From: John A Meinel <john@arbash-meinel.com>
To: Matthew Schumacher <matt.s@aptalaska.net>
Cc: pgsql-performance@postgresql.org
Subject: Re: Performance problems testing with Spamassassin 3.1.0
Date: Thu, 04 Aug 2005 17:42:52 -0500
Message-ID: <42F299EC.8080809@arbash-meinel.com> (raw)
In-Reply-To: <42F298B5.5010405@aptalaska.net>
References: <42E9749F.6000709@aptalaska.net>
<1122616211.5691.340.camel@lamb.mcmillan.net.nz>
<42EA6966.8090402@aptalaska.net>
<1122668371.5691.352.camel@lamb.mcmillan.net.nz>
<19405.1122670917@sss.pgh.pa.us>
<42EAA410.9010700@aptalaska.net>
<27897.1122748133@sss.pgh.pa.us>
<42EE8043.1050207@aptalaska.net>
<op.suulgxokth1vuj@localhost>
<42EE93F8.5010003@aptalaska.net>
<20050801214931.GF60019@decibel.org>
<42EE9D83.8000405@pobox.com>
<42F1CE43.8040801@aptalaska.net>
<15610.1123166222@sss.pgh.pa.us>
<42F23FB6.5010405@aptalaska.net>
<42F251F0.1070407@aptalaska.net>
<42F26E2C.8010607@arbash-meinel.com>
<42F298B5.5010405@aptalaska.net>
Matthew Schumacher wrote:
> John A Meinel wrote:
>
>
>>Surely this isn't what you have. You have *no* loop here, and you have
>>stuff like:
>> AND
>> (bayes_token_tmp) NOT IN (SELECT token FROM bayes_token);
>>
>>I'm guessing this isn't your last version of the function.
>>
>>As far as putting the CREATE TEMP TABLE inside the function, I think the
>>problem is that the first time it runs, it compiles the function, and
>>when it gets to the UPDATE/INSERT with the temporary table name, at
>>compile time it hard-codes that table id.
>>
>>I tried getting around it by using "EXECUTE" which worked, but it made
>>the function horribly slow. So I don't recommend it.
>>
>>Anyway, if you want us to evaluate it, you really need to send us the
>>real final function.
>>
>>John
>>=:->
>
>
> It is the final function. It doesn't need a loop because of the
> bayes_token_tmp function I added. The array is passed to it and it
> returns a record set so I can work off of it like it's a table. So the
> function works the same way it before, but instead of using SELECT
> intoken from TEMPTABLE, you use SELECT bayes_token_tmp from
> bayes_token_tmp(intokenary).
>
> I think this is more efficient than the create table overhead,
> especially because the incoming record set won't be to big.
>
> Thanks,
>
> schu
>
>
Well, I would at least recommend that you change the "WHERE
bayes_token_tmp NOT IN (SELECT token FROM bayes_token)"
with a
"WHERE NOT EXISTS (SELECT toke FROM bayes_token WHERE
token=bayes_token_tmp)"
You might try experimenting with the differences, but on my system the
NOT IN has to do a full sequential scan on bayes_token and load all
entries into a list, while NOT EXISTS can do effectively a nested loop.
The nested loop requires that there is an index on bayes_token(token),
but I'm pretty sure there is anyway.
Again, in my testing, it was a difference of 4200ms versus 180ms. (800k
rows in my big table, 2k in the temp one)
John
=:->
Attachments:
signature.asc (application/pgp-signature, 252B) download
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: pgsql-performance@postgresql.org
Cc: john@arbash-meinel.com, matt.s@aptalaska.net
Subject: Re: Performance problems testing with Spamassassin 3.1.0
In-Reply-To: <42F299EC.8080809@arbash-meinel.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by DDX for PostgreSQL; see mirroring instructions
for how to clone and mirror all data and code used for this inbox