
>There haven't been too much changes in this area between 7.3 and 7.4.
Here is the patch for 7.3.4 ...
Bruce, I noticed that the original patch submission didn't contain
anything useful as a cvs log message:
Make COPY FROM a bit more compatible with C
On Sun, 05 Oct 2003 19:12:50 -0400, Tom Lane <[EMAIL PROTECTED]>
wrote:
>it seems we have to compare the null representation string to the
>pre-debackslashing input.
Here is a patch that does this and adds a few regression tests.
>(This is probably fairly easy to make happen
>in CVS tip, but it m
On Thu, 11 Sep 2003 23:17:13 -0400 (EDT), Bruce Momjian
<[EMAIL PROTECTED]> wrote:
>> The stats collector is on by default (of course, that doesn't apply to
>> "make installcheck"...)
By doing "appropriate" changes to your installation I guess you can
break almost every test ;-)
>Oh I see now, ni
With pg_autovacuum becoming increasingly popular it's important to
have a working stats collector. This test is able to discover the
problem that was present in 7.4 Beta 2.
Servus
Manfred
diff -ruN ../base/src/test/regress/expected/stats.out
src/test/regress/expected/stats.out
--- ../base/src/t
On Mon, 8 Sep 2003 13:02:11 -0400 (EDT), Bruce Momjian
<[EMAIL PROTECTED]> wrote:
>
>Manfred, can I get a description for this patch? Thanks.
Try to reduce confusion about what is a lock method identifier, a lock
method control structure, or a table of control structures.
. Use type LOCKMASK whe
On Mon, 8 Sep 2003 11:31:05 +0200, "Zeugswetter Andreas SB SD"
<[EMAIL PROTECTED]> wrote:
>> As Tom mentioned, we might not want to keep the tid's in order after the
>> index is created because he wants the most recent tid's first, so the
>> expired ones migrate to the end.
>
>But on average this a
On Mon, 8 Sep 2003 00:51:45 -0400 (EDT), Bruce Momjian
<[EMAIL PROTECTED]> wrote:
>
>This has been saved for the 7.5 release:
>
> http:/momjian.postgresql.org/cgi-bin/pgpatches2
Here is the combined patch. Prior patches in this thread are hereby
obsolete.
Servus
Manfred
diff -ruN ../base/
On Sun, 07 Sep 2003 12:23:28 -0400, Tom Lane <[EMAIL PROTECTED]>
wrote:
>Maybe so, but it would degrade the performance in the unique-index case
>if we do it as the TODO is worded.
The patch would only hurt with a unique index, if there are lots of
duplicate tuples at CREATE INDEX time.
>My own o
On Sun, 7 Sep 2003 11:43:42 -0400 (EDT), Bruce Momjian
<[EMAIL PROTECTED]> wrote:
>I assume this completes this TODO:
>
> * Order duplicate index entries by tid for faster heap lookups
I don't think so, because the patch does nothing to keep the sort
order once the index is initially created
On Sun, 07 Sep 2003 10:19:07 -0400, Tom Lane <[EMAIL PROTECTED]>
wrote:
>> +#define BITS_OFF(i) ~(1 << (i))
>
>I'd put another pair of parens around that. Also, it might be worth
>moving into a header file. There is at least one place in proc.c that
>manipulates lock masks using explicit shifts,
Try to reduce confusion about what is a lock method identifier, a lock
method control structure, or a table of control structures.
. Use type LOCKMASK where an int is not a counter.
. Get rid of INVALID_TABLEID, use INVALID_LOCKMETHOD instead.
. Use INVALID_LOCKMETHOD instead of (LOCKMETHOD) NUL
On Mon, 01 Sep 2003 08:46:09 -0400, Tom Lane <[EMAIL PROTECTED]>
wrote:
>[EMAIL PROTECTED] writes:
>> it took 69 minutes to finish, 75% of this time was devoted to create 2
>> indexes on varchar(2) with value being 'O', 'N' or null;
>
>I still say it's either strcoll or qsort's fault.
If qsort is
On Thu, 31 Jul 2003 16:30:17 +0200, Andreas Pflug
<[EMAIL PROTECTED]> wrote:
>Well, to me it's not well-known that floating-point addition is not
>associative
This is a case of theory vs. practice mismatch: In theory addition is
associative, in practice there is only limited storage available fo
On Fri, 04 Jul 2003 15:29:37 -0400, Tom Lane <[EMAIL PROTECTED]>
wrote:
>The attached patch shows how initdb can dynamically determine reasonable
>shared_buffers and max_connections settings that will work on the
>current machine.
Can't this be done on postmaster startup? I think of two GUC
varia
34 matches
Mail list logo
